1 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3 PR tree-optimization/85720
4 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5 SCC if all partitions are builtins.
6 (version_loop_by_alias_check): New parameter. Generate cancelable
7 runtime alias check if all partitions are builtins.
8 (distribute_loop): Update call to above function.
10 2018-05-25 Bin Cheng <bin.cheng@arm.com>
12 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
13 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
14 (parm_default_def_partition_arg): Ditto.
15 (set_parm_default_def_partition): Ditto.
16 (get_parm_default_def_partitions): Ditto and make it static.
17 (get_undefined_value_partitions): Ditto and make it static.
18 (remove_ssa_form): Refactor call to init_var_map here.
19 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
20 computation for loop region.
21 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
22 (register_default_def): Delete.
23 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
24 (parm_default_def_partition_arg): Ditto.
25 (set_parm_default_def_partition): Ditto.
26 (get_parm_default_def_partitions): Ditto and make it static.
27 (get_undefined_value_partitions): Ditto and make it static.
28 (coalesce_with_default, coalesce_with_default): Update comment.
29 (create_coalesce_list_for_region): New func factored out from
30 create_outofssa_var_map.
31 (populate_coalesce_list_for_outofssa): New func factored out from
32 create_outofssa_var_map and coalesce_ssa_name.
33 (create_outofssa_var_map): Delete.
34 (coalesce_ssa_name): Refactor to support live range computation.
35 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
36 (get_parm_default_def_partitions): Delete.
37 (get_undefined_value_partitions): Ditto.
38 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
39 computation for loop region.
40 (new_tree_live_info, loe_visit_block): Ditto.
41 (live_worklist, set_var_live_on_entry): Ditto.
42 (calculate_live_on_exit, verify_live_on_entry): Ditto.
43 * tree-ssa-live.h (struct _var_map): New fields.
44 (init_var_map): Change decl.
45 (region_contains_p): New.
47 2018-05-25 Bin Cheng <bin.cheng@arm.com>
49 * tree-ssa-live.h (live_merge_and_clear): Delete.
51 2018-05-25 Richard Biener <rguenther@suse.de>
54 * tree-dump.c (dequeue_and_dump): Remove access to removed
55 operand 2 of a SWITCH_EXPR.
57 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
59 * doc/sourcebuild.texi (vect_double_cond_arith): Include
60 multiplication and division.
61 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
62 (cond_udiv@var{m}, cond_umod@var{m}): Document.
63 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
64 (cond_udiv_optab, cond_umod_optab): New optabs.
65 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
66 (IFN_COND_RDIV): New internal functions.
67 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
68 TRUNC_MOD_EXPR and RDIV_EXPR.
69 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
70 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
72 (SVE_INT_BINARY): Include mult.
73 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
74 (optab, sve_int_op): Handle mult.
75 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
77 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
78 for SVE_INT_BINARY_SD.
80 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
82 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
83 (optab, sve_int_op): Handle div and udiv.
84 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
85 for SVE_INT_BINARY_SD.
86 (*<optab><mode>3): New insn for the same.
88 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
90 * tree-vect-patterns.c: Include predict.h.
91 (vect_recog_divmod_pattern): Restrict check for division support
92 to when optimizing for size.
94 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
96 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
97 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
98 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
99 (gimple_match_op::set_op): Likewise.
100 (gimple_resimplify4): Declare.
101 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
102 (expr::gen_transform): Likewise.
103 (decision_tree::gen): Generate a simplification routine for 4 operands.
104 * gimple-match-head.c (gimple_simplify): Add an overload for
105 4 operands. In the top-level function, handle up to 4 call
106 arguments and call gimple_resimplify4.
107 (gimple_resimplify4): New function.
108 (build_call_internal): Pass a fourth operand.
109 (maybe_push_to_seq): Likewise.
110 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
111 Fold VEC_COND_EXPRs of an operation and a default value into
112 an IFN_COND_* function if possible.
113 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
115 (SVE_COND_FP_BINARY): Include them.
116 (optab, sve_fp_op): Handle them.
117 (SVE_INT_BINARY_REV): New code iterator.
118 (SVE_COND_FP_BINARY_REV): New int iterator.
119 (commutative): New int attribute.
120 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
122 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
124 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
125 (*cond_<optab><mode>): New patterns for reversed operands.
127 2018-05-25 Richard Biener <rguenther@suse.de>
129 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
130 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
131 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
132 (STMT_VINFO_GROUPED_ACCESS): Adjust.
133 * tree-vect-data-refs.c (everywhere): Adjust users.
134 * tree-vect-loop.c (everywhere): Likewise.
135 * tree-vect-slp.c (everywhere): Likewise.
136 * tree-vect-stmts.c (everywhere): Likewise.
137 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
139 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
141 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
143 (gcc_cv_as_section_exclude): ... this.
144 Try Solaris as #exclude syntax.
145 * configure: Regenerate.
146 * config.in: Regenerate.
147 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
149 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
150 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
152 * varasm.c (default_elf_asm_named_section): Don't check if
153 HAVE_GAS_SECTION_EXCLUDE is defined.
155 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
157 * doc/md.texi: Update the documentation of the cond_* optabs
158 to mention the new final operand. Fix GET_MODE_NUNITS call.
159 Describe the scalar case too.
160 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
161 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
163 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
164 (get_conditional_internal_fn): Update comment.
165 * tree-vect-loop.c (vectorizable_reduction): Pass the original
166 accumulator value as a final argument to conditional functions.
167 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
168 a define_expand and add an "else" operand. Assert for now that
169 the else operand is equal to operand 2. Use SVE_INT_BINARY and
170 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
171 (*cond_<optab><mode>): New patterns.
172 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
173 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
174 (UNSPEC_COND_EOR): Delete.
175 (optab): Remove associated mappings.
176 (SVE_INT_BINARY): New code iterator.
177 (sve_int_op): Remove int attribute and add "minus" to the code
179 (SVE_COND_INT_OP): Delete.
180 (SVE_COND_FP_OP): Rename to...
181 (SVE_COND_FP_BINARY): ...this.
183 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
185 * optabs.c (can_reuse_operands_p): New function.
186 (maybe_legitimize_operands): Try to reuse the results for
189 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
191 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
192 Add {q} suffix to insn mnemonic.
194 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
196 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
197 (msp430_warn_func_return): New.
199 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
201 * fold-const.c (tree_nonzero_bits): New function.
202 * fold-const.h (tree_nonzero_bits): Likewise.
203 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
204 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
206 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
210 * varasm.c (assemble_alias): Check ifunc_resolver only on
213 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
216 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
217 when memory input operand is handled.
219 2018-05-24 Luis Machado <luis.machado@linaro.org>
221 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
223 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
225 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
227 * match.pd: Delay FMA folds until after vectorization.
229 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
232 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
233 address check not strict.
235 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
237 * gimple-match.h (gimple_match_op): New class.
238 (mprts_hook): Replace parameters with a gimple_match_op *.
239 (maybe_build_generic_op): Likewise.
240 (gimple_simplified_result_is_gimple_val): Replace parameters with
241 a const gimple_match_op *.
242 (gimple_simplify): Replace code_helper * and tree * parameters with
243 a gimple_match_op * parameter.
244 (gimple_resimplify1): Replace code_helper *, tree and tree *
245 parameters with a gimple_match_op * parameter.
246 (gimple_resimplify2): Likewise.
247 (gimple_resimplify3): Likewise.
248 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
249 parameters with a gimple_match_op * parameter.
250 * gimple-match-head.c (gimple_simplify): Change prototypes of
251 auto-generated functions to take a gimple_match_op * instead of
252 separate code_helper * and tree * parameters. Make the same
253 change in the top-level overload and update calls to the
254 gimple_resimplify routines. Update calls to the auto-generated
255 functions and to maybe_push_res_to_seq in the publicly-facing
256 operation-specific gimple_simplify overloads.
257 (gimple_match_op::MAX_NUM_OPS): Define.
258 (gimple_resimplify1): Replace rcode and ops with a single res_op
259 parameter. Update call to gimple_simplify.
260 (gimple_resimplify2): Likewise.
261 (gimple_resimplify3): Likewise.
262 (mprts_hook): Replace parameters with a gimple_match_op *.
263 (maybe_build_generic_op): Likewise.
264 (build_call_internal): Replace type, nargs and ops with
266 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
267 with a single gimple_match_op *. Update calls to mprts_hook,
268 build_call_internal and gimple_simplified_result_is_gimple_val.
269 Factor out code that is common to the tree_code and combined_fn cases.
270 * genmatch.c (expr::gen_transform): Replace tem_code and
271 tem_ops with a gimple_match_op called tem_op. Update calls
272 to the gimple_resimplify functions and maybe_push_res_to_seq.
273 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
274 res_ops. Update call to the gimple_resimplify functions.
275 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
276 (decision_tree::gen): Make the functions take a gimple_match_op *
277 called res_op instead of separate res_code and res_ops parameters.
278 Update call accordingly.
279 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
280 and ops with a single res_op parameter. Update calls to
281 maybe_build_generic_op and maybe_push_res_to_seq.
282 (fold_stmt_1): Update calls to gimple_simplify and
283 replace_stmt_with_simplification.
284 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
285 and gimple_simplified_result_is_gimple_val.
286 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
288 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
289 with a gimple_match_op *.
290 (vn_nary_build_or_lookup): Likewise. Update call to
291 vn_nary_build_or_lookup_1.
292 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
293 gimple_match_op *. Update calls to the gimple_resimplify routines
294 and to gimple_simplified_result_is_gimple_val.
295 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
296 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
297 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
298 (visit_nary_op): Likewise.
299 (visit_reference_op_load): Likewise.
301 2018-05-23 Luis Machado <luis.machado@linaro.org>
303 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
304 modifier for printing the step amount.
306 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
309 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
312 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
314 * doc/sourcebuild.texi (Endianness): New subsubsection.
316 2018-05-23 Luis Machado <luis.machado@linaro.org>
318 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
319 <prefetch_dynamic_strides>: New const bool field.
320 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
321 prefetch_dynamic_strides.
322 (exynosm1_prefetch_tune): Likewise.
323 (thunderxt88_prefetch_tune): Likewise.
324 (thunderx_prefetch_tune): Likewise.
325 (thunderx2t99_prefetch_tune): Likewise.
326 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
328 (aarch64_override_options_internal): Update to set
329 PARAM_PREFETCH_DYNAMIC_STRIDES.
330 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
331 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
332 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
333 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
334 prefetch-dynamic-strides setting.
336 2018-05-23 Luis Machado <luis.machado@linaro.org>
338 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
339 <minimum_stride>: New const int field.
340 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
341 minimum_stride field defaulting to -1.
342 (exynosm1_prefetch_tune): Likewise.
343 (thunderxt88_prefetch_tune): Likewise.
344 (thunderx_prefetch_tune): Likewise.
345 (thunderx2t99_prefetch_tune): Likewise.
346 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
347 <default_opt_level>: Set to 3.
348 (aarch64_override_options_internal): Update to set
349 PARAM_PREFETCH_MINIMUM_STRIDE.
350 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
351 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
352 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
353 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
354 stride is constant and is below the minimum stride threshold.
356 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
358 * config/arm/arm-cpus.in (mode26): Delete.
359 (armv4): Delete mode26 reference.
360 * config/arm/arm.c (arm_configure_build_target): Delete use of
363 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
365 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
367 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
368 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
370 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
371 (floatunsdidf2): Ditto.
373 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
375 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
376 (fixuns_trunc<mode>si2_avx512f): Ditto.
377 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
378 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
379 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
381 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
383 PR rtl-optimization/79985
384 * df-scan.c (df_insn_refs_collect): Remove special case for
385 global registers and asm statements.
387 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
389 * extend.texi (Global Register Variables): Rewrite the bullet list.
390 Note that the register is available for allocation. Note that access
391 via inline asm must use constraints. Add note about async-signal
392 handlers. Remove paragraph about automagic register selection.
394 2018-05-23 Richard Biener <rguenther@suse.de>
396 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
397 of fixed offset from memset VN.
399 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
401 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
403 (alloc_cand_and_find_basis): Initialize first_interp field.
404 (slsr_process_mul): Modify first_interp field.
405 (slsr_process_add): Likewise.
406 (slsr_process_cast): Modify first_interp field for each new
408 (slsr_process_copy): Likewise.
409 (dump_candidate): Dump first_interp field.
410 (replace_mult_candidate): Process all interpretations, not just
412 (replace_rhs_if_not_dup): Likewise.
413 (replace_one_candidate): Likewise.
415 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
417 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
419 (aarch64_needs_frame_chain): New function.
420 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
422 2018-05-23 Sudakshina Das <sudi.das@arm.com>
425 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
426 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
427 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
428 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
429 as true for strict-align.
430 (aarch64_can_inline_p): Perform checks even when callee has no
431 attributes to check for strict alignment.
432 * doc/extend.texi (AArch64 Function Attributes): Document
434 * doc/invoke.texi: (AArch64 Options): Likewise.
436 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
438 PR tree-optimization/85853
439 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
440 the handling of the root of the node to...
441 (vect_slp_analyze_node_operations_1): ...this new function,
442 and run the whole thing with the child nodes' def types
443 set according to their SLP node's def type.
445 2018-05-23 Richard Biener <rguenther@suse.de>
448 * tree-data-ref.c (create_runtime_alias_checks): Defer
449 and ignore overflow warnings.
451 2018-05-23 Yury Gribov <tetra2005@gmail.com>
453 PR tree-optimization/85822
454 * tree-vrp.c (is_masked_range_test): Fix handling of negative
457 2018-05-23 Richard Biener <rguenther@suse.de>
459 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
460 memset constants via native_interpret_expr.
462 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
465 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
467 (cgraph_node::create_alias): Likewise.
468 (cgraph_node::get_availability): Check ifunc_resolver instead
469 of looking up ifunc attribute.
470 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
471 * varasm.c (do_assemble_alias): Likewise.
472 (assemble_alias): Likewise.
473 (default_binds_local_p_3): Likewise.
474 * cgraph.h (cgraph_node): Add ifunc_resolver.
475 (cgraph_node::only_called_directly_or_aliased_p): Return false
477 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
479 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
480 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
481 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
482 instead of looking up ifunc attribute.
484 2018-05-22 Luis Machado <luis.machado@linaro.org>
486 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
488 2018-05-22 Martin Sebor <msebor@redhat.com>
491 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
492 only when expasion succeeds.
493 (expand_builtin_strcmp): Same.
494 (expand_builtin_strncmp): Same.
496 2018-05-22 Martin Sebor <msebor@redhat.com>
498 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
500 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
501 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
503 * config/aarch64/aarch64-ldpstp.md: Replace uses of
504 aarch64_mem_pair_operand with memory_operand and delete operand swapping
506 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
507 Add check for legitimate_address.
508 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
509 (aarch64_swap_ldrstr_operands): New.
510 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
513 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
514 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
516 * config/aarch64/aarch64.md: New patterns to generate stp
518 (store_pair_sw, store_pair_dw): New patterns to generate stp for
519 single words and double words.
520 (load_pair_sw, load_pair_dw): Likewise.
521 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
523 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
525 * config/aarch64/aarch64-ldpstp.md: Modify peephole
526 for different mode ldpstp and add peephole for merged zero stores.
528 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
530 (aarch64_gen_store_pair): Rename calls to match new patterns.
531 (aarch64_gen_load_pair): Rename calls to match new patterns.
532 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
533 (load_pair<DREG:mode><DREG2:mode>): ... This.
534 (store_pair<mode>): Rename to...
535 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
536 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
538 (V_INT_EQUIV): Handle SImode.
539 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
542 2018-05-22 Martin Sebor <msebor@redhat.com>
545 * calls.c (maybe_warn_nonstring_arg): Use string length to set
546 or ajust the presumed bound on an operation to avoid unnecessary
549 2018-05-22 Martin Sebor <msebor@redhat.com>
551 PR tree-optimization/85826
552 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
553 assuming that a DECL necesarily has a constant size.
555 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
558 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
560 2018-05-22 Richard Biener <rguenther@suse.de>
562 PR tree-optimization/85834
563 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
564 non-constant and non-zero memset arguments.
566 2018-05-22 Martin Liska <mliska@suse.cz>
569 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
571 2018-05-22 Richard Biener <rguenther@suse.de>
573 PR tree-optimization/85863
574 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
575 comparisons when vectype is specified.
576 (vectorizable_condition): Do not specify vectype for
577 vect_is_simple_cond when SLP vectorizing.
579 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
582 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
583 define __ibm128 as long double.
584 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
585 as a distinct type when IEEE 128-bit support is enabled.
586 (init_float128_ieee): Fix up conversions between IFmode and IEEE
587 128-bit types to use the correct functions.
588 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
589 convert between 128-bit floating point types that have different
590 modes but the same representation, instead of using gen_lowpart to
592 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
594 (IFKF_reg): New attributes to give the register constraints for
596 (extend<mode>tf2_internal): New insns to mark an explicit
597 conversion between 128-bit floating point types that have a
598 different mode but share the same representation.
600 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
602 PR tree-optimization/85814
603 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
604 a null return from get_strinfo when unsharing the next
605 strinfo in the chain.
607 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
610 * varasm.c (weak_finish): Clean up weak_decls.
612 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
614 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
615 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
616 UNSPEC_UADALP values.
617 * config/aarch64/iterators.md (ABAL): New int iterator.
620 (sur): Add mappings for the above.
621 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
623 (aarch64_<sur>abal<mode>_4): Likewise.
624 (aarch64_<sur>adalp<mode>_3): Likewise.
625 (<sur>sadv16qi): New define_expand.
627 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
629 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
630 (*movdf_internal): Ditto.
631 (*rcpsf2_sse): Ditto.
632 (*rsqrtsf2_sse): Ditto.
633 (*sqrt<mode>2_sse): Ditto.
635 2018-05-21 Tamar Christina <tamar.christina@arm.com>
637 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
639 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
640 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
641 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
642 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
644 * config/aarch64/arm_neon.h: Likewise.
645 * config/aarch64/iterators.md (VQ_I): New.
647 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
649 * config.gcc: Add arc/t-multilib-linux to tmake_file for
651 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
654 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
656 * config/nds32/constraints.md (S): New constraint.
657 * config/nds32/nds32.md (call_internal): Use constraint S.
658 (call_value_internal): Likewise.
659 (sibcall_internal): Likewise.
660 (sibcall_value_internal): Likewise.
662 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
663 Chung-Ju Wu <jasonwucj@gmail.com>
665 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
668 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
669 Chung-Ju Wu <jasonwucj@gmail.com>
671 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
672 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
673 (nds32_rtx_costs_impl): Simplify.
674 (nds32_address_cost_impl): Simplify.
675 (nds32_init_rtx_costs): New function.
676 (nds32_rtx_costs_speed_prefer): Likewise.
677 (nds32_rtx_costs_size_prefer): Likewise.
678 (nds32_address_cost_speed_prefer): Likewise.
679 (nds32_address_cost_speed_fwprop): Likewise.
680 (nds32_address_cost_size_prefer): Likewise.
681 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
682 * config/nds32/nds32.c (nds32_option_override): Use
683 nds32_init_rtx_costs function.
685 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
687 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
688 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
689 (TARGET_PIPELINE_N8): Likewise.
690 (TARGET_PIPELINE_N10): Likewise.
691 (TARGET_PIPELINE_N13): Likewise.
692 (TARGET_PIPELINE_GRAYWOLF): Likewise.
694 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
696 * config/nds32/nds32-fpu.md: Update copyright year.
698 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
700 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
702 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
704 * config/nds32/nds32.c
705 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
706 * config/nds32/nds32.opt (minline-asm-r15): New option.
708 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
710 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
712 * config/nds32/nds32.md (abssi2): New pattern.
714 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
716 * config/i386/i386.md (rex64namesuffix): New mode attribute.
717 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
718 Merge insn pattern from sse_cvtsi2ss<round_name> and
719 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
720 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
721 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
722 using SWI48 mode iterator.
723 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
724 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
725 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
726 pattern from sse_cvttss2si<round_saeonly_name>
727 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
728 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
729 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
730 using SWI48 mode iterator.
731 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
732 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
733 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
734 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
735 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
736 using SWI48 mode iterator.
737 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
738 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
739 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
740 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
741 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
743 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
744 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
745 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
746 pattern from sse_cvttsd2si<round_saeonly_name>
747 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
749 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
751 * config/nds32/nds32-md-auxiliary.c
752 (nds32_valid_smw_lwm_base_p): Refine.
753 (nds32_output_smw_single_word): Refine.
754 (nds32_output_smw_double_word): New.
755 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
757 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
759 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
760 (nds32_output_stack_pop): Refine.
761 (nds32_expand_unaligned_load): Refine.
762 (nds32_expand_unaligned_store): Refine.
764 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
765 Chung-Ju Wu <jasonwucj@gmail.com>
767 * config/nds32/constants.md: Add TP_REGNUM constant.
768 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
769 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
771 * config/nds32/nds32-doubleword.md: Consider flag_pic.
772 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
773 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
774 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
775 and PIC code generation.
776 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
778 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
780 * config/nds32/nds32.md: Support TLS and PIC.
781 * config/nds32/nds32.c: Support TLS and PIC.
782 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
783 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
786 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
788 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
791 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
792 Chung-Ju Wu <jasonwucj@gmail.com>
794 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
795 * config/nds32/nds32-md-auxiliary.c
796 (symbolic_reference_mentioned_p): New.
797 (nds32_legitimize_ict_address): New.
798 (nds32_expand_ict_move): New.
799 (nds32_indirect_call_referenced_p): New.
800 (nds32_symbol_binds_local_p): Delete.
801 (nds32_long_call_p): Modify.
802 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
803 * config/nds32/nds32-protos.h
804 (symbolic_reference_mentioned_p): Declare.
805 (nds32_legitimize_ict_address): Declare.
806 (nds32_expand_ict_move): Declare.
807 (nds32_indirect_call_referenced_p): Declare.
808 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
809 (nds32_relax_group): Use nds32_ict_const_p as condition.
810 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
811 (nds32_asm_file_start): Output ict_model directive in asm code.
812 (nds32_legitimate_address_p): Consider indirect call.
813 (nds32_print_operand): Consider indirect call.
814 (nds32_print_operand_address): Consider indirect call.
815 (nds32_insert_attributes): Handle "indirect_call" attribute.
816 (TARGET_LEGITIMATE_ADDRESS_P): Define.
817 (TARGET_LEGITIMATE_CONSTANT_P): Define.
818 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
819 (TARGET_DELEGITIMIZE_ADDRESS): Define.
820 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
821 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
822 (TARGET_ICT_MODEL_SMALL): Define.
823 (TARGET_ICT_MODEL_LARGE): Define.
824 * config/nds32/nds32.md (movsi): Consider ict model.
825 (call, call_value): Consider ict model.
826 (sibcall, sibcall_value): Consider ict model.
827 * config/nds32/nds32.opt (mict-model): New option.
828 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
831 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
832 Monk Chiang <sh.chiang04@gmail.com>
833 Jim Wilson <jimw@sifive.com>
835 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
836 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
837 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
838 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
839 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
840 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
841 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
842 compute save_libcall_adjustment properly.
843 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
844 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
845 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
846 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
847 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
848 (ABI_SPEC): Handle mabi=ilp32e.
849 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
851 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
852 <-march>: Add rv32e as an example.
854 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
857 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
858 (intra_create_variable_infos): Handle C++ constructors.
860 2018-05-18 Martin Liska <mliska@suse.cz>
862 * passes.def: Remove a redundant pass.
864 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
867 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
869 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
871 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
873 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
875 (armv2, armv2a, armv3, armv3m): Delete architectures.
876 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
877 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
878 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
880 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
881 (*mulsidi3adddi): Likewise.
882 (mulsidi3): Likewise.
883 (*mulsidi3_nov6): Likewise.
884 (umulsidi3): Likewise.
885 (umulsidi3_nov6): Likewise.
886 (umaddsidi4): Likewise.
887 (*umulsidi3adddi): Likewise.
888 (smulsi3_highpart): Likewise.
889 (*smulsi3_highpart_nov6): Likewise.
890 (umulsi3_highpart): Likewise.
891 (*umulsi3_highpart_nov6): Likewise.
892 * config/arm/arm.h (arm_arch3m): Delete.
893 * config/arm/arm.c (arm_arch3m): Delete.
894 (arm_option_override_internal): Update armv3-related comment.
895 (arm_configure_build_target): Delete use of isa_bit_mode32.
896 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
897 (arm_rtx_costs_internal): Delete check of arm_arch3m.
898 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
901 * config/arm/arm-protos.h (arm_arch3m): Delete.
902 * config/arm/arm-tables.opt: Regenerate.
903 * config/arm/arm-tune.md: Likewise.
904 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
905 deleted architectures.
907 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
909 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
910 (armv5t, armv5te): New features.
911 (ARMv5, ARMv5e): Delete fgroups.
912 (ARMv5t, ARMv5te): Adjust for above changes.
914 (armv5, armv5e): Delete arches.
915 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
917 (*call_reg_arm): Likewise.
918 (*call_value_reg_armv5): Likewise.
919 (*call_value_reg_arm): Likewise.
920 (*call_symbol): Likewise.
921 (*call_value_symbol): Likewise.
922 (*sibcall_insn): Likewise.
923 (*sibcall_value_insn): Likewise.
925 (prefetch): Likewise.
926 (define_split and define_peephole2 dependent on arm_arch5):
928 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
930 (TARGET_ARM_QBIT): Likewise.
931 (TARGET_DSP_MULTIPLY): Likewise.
932 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
933 (arm_arch5, arm_arch5e): Delete.
934 (arm_arch5t, arm_arch5te): Declare.
935 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
936 (arm_arch5t): Declare.
937 (arm_option_reconfigure_globals): Update for the above.
938 (arm_options_perform_arch_sanity_checks): Update comment, replace
939 use of arm_arch5 with arm_arch5t.
940 (use_return_insn): Likewise.
941 (arm_emit_call_insn): Likewise.
942 (output_return_instruction): Likewise.
943 (arm_final_prescan_insn): Likewise.
944 (arm_coproc_builtin_available): Likewise.
945 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
946 arm_arch5e with arm_arch5t and arm_arch5te.
947 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
948 (arm_arch5t, arm_arch5te): Declare.
949 * config/arm/arm-tables.opt: Regenerate.
950 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
951 * config/arm/t-multilib: Likewise.
952 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
953 instead of arm_arch5.
954 (*call_reg_thumb1): Likewise.
955 (*call_value_reg_thumb1_v5): Likewise.
956 (*call_value_reg_thumb1): Likewise.
957 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
959 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
961 2018-05-18 Martin Liska <mliska@suse.cz>
963 PR gcov-profile/84846
964 * doc/gcov.texi: Document -t option of gcov tool.
966 2018-05-18 Martin Liska <mliska@suse.cz>
968 PR gcov-profile/84846
969 * gcov.c (print_usage): Add new -t option.
970 (process_args): Handle the option.
971 (generate_results): Use stdout as output when requested by
974 2018-05-18 Martin Liska <mliska@suse.cz>
976 PR gcov-profile/84846
977 * coverage.c (coverage_init): Write PWD to .gcno file.
978 * doc/gcov.texi: Document how working directory is printed.
979 * gcov-dump.c (dump_gcov_file): Print PWD.
980 * gcov.c (output_intermediate_file): Likewise.
981 (read_graph_file): Read PWD string.
982 (output_lines): Print PWD.
984 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
987 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
988 for retval and return false if all args to phi are zero.
990 2018-05-18 Richard Biener <rguenther@suse.de>
992 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
994 (evrp_dom_walker::before_dom_children): Call it.
996 2018-05-18 Richard Biener <rguenther@suse.de>
998 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
999 results when processing array refs with variable index.
1001 2018-05-18 Toon Moene <toon@moene.org>
1003 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1004 directly after that of -floop-interchange. Indicate that both
1005 options are enabled by default when specifying -O3.
1007 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1009 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1010 iterator. Delete separate integer-mode vec_set<mode> expander.
1011 (aarch64_simd_vec_setv2di): Delete.
1012 (vec_setv2di): Delete.
1013 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1014 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1015 the "w, r" alternative.
1017 2018-05-18 Martin Liska <mliska@suse.cz>
1019 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1020 * tree-pass.h (make_pass_lower_switch_O0): New function.
1021 * tree-switch-conversion.c (node_has_low_bound): Remove.
1022 (node_has_high_bound): Likewise.
1023 (node_is_bounded): Likewise.
1024 (class pass_lower_switch): Make it a template type and create
1026 (pass_lower_switch::execute): Add template argument.
1027 (make_pass_lower_switch): New function.
1028 (make_pass_lower_switch_O0): New function.
1029 (do_jump_if_equal): Remove.
1030 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1031 all the hard work to tree optimization passes.
1033 2018-05-18 Martin Liska <mliska@suse.cz>
1035 * dbgcnt.c (limit_low): Renamed from limit.
1036 (limit_high): New variable.
1037 (dbg_cnt_is_enabled): Check for upper limit.
1038 (dbg_cnt): Adjust dumping.
1039 (dbg_cnt_set_limit_by_index): Add new argument for high
1041 (dbg_cnt_set_limit_by_name): Likewise.
1042 (dbg_cnt_process_single_pair): Parse new format.
1043 (dbg_cnt_process_opt): Use strtok.
1044 (dbg_cnt_list_all_counters): Remove 'value' and add
1046 * doc/invoke.texi: Document changes.
1048 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1050 * doc/sourcebuild.texi (scalar_all_fma): Document.
1051 * tree.def (FMA_EXPR): Delete.
1052 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1053 * internal-fn.c (ternary_direct): New macro.
1054 (expand_ternary_optab_fn): Likewise.
1055 (direct_ternary_optab_supported_p): Likewise.
1056 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1057 * builtins.c (fold_builtin_fma): Delete.
1058 (fold_builtin_3): Don't call it.
1059 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1060 * expr.c (expand_expr_real_2): Likewise.
1061 * fold-const.c (operand_equal_p): Likewise.
1062 (fold_ternary_loc): Likewise.
1063 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1064 * gimple.c (DEFTREECODE): Likewise.
1065 * gimplify.c (gimplify_expr): Likewise.
1066 * optabs-tree.c (optab_for_tree_code): Likewise.
1067 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1068 * tree-eh.c (operation_could_trap_p): Likewise.
1069 (stmt_could_throw_1_p): Likewise.
1070 * tree-inline.c (estimate_operator_cost): Likewise.
1071 * tree-pretty-print.c (dump_generic_node): Likewise.
1072 (op_code_prio): Likewise.
1073 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1074 * tree-ssa-operands.c (get_expr_operands): Likewise.
1075 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1076 * fold-const-call.h (fold_fma): Delete.
1077 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1078 CFN_FNMA and CFN_FNMS.
1080 * genmatch.c (combined_fn): New enum.
1081 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1082 (commutative_op): New function.
1083 (commutate): Use it. Handle more than 2 operands.
1084 (dt_operand::gen_gimple_expr): Use commutative_op.
1085 (parser::parse_expr): Allow :c to be used with non-binary
1086 operators if the commutative operand is known.
1087 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1088 CFN_FMS, CFN_FNMA and CFN_FNMS.
1089 (backprop::process_assign_use): Remove FMA_EXPR handling.
1090 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1091 (gen_hsa_fma): New function.
1092 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1093 IFN_FNMA and IFN_FNMS.
1094 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1095 * gimple-fold.h (follow_all_ssa_edges): Declare.
1096 * gimple-fold.c (follow_all_ssa_edges): New function.
1097 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1098 gimple_build interface and use follow_all_ssa_edges to fold the result.
1099 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1100 instead of checking for optabs directly.
1101 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1102 rather than FMA_EXPRs.
1103 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1104 call to IFN_FMA instead of an FMA_EXPR.
1106 2018-05-17 Jim Wilson <jimw@sifive.com>
1108 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1109 sign extended promoted subreg, and the range does not have the sign bit
1110 set, then do a sign extend.
1112 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1113 test, check for sign extended subreg and/or constant operands, and
1114 do a sign extend in that case.
1116 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1118 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1119 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1121 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1122 Change logics_shift_reg to logics_shift_imm.
1123 (thunderx2t99_fp_loadpair_basic): Delete.
1124 (thunderx2t99_fp_storepair_basic): Delete.
1125 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1126 (thunderx2t99_asimd_polynomial): Delete.
1127 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1128 and neon_fp_mul_d_scalar_q.
1129 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1130 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1131 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1132 (thunderx2t99_asimd_lut): Add missing tbl types.
1133 (thunderx2t99_asimd_ext): Delete.
1134 (thunderx2t99_asimd_load1_1_mult): Delete.
1135 (thunderx2t99_asimd_load1_2_mult): Delete.
1136 (thunderx2t99_asimd_load1_ldp): New.
1137 (thunderx2t99_asimd_load1): New.
1138 (thunderx2t99_asimd_load2): Add missing *load2* types.
1139 (thunderx2t99_asimd_load3): New.
1140 (thunderx2t99_asimd_load4): New.
1141 (thunderx2t99_asimd_store1_1_mult): Delete.
1142 (thunderx2t99_asimd_store1_2_mult): Delete.
1143 (thunderx2t99_asimd_store2_mult): Delete.
1144 (thunderx2t99_asimd_store2_onelane): Delete.
1145 (thunderx2t99_asimd_store_stp): New.
1146 (thunderx2t99_asimd_store1): New.
1147 (thunderx2t99_asimd_store2): New.
1148 (thunderx2t99_asimd_store3): New.
1149 (thunderx2t99_asimd_store4): New.
1151 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1153 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1154 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1156 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1157 Segher Boessenkool <segher@kernel.crashing.org>
1160 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1163 2018-05-17 Richard Biener <rguenther@suse.de>
1165 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1166 for pruning loop and prune defs feeding only already visited PHIs.
1168 2018-05-17 Richard Biener <rguenther@suse.de>
1170 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1172 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1173 Richard Biener <rguenther@suse.de>
1175 PR tree-optimization/85793
1176 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1177 for VMAT_ELEMENTWISE.
1179 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1181 * internal-fn.h (lookup_internal_fn): Declare
1182 * internal-fn.c (lookup_internal_fn): New function.
1183 * gimple.c (gimple_build_call_from_tree): Handle calls to
1185 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1186 internal function names.
1187 * tree-pretty-print.c (dump_generic_node): Likewise.
1188 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1190 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1192 * gimple-fold.h (gimple_build): Make the function forms take
1193 combined_fn rather than built_in_function.
1194 (gimple_simplify): Likewise.
1195 * gimple-match-head.c (gimple_simplify): Likewise.
1196 * gimple-fold.c (gimple_build): Likewise.
1197 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1198 rather than gimple_build_call_internal.
1199 (get_initial_defs_for_reduction): Likewise.
1200 (vect_create_epilog_for_reduction): Likewise.
1201 (vectorizable_live_operation): Likewise.
1203 2018-05-17 Martin Liska <mliska@suse.cz>
1205 * gimple-ssa-sprintf.c (format_directive): Do not use
1206 space in between 'G_' and '('.
1208 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1211 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1212 even if the mask is not all ones.
1215 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1217 (ix86_gimple_fold_builtin): Likewise.
1220 * config/i386/i386.c: Include tree-vector-builder.h.
1221 (ix86_vector_shift_count): New function.
1222 (ix86_fold_builtin): Fold shift builtins by scalar count.
1223 (ix86_gimple_fold_builtin): Likewise.
1225 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1226 _mm512_setzero): New intrinsics.
1228 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1229 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1231 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1232 code generation for cases where splatting a value is not useful.
1233 * simplify-rtx.c (simplify_ternary_operation): Simplify
1234 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1235 mode to a vec_concat.
1237 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1239 * config.gcc: Support "goldmont-plus".
1240 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1242 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1243 PROCESSOR_GOLDMONT_PLUS.
1244 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1245 (processor_target_table): Add "goldmont-plus".
1246 (PTA_GOLDMONT_PLUS): Define.
1247 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1248 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1249 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1250 (fold_builtin_cpu): Add "goldmont-plus".
1251 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1252 (ix86_option_override_internal): Add "goldmont-plus".
1253 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1254 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1255 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1256 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1258 2018-05-17 Richard Biener <rguenther@suse.de>
1260 PR tree-optimization/85757
1261 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1262 remove defs that only feed that PHI from further processing.
1264 2018-05-16 Jim Wilson <jimw@sifive.com>
1266 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1268 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1270 2018-05-16 Mark Wielaard <mark@klomp.org>
1272 * dwarf2out.c (count_index_strings): New function.
1273 (output_indirect_strings): Call count_index_strings and generate
1274 header for dwarf_version >= 5.
1276 2018-05-16 Mark Wielaard <mark@klomp.org>
1278 * dwarf2out.c (dwarf_FORM): New function.
1279 (set_indirect_string): Use dwarf_FORM.
1280 (reset_indirect_string): Likewise.
1281 (size_of_die): Likewise.
1282 (value_format): Likewise.
1283 (output_die): Likewise.
1284 (add_skeleton_AT_string): Likewise.
1285 (output_macinfo_op): Likewise.
1286 (index_string): Likewise.
1287 (output_index_string_offset): Likewise.
1288 (output_index_string): Likewise.
1289 (count_index_strings): Likewise.
1291 2018-05-16 Carl Love <cel@us.ibm.com>
1293 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1294 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1296 2018-05-16 Martin Jambor <mjambor@suse.cz>
1298 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1299 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1301 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1303 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1304 (fnma<mode>4): Likewise.
1305 (fms<mode>4): Likewise.
1306 (fnms<mode>4): Likewise.
1307 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1308 (aarch64_fnma<mode>4): Likewise.
1309 (aarch64_fms<mode>4): Likewise.
1310 (aarch64_fnms<mode>4): Likewise.
1311 (aarch64_fnmadd<mode>4): Likewise.
1313 2018-05-16 Jason Merrill <jason@redhat.com>
1315 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1317 2018-05-16 Richard Biener <rguenther@suse.de>
1319 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1320 (dump_stmt_cost): Declare.
1321 (add_stmt_cost): Dump cost we add.
1322 (add_stmt_costs): New function.
1323 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1325 (vect_analyze_stmt): Adjust prototype.
1326 (vectorizable_condition): Likewise.
1327 (vectorizable_live_operation): Likewise.
1328 (vectorizable_reduction): Likewise.
1329 (vectorizable_induction): Likewise.
1330 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1331 cost vector to pass to vectorizable_ and record afterwards.
1332 (vect_model_reduction_cost): Take cost vector argument and adjust.
1333 (vect_model_induction_cost): Likewise.
1334 (vectorizable_reduction): Likewise.
1335 (vectorizable_induction): Likewise.
1336 (vectorizable_live_operation): Likewise.
1337 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1338 SLP_TREE_NUMBER_OF_VEC_STMTS.
1339 (vect_analyze_slp_cost_1): Remove.
1340 (vect_analyze_slp_cost): Likewise.
1341 (vect_slp_analyze_node_operations): Take visited args and
1342 a target cost vector. Avoid processing already visited stmt sets.
1343 (vect_slp_analyze_operations): Use a local cost vector to gather
1344 costs and register those of non-discarded instances.
1345 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1346 (vect_schedule_slp_instance): Remove copying of
1347 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1349 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1350 adding cost. Record cost entry location.
1351 (vect_prologue_cost_for_slp_op): Function to compute cost of
1352 a constant or invariant generated for SLP vect in the prologue,
1353 split out from vect_analyze_slp_cost_1.
1354 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1355 (vect_model_promotion_demotion_cost): Likewise.
1356 (vect_model_store_cost): Likewise, make static.
1357 (vect_model_load_cost): Likewise.
1358 (vectorizable_bswap): Add cost vector arg and adjust.
1359 (vectorizable_call): Likewise.
1360 (vectorizable_simd_clone_call): Likewise.
1361 (vectorizable_conversion): Likewise.
1362 (vectorizable_assignment): Likewise.
1363 (vectorizable_shift): Likewise.
1364 (vectorizable_operation): Likewise.
1365 (vectorizable_store): Likewise.
1366 (vectorizable_load): Likewise.
1367 (vectorizable_condition): Likewise.
1368 (vectorizable_comparison): Likewise.
1369 (can_vectorize_live_stmts): Likewise.
1370 (vect_analyze_stmt): Likewise.
1371 (vect_transform_stmt): Adjust calls to vectorizable_*.
1372 * tree-vectorizer.c: Include gimple-pretty-print.h.
1373 (dump_stmt_cost): New function.
1375 2018-05-16 Richard Biener <rguenther@suse.de>
1377 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1378 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1379 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1380 (check_name): New callback.
1381 (dse_classify_store): Track cycles via a visited bitmap of PHI
1382 defs and simplify handling of in-loop and across loop dead stores
1383 and properly fail for loop-variant refs. Handle byte-tracking with
1384 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1387 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1389 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1390 (vect_get_mask_type_for_stmt): Likewise.
1391 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1393 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1394 to determine the statement's vector type and the vector type that
1395 should be used for calculating nunits. Deal with cases in which
1396 the type has to be deferred.
1397 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1398 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1399 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1400 (vect_determine_vf_for_stmt): New functions, split out from...
1401 (vect_determine_vectorization_factor): ...here.
1402 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1403 (vect_get_mask_type_for_stmt): New functions, split out from
1404 vect_determine_vectorization_factor.
1406 2018-05-16 Richard Biener <rguenther@suse.de>
1408 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1409 verify the [VEC_]COND_EXPR embedded comparison.
1411 2018-05-15 Martin Sebor <msebor@redhat.com>
1413 PR tree-optimization/85753
1414 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1415 RECORD_TYPE in addition to ARRAY_TYPE.
1417 2018-05-15 Martin Sebor <msebor@redhat.com>
1420 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1422 2018-05-15 Richard Biener <rguenther@suse.de>
1424 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1425 add by_clobber_p one. Change algorithm to collect all defs
1426 representing uses we need to walk and try reducing them to
1427 a single one before failing.
1428 (dse_dom_walker::dse_optimize_stmt): Adjust.
1430 2018-05-13 Mark Wielaard <mark@klomp.org>
1432 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1433 (size_of_loc_descr): Likewise.
1434 (output_loc_operands): Likewise.
1435 (output_loc_operands_raw): Likewise.
1436 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1437 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1438 (hash_loc_operands): Likewise.
1439 (compare_loc_operands): Likewise.
1441 2018-05-14 Mark Wielaard <mark@klomp.org>
1443 * dwarf2out.c (count_index_addrs): New function.
1444 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1446 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1448 PR tree-optimization/83648
1449 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1450 return value as malloc candidate.
1452 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1455 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1456 as true in call to suggest_attribute.
1458 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1460 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1461 -mreadonly-in-sdata.
1463 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1465 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1467 (aarch64_crypto_aesd_fused): Likewise.
1469 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1471 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1472 (movsi_aarch64): Likewise.
1473 (load_pairsi): Likewise.
1474 (load_pairdi): Likewise.
1475 (store_pairsi): Likewise.
1476 (store_pairdi): Likewise.
1477 (load_pairsf): Likewise.
1478 (load_pairdf): Likewise.
1479 (store_pairsf): Likewise.
1480 (store_pairdf): Likewise.
1481 (zero_extend): Likewise.
1482 (trunc): Swap alternatives.
1483 (fcvt_target): Add '?' to prefer w over r.
1485 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1488 * config/i386/i386.md: Disallow non-commutative arithmetics in
1489 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1490 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1491 in the peephole2 before it.
1493 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1495 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1496 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1497 (ix86_handle_option): Handle -mcldemote.
1498 * config.gcc: New header.
1499 * config/i386/cldemoteintrin.h: New file.
1500 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1501 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1503 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1504 OPTION_MASK_ISA_CLDEMOTE.
1505 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1506 (ix86_valid_target_attribute_inner_p): Ditto.
1507 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1508 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1509 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1510 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1511 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1513 * config/i386/i386.opt: Add -mcldemote.
1514 * config/i386/x86intrin.h: New header.
1515 * doc/invoke.texi: Add -mcldemote.
1517 2018-05-14 Richard Biener <rguenther@suse.de>
1519 * doc/match-and-simplify.texi: Adjust :s documentation.
1521 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1523 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1524 intended memcpy size.
1525 (REORDER_45): Likewise.
1527 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1529 * sort.cc: New file.
1530 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1531 * vec.c (qsort_chk): Use gcc_qsort.
1532 * Makefile.in (OBJS-libcommon): Add sort.o.
1533 (build/sort.o): New target. Use it...
1534 (BUILD_RTL): ... here, and...
1535 (build/gencfn-macros): ... here, and...
1536 (build/genmatch): ... here.
1538 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1539 Chung-Ju Wu <jasonwucj@gmail.com>
1541 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1542 * config/nds32/nds32-graywolf.md: New file.
1543 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1544 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1546 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1547 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1548 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1549 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1550 * config/nds32/pipelines.md: Include n15 settings.
1552 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1553 Chung-Ju Wu <jasonwucj@gmail.com>
1555 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1556 * config/nds32/nds32-n13.md: New file.
1557 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1558 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1560 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1561 * config/nds32/nds32.md (pipeline_model): Add n13.
1562 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1563 * config/nds32/pipelines.md: Include n13 settings.
1565 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1566 Chung-Ju Wu <jasonwucj@gmail.com>
1568 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1569 * config/nds32/nds32-n10.md: New file.
1570 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1571 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1573 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1574 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1575 * config/nds32/nds32.md (pipeline_model): Add n10.
1576 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1577 * config/nds32/pipelines.md: Include n10 settings.
1579 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1580 Kito Cheng <kito.cheng@gmail.com>
1581 Chung-Ju Wu <jasonwucj@gmail.com>
1583 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1584 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1585 Add enum values for DSP extension instructions.
1586 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1588 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1589 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1591 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1592 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1593 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1594 * config/nds32/nds32-intrinsic.md: Likewise.
1595 * config/nds32/nds32_intrinsic.h: Likewise.
1596 * config/nds32/nds32-md-auxiliary.c: Likewise.
1597 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1598 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1599 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1600 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1601 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1602 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1603 TYPE_DMAC in switch statement.
1604 * config/nds32/nds32.c: New checking and implementation for DSP
1605 extension instructions.
1606 * config/nds32/nds32.h: Likewise.
1607 * config/nds32/nds32.md: Likewise.
1608 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1609 * config/nds32/predicates.md: Implement new predicates for DSP
1612 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1614 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1615 Reformat alternatives and attributes so it is easier to identify
1616 which constraints/attributes go with which instruction.
1617 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1618 (mov<mode>_softfloat32, FMOVE64): Likewise.
1619 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1620 (mov<mode>_softfloat64, FMOVE64): Likewise.
1622 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1624 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1626 (Basic PowerPC Built-in Functions): The new name of the
1627 subsection previously known as "PowerPC Built-in Functions".
1628 (Basic PowerPC Built-in Functions Available on all Configurations):
1630 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1631 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1632 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1633 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1635 2018-05-11 Martin Jambor <mjambor@suse.cz>
1638 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1641 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1644 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1646 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1648 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1649 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1650 (ix86_handle_option): Handle -mwaitpkg.
1651 * config.gcc: New header.
1652 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1653 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1654 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1656 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1657 OPTION_MASK_ISA_WAITPKG.
1658 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1659 (ix86_option_override_internal): Add PTA_WAITPKG.
1660 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1661 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1662 IX86_BUILTIN_TPAUSE.
1663 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1664 __builtin_ia32_umwait and __builtin_ia32_tpause.
1665 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1666 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1667 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1668 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1669 UNSPECV_TPAUSE): New.
1670 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1671 * config/i386/i386.opt: Add -mwaitpkg.
1672 * config/i386/waitpkgintrin.h: New file.
1673 * config/i386/x86intrin.h: New header.
1674 * doc/invoke.texi: Add -mwaitpkg.
1676 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1679 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1681 (cortex-m0): Use armv6s-m isa.
1682 (cortex-m0plus): Likewise.
1683 (cortex-m1): Likewise.
1684 (cortex-m0.small-multiply): Likewise.
1685 (cortex-m0plus.small-multiply): Likewise.
1686 (cortex-m1.small-multiply): Likewise.
1688 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1689 Jakub Jelinek <jakub@redhat.com>
1691 PR tree-optimization/85692
1692 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1693 source permute as well.
1695 2018-05-11 Martin Liska <mliska@suse.cz>
1698 * doc/extend.texi: Document LLVM style format for no_sanitize
1701 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1703 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1704 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1705 (mode_supports_vsx_dform_quad): Likewise.
1706 (mode_supports_vmx_dform): Move these functions to be next to the
1707 other mode_supports functions.
1708 (mode_supports_dq_form): Likewise.
1709 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1710 mode_supports_dq_form.
1711 (reg_offset_addressing_ok_p): Likewise.
1712 (offsettable_ok_by_alignment): Likewise.
1713 (rs6000_legitimate_offset_address_p): Likewise.
1714 (legitimate_lo_sum_address_p): Likewise.
1715 (rs6000_legitimize_address): Likewise.
1716 (rs6000_legitimize_reload_address): Likewise.
1717 (rs6000_secondary_reload_inner): Likewise.
1718 (rs6000_preferred_reload_class): Likewise.
1719 (rs6000_output_move_128bit): Likewise.
1721 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1723 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1724 Generate SImode target register for null target.
1725 <case IX86_BUILTIN_XGETBV>: Ditto.
1726 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1727 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1729 2018-05-10 Carl Love <cel@us.ibm.com>
1731 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1732 dcbtt and dcbtstt if operands[2] is 0.
1734 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1737 * config/i386/sse.md (usadv64qi): New expander.
1739 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1741 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1742 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1743 -maltivec=be support.
1744 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1745 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1746 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1747 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1748 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1749 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1750 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1751 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1752 altivec_vsumsws): Adjust.
1753 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1754 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1755 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1757 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1758 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1759 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1760 (altivec_lve<VI_char>x): Delete expand.
1761 (*altivec_lve<VI_char>x_internal): Rename to...
1762 (altivec_lve<VI_char>x): ... this.
1763 (altivec_lvxl_<mode>): Delete expand.
1764 (*altivec_lvxl_<mode>_internal): Rename to ...
1765 (altivec_lvxl_<mode>): ... this.
1766 (altivec_stvxl_<mode>): Delete expand.
1767 (*altivec_stvxl_<mode>_internal): Rename to ...
1768 (altivec_stvxl_<mode>): ... this.
1769 (altivec_stve<VI_char>x): Delete expand.
1770 (*altivec_stve<VI_char>x_internal): Rename to ...
1771 (altivec_stve<VI_char>x): ... this.
1772 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1773 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1774 reduc_plus_scal_<mode>): Adjust.
1775 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1777 (rs6000_cpu_cpp_builtins): Adjust.
1778 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1779 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1780 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1781 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1782 -maltivec=be support.
1783 (rs6000_split_vec_extract_var): Adjust.
1784 (rs6000_split_v4si_init): Adjust.
1785 (swap_selector_for_mode): Delete.
1786 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1787 altivec_expand_stvex_be): Delete.
1788 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1789 -maltivec=be support.
1790 (rs6000_gimple_fold_builtin): Ditto.
1791 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1793 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1794 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1795 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1796 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1797 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1798 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1799 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1800 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1801 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1802 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1803 anonymous split): Adjust.
1804 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1805 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1807 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1809 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1810 when --with-gxx-include-dir is also specified.
1811 * configure: Regenerate.
1813 2018-05-09 Jim Wilson <jimw@sifive.com>
1816 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1817 * config/riscv/t-withmultilib: New.
1818 * config/riscv/withmultilib.h: New.
1819 * doc/install.texi: Document RISC-V --with-multilib-list support.
1821 2018-05-09 Richard Biener <rguenther@suse.de>
1823 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1825 (vect_bb_vectorization_profitable_p): Adjust. Compute
1826 actual scalar cost using the cost vector and the add_stmt_cost
1829 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1831 PR rtl-optimization/85645
1832 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1833 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1835 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1837 PR rtl-optimization/85645
1838 * shrink-wrap.c (spread_components): Return a boolean saying if
1839 anything was changed.
1840 (try_shrink_wrapping_separate): Iterate spread_components until
1841 nothing changes anymore.
1843 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1845 PR rtl-optimization/85645
1846 * regrename.c (build_def_use): Also kill the chains that include the
1847 destination of a REG_CFA_REGISTER note.
1849 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1851 PR rtl-optimization/85645
1852 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1853 insn that has a REG_CFA_REGISTER note.
1855 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1857 * cfgexpand.c (expand_clobber): New function.
1858 (expand_gimple_stmt_1): Use it.
1859 * tree-vect-stmts.c (vect_clobber_variable): New function,
1861 (vectorizable_simd_clone_call): ...here.
1862 (vectorizable_store): Emit a clobber either side of an
1863 IFN_STORE_LANES sequence.
1864 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1866 2018-05-09 Tom de Vries <tom@codesourcery.com>
1869 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1870 (define_insn "trap_if_false"): Add exit after trap.
1872 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1874 PR rtl-optimization/85638
1875 * bb-reorder.c: Include common/common-target.h.
1876 (create_forwarder_block): New function extracted from...
1877 (fix_up_crossing_landing_pad): ...here. Rename into...
1878 (dw2_fix_up_crossing_landing_pad): ...this.
1879 (sjlj_fix_up_crossing_landing_pad): New function.
1880 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1881 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1882 from both partitions and exit the loop after one iteration.
1884 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1887 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1889 (Basic PowerPC Built-in Functions): The new name of the
1890 subsection previously known as "PowerPC Built-in Functions".
1891 (Basic PowerPC Built-in Functions Available on all Configurations):
1893 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1895 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1897 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1899 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1902 2018-05-08 Jim Wilson <jimw@sifive.com>
1904 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1905 (LD_EMUL_SUFFIX): New.
1906 (LINK_SPEC): Use it.
1908 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1910 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1912 (Basic PowerPC Built-in Functions): The new name of the
1913 subsection previously known as "PowerPC Built-in Functions".
1914 (Basic PowerPC Built-in Functions Available on all Configurations):
1916 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1918 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1920 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1922 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1925 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1928 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1929 after cmpelim optimization.
1931 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1933 * config.gcc: Support "goldmont".
1934 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1935 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1937 * config/i386/i386.c (m_GOLDMONT): Define.
1938 (processor_target_table): Add "goldmont".
1939 (PTA_GOLDMONT): Define.
1940 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1941 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1942 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1943 (fold_builtin_cpu): Add "goldmont".
1944 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1945 (ix86_option_override_internal): Add "goldmont".
1946 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1947 (processor_type): Add PROCESSOR_GOLDMONT.
1948 * config/i386/i386.md: Add CPU "glm".
1949 * config/i386/glm.md: New file.
1950 * config/i386/x86-tune.def: Add m_GOLDMONT.
1951 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1953 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1956 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1958 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1959 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1960 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1963 * config/i386/i386.c (ix86_fold_builtin): Handle
1964 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1967 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1968 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1970 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1973 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1974 (check_arch): Likewise.
1975 (check_fpu): Return the result rather than printing it.
1976 (end arch): Fix operator precedence.
1977 (end cpu): Likewise.
1978 (END): Print the result from check_fpu.
1980 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1981 Alan Hayward <alan.hayward@arm.com>
1982 David Sherwood <david.sherwood@arm.com>
1984 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1985 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1986 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1987 (*fcmuo<mode>_and): New patterns.
1989 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1991 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1992 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1993 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1994 (cmp_op, sve_imm_con): New code attributes.
1995 (SVE_COND_INT_CMP, imm_con): Delete.
1996 (cmp_op): Remove above unspecs from int attribute.
1997 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1999 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2000 comparison-specific unspecs.
2001 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2002 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2003 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2004 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2005 (*vec_fcm<cmp_op><mode>): Rename to...
2006 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2007 (*vec_fcmuo<mode>): Rename to...
2008 (*fcmuo<mode>): ...this and adjust likewise.
2009 (*pred_fcm<cmp_op><mode>): New pattern.
2010 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2011 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2013 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2015 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2016 (aarch64_emit_sve_predicated_cond): New function.
2017 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2018 (aarch64_emit_unspec_cond_or): Replace with...
2019 (aarch64_emit_sve_or_conds): ...this new function. Use
2020 aarch64_emit_sve_ptrue_op for the individual comparisons and
2021 aarch64_emit_binop to OR them together.
2022 (aarch64_emit_inverted_unspec_cond): Replace with...
2023 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2024 aarch64_emit_sve_ptrue_op for the comparison and
2025 aarch64_emit_unop to invert the result.
2026 (aarch64_expand_sve_vec_cmp_float): Update after the above
2027 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2029 2018-05-07 Nathan Sidwell <nathan@acm.org>
2031 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2032 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2033 (Backwards Compatibility): Likewise.
2035 2018-05-07 Luis Machado <luis.machado@linaro.org>
2039 2018-05-07 Luis Machado <luis.machado@linaro.org>
2041 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2042 <prefetch_dynamic_strides>: New const bool field.
2043 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2044 prefetch_dynamic_strides.
2045 (exynosm1_prefetch_tune): Likewise.
2046 (thunderxt88_prefetch_tune): Likewise.
2047 (thunderx_prefetch_tune): Likewise.
2048 (thunderx2t99_prefetch_tune): Likewise.
2049 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2050 (aarch64_override_options_internal): Update to set
2051 PARAM_PREFETCH_DYNAMIC_STRIDES.
2052 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2053 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2054 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2055 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2056 prefetch-dynamic-strides setting.
2058 2018-05-07 Luis Machado <luis.machado@linaro.org>
2060 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2061 <minimum_stride>: New const int field.
2062 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2063 minimum_stride field.
2064 (exynosm1_prefetch_tune): Likewise.
2065 (thunderxt88_prefetch_tune): Likewise.
2066 (thunderx_prefetch_tune): Likewise.
2067 (thunderx2t99_prefetch_tune): Likewise.
2068 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2069 (aarch64_override_options_internal): Update to set
2070 PARAM_PREFETCH_MINIMUM_STRIDE.
2071 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2072 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2073 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2074 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2075 stride is constant and is below the minimum stride threshold.
2077 2018-05-07 Luis Machado <luis.machado@linaro.org>
2079 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2082 2018-05-07 Luis Machado <luis.machado@linaro.org>
2084 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2085 <prefetch_dynamic_strides>: New const bool field.
2086 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2087 prefetch_dynamic_strides.
2088 (exynosm1_prefetch_tune): Likewise.
2089 (thunderxt88_prefetch_tune): Likewise.
2090 (thunderx_prefetch_tune): Likewise.
2091 (thunderx2t99_prefetch_tune): Likewise.
2092 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2093 (aarch64_override_options_internal): Update to set
2094 PARAM_PREFETCH_DYNAMIC_STRIDES.
2095 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2096 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2097 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2098 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2099 prefetch-dynamic-strides setting.
2101 2018-05-07 Luis Machado <luis.machado@linaro.org>
2103 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2104 <minimum_stride>: New const int field.
2105 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2106 minimum_stride field.
2107 (exynosm1_prefetch_tune): Likewise.
2108 (thunderxt88_prefetch_tune): Likewise.
2109 (thunderx_prefetch_tune): Likewise.
2110 (thunderx2t99_prefetch_tune): Likewise.
2111 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2112 (aarch64_override_options_internal): Update to set
2113 PARAM_PREFETCH_MINIMUM_STRIDE.
2114 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2115 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2116 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2117 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2118 stride is constant and is below the minimum stride threshold.
2120 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2123 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2124 the type is addressable. Don't force op into register if it has
2127 2018-05-05 Roland McGrath <mcgrathr@google.com>
2130 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2131 any section for which we don't know a specific type it should have,
2132 regardless of name. Previously this was done only for the exact
2133 names ".init_array", ".fini_array", and ".preinit_array".
2134 (default_elf_asm_named_section): Add comment about
2135 relationship with default_section_type_flags and SECTION_NOTYPE.
2136 (get_section): Don't consider it a type conflict if one side has
2137 SECTION_NOTYPE and the other doesn't, as long as neither has the
2138 SECTION_BSS et al used in the default_section_type_flags logic.
2140 2018-05-05 Tom de Vries <tom@codesourcery.com>
2143 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2144 (workaround_barsyncs): New function.
2145 (nvptx_reorg): Use workaround_barsyncs.
2146 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2147 (define_expand "nvptx_membar_cta"): New define_expand.
2148 (define_insn "*nvptx_membar_cta"): New insn.
2150 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2152 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2153 To improve optimization opportunities.
2154 * builtin-types.def: The new needed builtin types for the above.
2156 2018-05-04 Richard Biener <rguenther@suse.de>
2158 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2159 * gimple-ssa-store-merging.c
2160 (imm_store_chain_info::output_merged_store): Remove redundant create,
2161 release split_store vector contents on failure.
2162 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2163 scalar stmt vector on cache hit.
2165 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2167 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2169 * config.gcc (powerpc-xilinx-eabi*): Remove.
2170 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2172 (fusion_addis_mem_combo_load): Ditto.
2173 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2175 (rs6000_cpu_cpp_builtins): Ditto.
2176 * config/rs6000/rs6000-linux.c
2177 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2178 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2179 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2181 (rs6000_setup_reg_addr_masks): Ditto.
2182 (rs6000_init_hard_regno_mode_ok): Ditto.
2183 (rs6000_option_override_internal): Ditto.
2184 (legitimate_lo_sum_address_p): Ditto.
2185 (rs6000_legitimize_address): Ditto.
2186 (rs6000_legitimize_reload_address): Ditto.
2187 (rs6000_legitimate_address_p): Ditto.
2188 (abi_v4_pass_in_fpr): Ditto.
2189 (setup_incoming_varargs): Ditto.
2190 (rs6000_gimplify_va_arg): Ditto.
2191 (rs6000_split_multireg_move): Ditto.
2192 (rs6000_savres_strategy): Ditto.
2193 (rs6000_emit_prologue_components): Ditto.
2194 (rs6000_emit_epilogue_components): Ditto.
2195 (rs6000_emit_prologue): Ditto.
2196 (rs6000_emit_epilogue): Ditto.
2197 (rs6000_elf_file_end): Ditto.
2198 (rs6000_function_value): Ditto.
2199 (rs6000_libcall_value): Ditto.
2200 * config/rs6000/rs6000.h: Ditto.
2201 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2202 (TARGET_MINMAX): ... this. New.
2203 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2204 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2205 (*movsi_internal1_single): Delete.
2206 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2207 mfpu=, mxilinx-fpu): Delete.
2208 * config/rs6000/singlefp.h: Delete.
2209 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2210 * config/rs6000/t-rs6000: Ditto.
2211 * config/rs6000/t-xilinx: Delete.
2212 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2213 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2214 (VStype_simple): Delete.
2215 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2216 * config/rs6000/xfpu.h: Delete.
2217 * config/rs6000/xfpu.md: Delete.
2218 * config/rs6000/xilinx.h: Delete.
2219 * config/rs6000/xilinx.opt: Delete.
2220 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2221 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2223 2018-05-04 Tom de Vries <tom@codesourcery.com>
2226 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2229 2018-05-04 Richard Biener <rguenther@suse.de>
2232 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2233 (expand_complex_div_wide): Likewise.
2234 (expand_complex_operations_1): Likewise.
2235 (expand_complex_libcall): Preserve EH info of the original stmt.
2236 (tree_lower_complex): Handle removed blocks.
2237 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2238 on complex multiplication and division libcall builtins.
2240 2018-05-04 Richard Biener <rguenther@suse.de>
2243 * fold-const.c (negate_expr_p): Restrict negation of operand
2244 zero of a division to when we know that can happen without
2246 (fold_negate_expr_1): Likewise.
2248 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2251 * real.h (real_nextafter): Declare.
2252 * real.c (real_nextafter): New function.
2253 * fold-const-call.c (fold_const_nextafter): New function.
2254 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2255 CASE_CFN_NEXTTOWARD.
2256 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2257 even when arg1_mode is different from arg0_mode.
2259 2018-05-03 Nathan Sidwell <nathan@acm.org>
2261 * doc/extend.texi (Deprecated Features): Remove
2263 (Backwards Compatibility): Likewise.
2264 * doc/invoke.texi (C++ Language Options): Likewise.
2265 (C++ Dialect Options): Likewise.
2267 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2270 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2271 _mm512_mask_mullox_epi64): New intrinsics.
2273 2018-05-03 Tom de Vries <tom@codesourcery.com>
2276 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2277 dump files): Add offload-tree.
2279 2018-05-03 Richard Biener <rguenther@suse.de>
2281 PR tree-optimization/85615
2282 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2283 to loops not nested in BBs loop father to avoid creating multi-entry
2286 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2288 PR tree-optimization/70291
2289 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2290 arguments. Change return type to tree. Emit libcall as a new
2291 statement rather than replacing existing one when inplace_p is true.
2292 (expand_complex_multiplication_components): New function.
2293 (expand_complex_multiplication): Expand floating-point complex
2294 multiplication using the above.
2295 (expand_complex_division): Rename inner_type parameter to type.
2296 Update expand_complex_libcall call-site.
2297 (expand_complex_operations_1): Update expand_complex_multiplication
2298 and expand_complex_division call-sites.
2300 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2303 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2304 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2305 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2306 the highest significant bit of the shift count mask is clear. In
2307 check whether and[sq]i3 is needed verify that all significant bits
2308 of the shift count other than the highest are set.
2310 2018-05-02 Tom de Vries <tom@codesourcery.com>
2313 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2314 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2315 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2316 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2317 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2318 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2319 __builtin_goacc_parlevel_size.
2321 2018-05-02 Richard Biener <rguenther@suse.de>
2323 PR tree-optimization/85597
2324 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2325 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2328 2018-05-02 Tom de Vries <tom@codesourcery.com>
2331 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2332 dump files): Add ltrans-tree.
2334 2018-05-02 Tom de Vries <tom@codesourcery.com>
2337 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2338 dump files): Add wpa-ipa.
2340 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2342 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2343 powerpc*-*-linux*paired* target.
2344 * config/rs6000/750cl.h: Delete.
2345 * config/rs6000/paired.h: Delete.
2346 * config/rs6000/paired.md: Delete.
2347 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2349 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2350 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2351 comment. Remove paired float support.
2352 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2353 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2355 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2356 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2358 * config/rs6000/rs6000.c: Remove paired float support.
2359 (paired_expand_vector_init, paired_expand_vector_move,
2360 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2361 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2362 paired_expand_builtin, paired_expand_predicate_builtin,
2363 paired_init_builtins): Delete.
2364 * config/rs6000/rs6000.h: Remove paired float support.
2365 * config/rs6000/rs6000.md: Remove paired float support.
2366 (move_from_CR_ov_bit): Delete.
2367 * config/rs6000/rs6000.opt (mpaired): Delete.
2368 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2369 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2371 2018-05-02 Richard Biener <rguenther@suse.de>
2374 * gimplify.c (gimplify_save_expr): When in SSA form allow
2375 SAVE_EXPRs to compute to SSA vars.
2377 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2380 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2381 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2382 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2383 clobber operands[2], instead use a new pseudo. Formatting fixes.
2385 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2387 PR tree-optimization/85586
2388 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2389 exit early for statements in the same group if the accesses are
2392 2018-05-02 Tom de Vries <tom@codesourcery.com>
2395 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2398 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2400 PR tree-optimization/85143
2401 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2403 2018-05-01 Tom de Vries <tom@codesourcery.com>
2406 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2407 not found" error message.
2409 2018-05-01 Tom de Vries <tom@codesourcery.com>
2412 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2413 * vec.c (test_ordered_remove_if): New function.
2414 (vec_c_tests): Call test_ordered_remove_if.
2415 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2416 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2417 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2418 VEC_ORDERED_REMOVE_IF.
2420 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2422 PR tree-optimization/82665
2423 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2424 pointer subtraction where arguments come from a memchr call.
2426 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2428 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2429 --push-state --as-needed and --pop-state instead of --as-needed and
2430 --no-as-needed if ld supports it.
2431 * configure: Regenerated.
2434 * doc/install.texi2html: Replace _002d with - and _002a with * in
2435 generated html files using sed.
2437 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2440 * gcc-rich-location.c (blank_line_before_p): New function.
2441 (use_new_line): New function.
2442 (gcc_rich_location::add_fixit_insert_formatted): New function.
2443 * gcc-rich-location.h
2444 (gcc_rich_location::add_fixit_insert_formatted): New function.
2446 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2448 * selftest.c (assert_streq): Rename "expected" and "actual" to
2449 "val1" and "val2". Extend NULL-handling to cover both inputs
2450 symmetrically, while still requiring both to be non-NULL for a pass.
2451 * selftest.h (assert_streq): Rename "expected" and "actual" to
2453 (ASSERT_EQ): Likewise.
2454 (ASSERT_EQ_AT): Likewise.
2455 (ASSERT_KNOWN_EQ): Likewise.
2456 (ASSERT_KNOWN_EQ_AT): Likewise.
2457 (ASSERT_NE): Likewise.
2458 (ASSERT_MAYBE_NE): Likewise.
2459 (ASSERT_MAYBE_NE_AT): Likewise.
2460 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2461 the assertion to pass.
2462 (ASSERT_STREQ_AT): Likewise.
2464 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2466 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2467 interaction with -pie.
2469 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2471 * selftest.h: Fix alphabetization of per-source-file selftest
2474 2018-04-30 Jason Merrill <jason@redhat.com>
2476 PR c++/61982 - dead stores to destroyed objects.
2477 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2480 2018-04-30 Jason Merrill <jason@redhat.com>
2482 * tree.c (build_clobber): New.
2483 * tree.h: Declare it.
2484 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2486 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2488 * diagnostic-show-locus.c (layout::layout): Update for
2489 location_get_source_line returning a char_span.
2490 (struct char_span): Move to input.h.
2491 (struct correction): Update for fields in char_span becoming
2493 (struct source_line): Update for location_get_source_line
2494 returning a char_span.
2495 (layout::print_line): Likewise.
2496 * edit-context.c (edited_file::print_content): Likewise.
2497 (edited_file::print_diff_hunk): Likewise.
2498 (edited_file::print_run_of_changed_lines): Likewise.
2499 (edited_file::get_num_lines): Likewise.
2500 (edited_line::edited_line): Likewise.
2501 * final.c (asm_show_source): Likewise.
2502 * input.c (location_get_source_line): Convert return type
2503 from const char * to char_span, losing the final "line_len"
2505 (dump_location_info): Update for the above.
2506 (get_substring_ranges_for_loc): Likewise. Use a char_span
2507 when handling the literal within the line.
2508 (test_reading_source_line): Update for location_get_source_line
2509 returning a char_span.
2510 * input.h (class char_span): Move here from
2511 diagnostic-show-locus.c, converting from a struct to a class.
2512 Make data members private.
2513 (char_span::operator bool): New.
2514 (char_span::length): New.
2515 (char_span::get_buffer): New.
2516 (char_span::operator[]): New.
2517 (char_span::subspan): Make const.
2518 (char_span::xstrdup): New.
2519 (location_get_source_line): Convert return type from const char *
2520 to char_span, losing the final "line_size" param.
2522 2018-04-30 Jan Hubicka <jh@suse.cz>
2524 * lto-wrapper.c (ltrans_priorities): New static var.
2525 (cmp_priority): New.
2526 (run_gcc): Read priorities and if doing parallel build order
2527 the Makefile by them.
2529 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2531 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2533 2018-04-30 Richard Biener <rguenther@suse.de>
2535 * tree-cfg.c (verify_address): Remove base argument, add
2536 flag whether to check TREE_ADDRESSABLE and do that.
2537 (verify_expr): Remove.
2538 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2539 (verify_gimple_assign_single): Likewise.
2540 (verify_gimple_switch): Likewise.
2541 (verify_expr_location_1): Dereference tp once. Add (disabled)
2542 piece from verify_expr.
2543 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2545 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2547 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2549 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2551 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2552 (small_data_pattern): Likewise.
2553 (arc_rewrite_small_data): Likewise.
2554 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2555 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2556 (get_symbol_alignment): New function.
2557 (legitimate_small_data_address_p): Likewise.
2558 (legitimate_scaled_address): Update, call
2559 legitimate_small_data_address_p.
2560 (output_sdata): New static variable.
2561 (arc_print_operand): Update how we handle small data operands.
2562 (arc_print_operand_address): Likewise.
2563 (arc_legitimate_address_p): Update, use
2564 legitimate_small_data_address_p.
2565 (arc_rewrite_small_data_p): Remove.
2566 (arc_rewrite_small_data_1): Likewise.
2567 (arc_rewrite_small_data): Likewise.
2568 (small_data_pattern): Likewise.
2569 (compact_sda_memory_operand): Update to use
2570 legitimate_small_data_address_p and get_symbol_alignment.
2571 (prepare_move_operands): Don't rewite sdata pattern.
2572 (prepare_extend_operands): Remove.
2573 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2575 (zero_extendqisi2): Likewise.
2576 (zero_extendhisi2): Likewise.
2577 (extendqihi2): Likewise.
2578 (extendqisi2): Likewise.
2579 (extendhisi2): Likewise.
2583 * config/arc/constraints.md (Usd): Change it to memory constraint.
2585 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2587 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2588 as source of std instructions.
2589 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2590 allow 6-bit constants as source for store instructions.
2591 (movdi_insn): Update instruction pattern to allow 6-bit constants
2592 as source for store instructions.
2594 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2596 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2598 2018-04-30 Nathan Sidwell <nathan@acm.org>
2599 Sandra Loosemore <sandra@codesourcery.com>
2601 * dumpfile.c (dump_open): Allow '-' for stdout.
2602 * doc/invoke.texi (Developer Options): Document dump filename
2603 determination early. Document stdin/stdout selection.
2605 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2607 Microblaze Target: PIC data text relative
2609 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2610 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2612 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2613 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2614 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2615 New addressing mode for data-text relative position indepenedent code.
2616 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2617 'ADDRESS_SYMBOLIC_TXT_REL'.
2618 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2619 (microblaze_legitimate_pic_operand): Exclude function calls from
2620 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2621 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2623 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2624 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2625 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2626 for 'address + offset'.
2627 (microblaze_expand_prologue): Add new function prologue call for
2629 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2630 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2631 table in case of TARGET_PIC_DATA_TEXT_REL.
2632 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2633 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2634 Add new macros 'UNSPEC_TEXT',
2635 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2636 + exclude function calls from 'UNSPEC_PLT' in case of data text
2638 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2639 new target hook for generating address diff vector tables in case of
2641 * doc/tm.texi : Regenerate.
2642 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2643 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2644 of addr diff vector generation.
2645 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2646 target hook definition.
2647 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2648 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2649 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2650 option for data text relative.
2652 2018-04-30 Richard Biener <rguenther@suse.de>
2654 * tree-chrec.h (evolution_function_is_constant_p): Remove
2656 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2658 2018-04-30 Richard Biener <rguenther@suse.de>
2661 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2663 2018-04-30 Richard Biener <rguenther@suse.de>
2665 PR tree-optimization/28364
2666 PR tree-optimization/85275
2667 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2668 copying first exit test.
2670 2018-04-28 Mark Wielaard <mark@klomp.org>
2672 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2674 (dwarf_AT): Handle DW_AT_addr_base.
2675 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2677 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2680 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2681 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2682 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2683 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2685 2018-04-28 Richard Biener <rguenther@suse.de>
2687 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2688 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2689 to reflect use. Only add interesting stmts.
2691 2018-04-27 Martin Jambor <mjambor@suse.cz>
2694 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2695 the jump function allows for passing through aggregate values.
2697 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2699 * input.h (in_system_header_at): Convert from macro to inline
2701 (from_macro_expansion_at): Likewise.
2702 (from_macro_definition_at): Likewise.
2704 2018-04-27 Jeff Law <law@redhat.com>
2706 * config.gcc: Mark tile* targets as deprecated/obsolete.
2708 2018-04-27 Richard Biener <rguenther@suse.de>
2710 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2713 2018-04-27 Richard Biener <rguenther@suse.de>
2715 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2717 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2719 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2720 with Yd constraint. Set "preferred_for_speed" attribute from
2721 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2723 (*movdi_internal): Ditto.
2724 (movti_interunit splitters): Remove
2725 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2726 (movdi_interunit splitters): Ditto.
2727 * config/i386/constraints.md (Ye): Remove.
2728 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2730 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2733 * config/aarch64/constraints.md (Usg): Limit to 31.
2736 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2738 PR tree-optimization/85529
2739 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2740 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2741 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2742 zero extension or masking of the MSB bit.
2743 (optimize_range_tests): Add FIRST_BB argument, pass it through
2744 to optimize_range_tests_var_bound.
2745 (maybe_optimize_range_tests, reassociate_bb): Adjust
2746 optimize_range_tests callers.
2748 2018-04-26 Richard Biener <rguenther@suse.de>
2749 Jakub Jelinek <jakub@redhat.com>
2751 * cgraph.h (symbol_table): Just declare debug method here.
2752 * symtab.c (symbol_table::debug): Define.
2754 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2756 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2758 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2760 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2761 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2762 (*movdi_internal): Substitute Yi and Yj constraint with x
2763 and Ym and Yn constraint with y constraint. Update "isa"
2764 attribute and set "preferred_for_speed" attribute from
2765 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2766 (*movsi_internal): Ditto.
2767 (*movdf_internal): Ditto.
2768 (*movsf_internal): Ditto.
2769 (*zero_extendsidi2): Ditto.
2770 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2771 (sse2_loadld): Ditto.
2772 (*vec_extract<ssevecmodelower>_0): Ditto.
2773 (*vec_extractv4si_0_zext_sse4): Ditto.
2774 (vec_concatv2di): Ditto.
2775 (*vec_dup<mode>): Ditto.
2776 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2777 * config/i386/constraints.md (Yi): Remove.
2782 2018-04-26 Nathan Sidwell <nathan@acm.org>
2784 * dumpfile.c (dump_open): New.
2785 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2786 (dump_finish): Detect stdio/stderr by value not name.
2788 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2790 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2792 2018-04-26 Tom de Vries <tom@codesourcery.com>
2795 * config/nvptx/nvptx.c (verify_neutering_jumps)
2796 (verify_neutering_labels): New function
2797 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2799 2018-04-26 Tom de Vries <tom@codesourcery.com>
2802 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2803 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2806 2018-04-26 Richard Biener <rguenther@suse.de>
2807 Tom de Vries <tom@codesourcery.com>
2810 * lto-streamer-out.c (output_function): Fixup loops if required to match
2811 discovery done in the reader.
2813 2018-04-26 Richard Biener <rguenther@suse.de>
2815 PR tree-optimization/85116
2816 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2817 have a loop exit from the single latch predecessor. Remove
2818 case of header with just condition.
2819 (ch_base::copy_headers): Exclude infinite loops from any
2821 (pass_ch::execute): Record exits.
2823 2018-04-26 Richard Biener <rguenther@suse.de>
2825 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2826 prologue cost vector and pass it to vect_get_load_cost.
2827 (vect_get_peeling_costs_all_drs): Likewise.
2828 (vect_peeling_hash_get_lowest_cost): Likewise.
2829 (vect_enhance_data_refs_alignment): Likewise.
2831 2018-04-26 Richard Biener <rguenther@suse.de>
2834 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2835 checking of integer<->pointer conversions.
2836 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2837 sign-/zero-extending pointer types.
2838 (expand_omp_for_static_chunk): Likewise.
2840 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2841 Jean Lee <xiaoyur347@gmail.com>
2843 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2844 (TARGET_ASAN_SHADOW_OFFSET): Define.
2845 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2846 true for -fsanitize=address.
2848 2018-04-25 Mark Wielaard <mark@klomp.org>
2850 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2853 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2855 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2856 than "alu", remove explicit "memory" and "imm_disp" attributes.
2857 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2860 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2861 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2862 gen_lowpart_no_emit.
2864 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2867 * config/i386/i386.c (ix86_expand_builtin): Change memory
2868 operand to XI, extend p0 to Pmode.
2869 * config/i386/i386.md: Change unspec volatile and operand
2870 1 mode to XI, change operand 0 mode to P.
2872 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2874 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2875 GET_MODE_MASK before any checking.
2876 (nds32_can_use_bset_p): Likewise.
2877 (nds32_can_use_btgl_p): Likewise.
2879 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2881 * config/nds32/nds32-doubleword.md: New define_split pattern for
2882 illegal register number.
2884 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2886 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2888 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2890 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2892 2018-04-25 Richard Biener <rguenther@suse.de>
2894 * lto-streamer.h (LTO_major_version): Bump to 8.
2896 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2898 * BASE-VER: Set to 9.0.0.
2900 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2902 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2903 in __abskf2 and __powikf2.
2905 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2908 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2909 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2910 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2911 Use the above on operand 2. Reindent.
2912 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2914 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2917 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2919 * config/i386/i386.opt (mcet): Removed.
2920 * doc/install.texi: Remove -mcet documentation.
2921 * doc/invoke.texi: Likewise.
2923 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2926 * doc/install.texi: Remove -mcet from bootstrap-cet.
2928 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2931 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2932 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2936 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2937 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2938 containing a CONST_VECTOR.
2940 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2942 * doc/install.texi: Update newlib dependency for nvptx.
2944 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2947 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2948 instead of INTVAL when shifting x left.
2950 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2952 PR tree-optimization/85478
2953 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2954 vect_grouped_store_supported for single element vectors.
2956 2018-04-24 Richard Biener <rguenther@suse.de>
2959 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2960 load cost increase to the case of non-constant step.
2962 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2965 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2966 destination if any_malformed_asm.
2968 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2971 * expr.c (store_field): In the bitfield case, if the value comes from
2972 a function call and is returned in registers by means of a PARALLEL,
2973 do not change the mode of the temporary unless BLKmode and VOIDmode.
2975 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2977 PR rtl-optimization/85423
2978 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2979 dependencies to debug insns when the previous insn is non-debug.
2981 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2983 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2984 enums into a single definition.
2985 (fls): Fix predicates and printing.
2988 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2990 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2991 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2992 and short u6 immediate.
2993 (check_if_valid_sleep_operand): Remove.
2994 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2996 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2998 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2999 flag_always_save_lp condition.
3000 * config/nds32/nds32.opt (malways-save-lp): New option.
3002 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3004 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3005 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3006 * config/nds32/nds32.h
3007 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3008 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3010 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3012 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3013 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3015 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3016 Chung-Ju Wu <jasonwucj@gmail.com>
3018 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3020 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3021 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3023 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3025 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3027 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3029 * config/nds32/nds32-protos.h (nds32_data_alignment,
3030 nds32_local_alignment): Declare.
3031 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3032 nds32_local_alignment): New functions.
3033 (TARGET_CONSTANT_ALIGNMENT): Define.
3034 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3036 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3038 * config/nds32/nds32.c
3039 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3040 (TARGET_MODES_TIEABLE_P): Likewise.
3042 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3044 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3047 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3048 Chung-Ju Wu <jasonwucj@gmail.com>
3050 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3051 for unaligned access.
3052 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3054 * config/nds32/nds32-intrinsic.md: Likewise.
3055 * config/nds32/nds32_intrinsic.h: Likewise.
3056 * config/nds32/nds32.h (nds32_builtins): Likewise.
3057 * config/nds32/nds32.opt (munaligned-access): New option.
3058 * config/nds32/nds32.c (nds32_asm_file_start): Display
3059 flag_unaligned_access status.
3061 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3063 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3064 -mno-relax is present.
3065 * config/riscv/linux.h (LINK_SPEC): Ditto.
3067 2018-04-20 Martin Sebor <msebor@redhat.com>
3070 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3072 (gimple_fold_builtin_stxcpy_chk): Same.
3073 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3075 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3078 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3079 __powikf2 when long double is IEEE 128-bit.
3081 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3083 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3084 step to make sure stack always aligned.
3086 2018-04-20 Carl Love <cel@us.ibm.com>
3089 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3090 size check for arg0.
3092 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3093 Tom de Vries <tom@codesourcery.com>
3096 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3097 Emit insns for calls too.
3098 (nvptx_find_par): Always look for worker-level predecessor insn.
3099 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3101 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3102 (nvptx_process_pars): Propagate frames for calls.
3104 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3107 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3109 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3110 (ix86_handle_option): Don't handle OPT_mibt.
3111 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3113 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3115 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3116 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3117 (ix86_target_macros): Define __CET__ with flag_cf_protection
3118 for -fcf-protection.
3119 * config/i386/i386.c (isa2_opts): Remove -mibt.
3120 * config/i386/i386.h (TARGET_IBT): Removed.
3121 (TARGET_IBT_P): Likewise.
3122 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3123 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3124 * config/i386/i386.opt (mcet): Update help message.
3127 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3128 -mcet as an alias for -mshstk.
3130 2018-04-20 Richard Biener <rguenther@suse.de>
3133 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3134 complexity by forcing a single use of the multiply operand.
3136 2018-04-20 Martin Jambor <mjambor@suse.cz>
3139 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3140 recursion dependency to only apply to non-clones.
3142 2018-04-20 Martin Jambor <mjambor@suse.cz>
3145 * ipa-cp.c (create_specialized_node): Check that clones of
3146 self-recursive edges exist during IPA-CP.
3148 2018-04-19 Toon Moene <toon@moene.org>
3150 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3153 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3155 PR tree-optimization/85467
3156 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3157 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3158 VECTOR_CST element to type.
3160 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3163 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3164 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3165 (builtin_longjmp): Likewise.
3166 (save_stack_nonlocal): New pattern.
3167 (restore_stack_nonlocal): Likewise.
3169 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3172 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3173 Replace ASM_OUTPUT_LABEL with fprintf.
3175 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3178 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3179 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3180 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3181 define __IBT__ and __SHSTK__ for -fcf-protection.
3182 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3184 (ix86_trampoline_init): Likewise.
3185 (x86_output_mi_thunk): Likewise.
3186 (ix86_notrack_prefixed_insn_p): Likewise.
3187 (ix86_option_override_internal): Don't disallow -fcf-protection.
3188 * config/i386/i386.md (rdssp<mode>): Also enable for
3190 (incssp<mode>): Likewise.
3191 (nop_endbr): Likewise.
3192 * config/i386/i386.opt (mcet): Change help message to built-in
3196 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3197 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3198 enable CET built-in functions.
3200 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3202 * common/config/i386/i386-common.c
3203 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3204 OPTION_MASK_ISA_MOVDIRI_UNSET,
3205 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3206 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3207 * config.gcc (movdirintrin.h): New header.
3208 * config/i386/cpuid.h (bit_MOVDIRI,
3209 bit_MOVDIR64B): New bits.
3210 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3212 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3213 (VOID, PVOID, PCVOID)): New function types.
3214 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3215 __builtin_ia32_directstoreu_u64,
3216 __builtin_ia32_movdir64b): New builtins.
3217 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3218 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3220 (ix86_valid_target_attribute_inner_p): Ditto.
3221 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3222 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3223 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3224 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3225 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3226 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3227 (movdiri<mode>, movdir64b_<mode>): New.
3228 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3229 * config/i386/immintrin.h: Include movdirintrin.h.
3230 * config/i386/movdirintrin.h: New file.
3231 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3233 2018-04-19 Richard Biener <rguenther@suse.de>
3236 * cfg.c (clear_bb_flags): When loop state says we have
3237 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3239 2018-04-19 Richard Biener <rguenther@suse.de>
3241 PR tree-optimization/84737
3242 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3243 copying restrict info.
3244 (vect_setup_realignment): Use it.
3245 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3246 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3247 the first DR to all generated stores.
3248 (vectorizable_load): Likewise for loads.
3250 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3252 PR tree-optimization/85446
3253 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3254 the integral and pointer types to have the same precision.
3256 * doc/install.texi: Document --disable-cet being the default and
3259 2018-04-18 Martin Liska <mliska@suse.cz>
3261 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3264 2018-04-18 Martin Liska <mliska@suse.cz>
3267 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3270 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3271 arguments if they are comparable.
3273 2018-04-18 Martin Liska <mliska@suse.cz>
3276 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3279 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3282 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3285 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3286 ENDBR after calling __morestack.
3288 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3291 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3292 by using gcc_base_ver to generate a gcc_driver_version, and use
3293 it when generating GCC_DRIVER_NAME.
3294 * configure: Regenerate.
3296 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3299 * config.gcc: Obsolete powerpc*-*-*spe*.
3301 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3304 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3305 (stabstr_D): Change type of unum from unsigned int to
3306 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3309 2018-04-17 Jim Wilson <jimw@sifive.com>
3312 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3313 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3314 Set arg_pointer_offset after using pretend_args_size.
3316 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3318 PR rtl-optimization/85431
3319 * dse.c (record_store): Ignore zero width stores.
3322 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3323 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3324 __builtin_stack_restore rather than after it.
3325 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3326 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3327 argument instead of virtual_dynamic_stack_rtx.
3329 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3331 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3333 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3334 Add note to error message to explain internal mapping of overloaded
3335 built-in function name to non-overloaded built-in function name.
3336 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3339 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3342 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3343 where the inputs overlap with the output.
3345 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3347 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3348 (=v, v) alternative and explicit "memory" attribute.
3349 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3350 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3352 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3353 "sselog1" type instead of "sselog".
3354 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3355 "sselog". Remove explicit "memory" attribute.
3356 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3357 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3359 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3360 "sselog1" type instead of "sselog". Remove explicit "memory"
3362 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3363 use "sselog1" type instead of "sselog". Remove explicit "memory"
3365 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3366 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3368 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3369 "sselog1" type instead of "sselog". Remove explicit "memory"
3371 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3372 use "sselog1" type instead of "sselog". Remove explicit "memory"
3376 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3379 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3382 2018-04-17 Martin Jambor <mjambor@suse.cz>
3385 * ipa-cp.c (create_specialized_node): Call
3386 expand_all_artificial_thunks if necessary.
3388 2018-04-17 Martin Liska <mliska@suse.cz>
3391 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3392 in message, remote space in between '_G' and '('.
3394 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3397 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3398 avx512f_vmcmp<mode>3<round_saeonly_name>,
3399 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3400 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3401 avx512f_rndscale<mode><round_saeonly_name>,
3402 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3403 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3404 Use %<iptr>2 instead of %2 for -masm=intel.
3405 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3406 avx512f_vcvttss2usi<round_saeonly_name>,
3407 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3409 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3410 avx512f_vcvttsd2usi<round_saeonly_name>,
3411 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3412 Use %q1 instead of %1 for -masm=intel.
3413 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3414 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3415 of %3 for -masm=intel.
3416 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3418 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3420 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3422 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3423 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3425 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3427 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3428 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3429 %g1 and one with %0 and %1.
3430 (avx512er_vmrcp28<mode><round_saeonly_name>,
3431 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3433 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3434 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3435 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3436 of %0 and %{%4%} for -masm=intel.
3437 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3438 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3439 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3440 order of %0 and %{%5%}%{z%} for -masm=intel.
3442 2018-04-17 Jan Hubicka <jh@suse.cz>
3445 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3447 2018-04-17 Martin Liska <mliska@suse.cz>
3450 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3451 for target_clone error message. Make default implementation
3452 clone to be a local declaration.
3453 (separate_attrs): Add new argument and check for an empty
3455 (expand_target_clones): Handle it.
3456 (ipa_target_clone): Make redirection just for target_clones
3459 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3460 Tom de Vries <tom@codesourcery.com>
3463 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3464 tiled basic blocks without omp continue statements.
3466 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3469 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3470 vec_extract expression as having side effects to make sure it gets
3473 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3476 * config/i386/i386.c (get_builtin_code_for_version): Check
3479 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3482 * gcc/config.gcc: Support "skylake".
3483 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3485 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3486 (processor_target_table): Add "skylake".
3487 (ix86_option_override_internal): Add "skylake".
3488 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3489 PROCESSOR_CANNONLAKE.
3490 (get_builtin_code_for_version): Fix priority for
3491 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3492 PROCESSOR_SKYLAKE-AVX512.
3493 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3494 (processor_type): Add PROCESSOR_SKYLAKE.
3496 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3497 Jason Merrill <jason@redhat.com>
3500 * convert.c (convert_to_integer_1): Use direct recursion for
3501 enumeral types and types with a precision less than the number
3502 of bits in their mode.
3504 2018-04-16 Julia Koval <julia.koval@intel.com>
3507 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3508 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3510 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3513 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3514 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3515 and -mno-direct-move.
3517 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3520 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3521 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3522 Ensure negative shifts result in {0}.
3524 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3526 PR rtl-optimization/79916
3527 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3528 regs (if any) to define how to gnerate SD moves when LRA is in
3531 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3533 PR rtl-optimization/85393
3534 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3535 * except.c (expand_dw2_landing_pad_for_region): Make static.
3536 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3537 a label and unconditional jump to old_bb, rather than
3538 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3541 PR rtl-optimization/85376
3542 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3543 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3544 instead of a specific value.
3546 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3547 Bin Cheng <bin.cheng@arm.com>
3549 PR tree-optimization/82965
3550 PR tree-optimization/83991
3551 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3552 by_profile_only parameter.
3553 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3554 information if the loop was predicted to iterate too many times.
3555 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3557 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3560 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3563 2018-04-13 Martin Liska <mliska@suse.cz>
3564 Jakub Jelinek <jakub@redhat.com>
3567 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3568 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3569 * builtins.c (expand_builtin_memory_copy_args): Use
3570 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3571 handle dest_addr == pc_rtx.
3573 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3576 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3577 asked to not generate direct moves.
3578 (fix_trunc<mode>si2_stfiwx): Similar.
3579 (fix_trunc<mode>si2_internal): Similar.
3581 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3584 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3585 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3586 lookup if dest in some wider mode is known to be const0_rtx and
3587 if so, record permanent equivalence for it to be ZERO_EXTEND of
3588 the narrower mode destination.
3590 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3592 * lto-streamer-out.c (output_function): Revert 259346.
3593 * omp-expand.c (expand_oacc_for): Likewise.
3595 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3597 PR rtl-optimization/85354
3598 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3599 * sel-sched.c (sel_global_init): ... here.
3601 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3604 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3605 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3606 mode for PE-COFF targets.
3607 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3608 (i386_pe_asm_lto_end): Likewise.
3609 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3610 (TARGET_ASM_LTO_END): Likewise.
3611 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3612 (i386_pe_asm_lto_start): New function.
3613 (i386_pe_asm_lto_end): Likewise.
3615 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3616 Richard Biener <rguenther@suse.de>
3619 * lto-streamer-out.c (output_function): Fix CFG loop state before
3621 * omp-expand.c (expand_oacc_for): Handle calls to internal
3622 functions like regular functions.
3624 2018-04-12 Richard Biener <rguenther@suse.de>
3627 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3628 for the early LTO debug to properly generate references to it
3629 during DIE emission. Do not re-use that for the skeleton for
3631 (dwarf2out_early_finish): Likewise.
3633 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3636 * config/i386/sse.md
3637 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3638 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3639 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3640 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3641 and output is a reg, avoid creating invalid lowpart subreg, but
3642 instead split into a 512-bit move. Don't split if not AVX512VL,
3643 input is xmm16+ reg and output is a mem.
3644 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3645 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3646 xmm16+ reg and output is a mem.
3648 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3650 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3651 also for flag_dwarf2_cfi_asm.
3653 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3655 PR rtl-optimization/85342
3656 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3657 a bool scalar var inside of the loop instead. Don't try to update
3658 recog_data.operand after failed apply_change_group.
3660 2018-04-12 Tom de Vries <tom@codesourcery.com>
3663 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3664 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3665 array with flexible array member as array without given dimension.
3666 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3667 argument for undefined param to true.
3669 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3672 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3673 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3674 from PowerPC section.
3675 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3676 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3677 help text that is too long.
3678 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3679 help text that is too long.
3680 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3681 help text that is too long.
3683 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3685 * config/alpha/alpha.md (stack_probe_internal): Rename
3686 from "probe_stack". Update all callers.
3688 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3690 PR rtl-optimization/84566
3691 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3692 sched_macro_fuse_insns.
3694 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3697 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3698 (compute_block_dependences): ... from here.
3700 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3702 PR tree-optimization/85331
3703 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3704 from int to HOST_WIDE_INT.
3706 2018-04-11 Martin Jambor <mjambor@suse.cz>
3709 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3710 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3711 not the same as the source val.
3712 (cgraph_edge_brings_value_p): New parameter.
3713 (gather_edges_for_value): Pass destination value to
3714 cgraph_edge_brings_value_p.
3715 (perhaps_add_new_callers): Likewise.
3716 (get_info_about_necessary_edges): Likewise and exclude values brought
3717 only by self-recursive edges.
3718 (create_specialized_node): Redirect only clones of self-calling edges.
3719 (+self_recursive_pass_through_p): New function.
3720 (find_more_scalar_values_for_callers_subset): Use it.
3721 (find_aggregate_values_for_callers_subset): Likewise.
3722 (known_aggs_to_agg_replacement_list): Removed.
3723 (decide_whether_version_node): Re-calculate known constants for all
3724 remaining context clones.
3726 2018-04-11 Richard Biener <rguenther@suse.de>
3729 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3730 from early DWARF output.
3731 (dwarf2out_early_finish): Output line info unconditionally into
3732 early DWARF and add reference to it.
3734 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3737 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3738 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3739 other than V2DFmode using iptr mode attribute.
3740 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3742 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3744 PR rtl-optimization/84659
3745 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3747 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3750 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3752 (output_loc_list): Pass address of a dummy size variable even in the
3753 locview handling loop.
3754 (index_location_lists): Add comment on why skip_loc_list_entry can't
3757 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3760 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3763 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3766 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3767 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3768 and -mstring-compare-inline-limit.
3770 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3773 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3774 for stack clash protection in a register whenever we need it to be in
3777 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3779 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3780 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3782 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3785 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3787 (mlong-double-): Ditto.
3788 * config/rs6000/sysv4.opt (msdata=): Ditto.
3789 (mtls-size=): Ditto.
3791 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3793 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3794 erroneous entries for
3795 "vector int vec_ldl (int, long int *)", and
3796 "vector unsigned int vec_ldl (int, unsigned long int *)".
3797 Add comments and entries for
3798 "vector bool char vec_ldl (int, bool char *)",
3799 "vector bool short vec_ldl (int, bool short *)",
3800 "vector bool int vec_ldl (int, bool int *)",
3801 "vector bool long long vec_ldl (int, bool long long *)",
3802 "vector pixel vec_ldl (int, pixel *)",
3803 "vector long long vec_ldl (int, long long *)",
3804 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3805 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3806 type tree bool_long_long_type_node and correct definition of
3807 bool_V2DI_type_node to make reference to this new type tree.
3808 (rs6000_mangle_type): Replace erroneous reference to
3809 bool_long_type_node with bool_long_long_type_node.
3810 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3811 comments to emphasize sign distinctions for char and int types and
3812 replace RS6000_BTI_bool_long constant with
3813 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3814 use of RS6000_BTI_pixel.
3815 (bool_long_type_node): Remove this macro definition.
3816 (bool_long_long_type_node): New macro definition
3818 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3820 PR rtl-optimization/85300
3821 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3822 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3823 simplify_unary_operation fails.
3825 2018-04-10 Martin Liska <mliska@suse.cz>
3827 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3828 cgraph_edge and ipa_ref.
3830 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3834 * config/i386/sse.md
3835 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3836 computation of the VEC_MERGE selector from mask.
3837 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3838 Fix decoding of the VEC_MERGE selector into mask.
3840 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3842 PR tree-optimization/85286
3843 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3845 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3847 * final.c (final_1): Set insn_last_address as well as
3848 insn_current_address.
3850 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3853 * explow.c (emit_stack_probe): Call validize_mem on memory location
3854 before passing it to gen_probe_stack. Create address operand and
3855 legitimize it for the probe_stack_address case.
3857 2018-04-09 Jan Hubicka <jh@suse.cz>
3860 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3861 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3862 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3864 (free_lang_data): Rebuild type inheritance graph.
3866 2018-04-09 Martin Sebor <msebor@redhat.com>
3868 * invoke.texi (-finline-small-functions): Mention other optimization
3870 (-findirect-inlining, -fpartial-inlining): Same.
3871 (-finline-functions-called-once): Same.
3872 (-freorder-blocks-and-partition): Same.
3874 2018-04-09 Jan Hubicka <jh@suse.cz>
3877 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3878 jumps; choose last target that matches the criteria (i.e.
3879 no partition changes for non-crossing jumps).
3880 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3881 support for redirecting crossing jumps to non-crossing.
3883 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3885 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3886 also for naked functions.
3888 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3890 * config/arc/arc.md (add_shift): New pattern.
3891 (add_shift2): Likewise.
3892 (sub_shift): Likewise.
3893 (sub_shift_cmp0_noout): Likewise.
3894 (compare_si_ashiftsi): Likewise.
3895 (xbfu_cmp0_noout): New combine pattern.
3896 (xbfu_cmp0"): Likewise.
3897 (movsi_set_cc_insn): Place the predicable variant first.
3898 (commutative_binary_cmp0_noout): Remove clobber.
3899 (commutative_binary_cmp0): New pattern.
3900 (noncommutative_binary_cmp0): Likewise.
3901 (noncommutative_binary_cmp0_noout): Likewise.
3902 (noncommutative_binary_comparison_result_used): Removed.
3903 (rsub_cmp0): New pattern.
3904 (rsub_cmp0_noout): Likewise.
3905 (extzvsi): Changed, keep only meaningful variants.
3906 (SQH, SEZ): New iterators.
3907 (SQH_postfix): New mode attribute.
3908 (SEZ_prefix): New code attribute.
3909 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3910 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3911 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3913 (noncommutative_operator): Check the availability of barrel
3916 2018-04-09 Richard Biener <rguenther@suse.de>
3918 PR tree-optimization/85284
3919 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3920 Only use the niter constraining form of simple_iv when the exit
3923 2018-04-09 Tom de Vries <tom@codesourcery.com>
3926 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3927 (define_expand "*memory_barrier"): New define_expand.
3928 (define_insn "memory_barrier"): New insn.
3930 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3932 PR rtl-optimization/80463
3933 PR rtl-optimization/83972
3934 PR rtl-optimization/83480
3936 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3937 correct producer for the insn.
3938 (tidy_control_flow): Fixup seqnos in case of debug insns.
3940 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3942 PR rtl-optimization/83913
3944 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3945 different sched-times when merging exprs.
3947 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3949 PR rtl-optimization/83962
3951 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3952 tidy_fallthru_edge and tidy_control_flow.
3954 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3956 PR rtl-optimization/83530
3958 * sel-sched.c (force_next_insn): New global variable.
3959 (remove_insn_for_debug): When force_next_insn is true, also leave only
3960 next insn in the ready list.
3961 (sel_sched_region): When the region wasn't scheduled, make another pass
3962 over it with force_next_insn set to 1.
3964 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3966 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3968 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3969 for interrupt control.
3970 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3971 functions for interrupt control.
3972 * config/nds32/nds32-intrinsic.md: Likewise.
3973 * config/nds32/nds32_intrinsic.h: Likewise.
3974 * config/nds32/nds32.h (nds32_builtins): Likewise.
3976 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3978 * config/nds32/nds32.c (nds32_init_machine_status,
3979 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3980 strict_aligned_p field.
3981 (nds32_expand_to_rtl_hook): New function.
3982 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3983 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3985 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3986 Chung-Ju Wu <jasonwucj@gmail.com>
3988 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3989 * config/nds32/nds32-n7.md: New file.
3990 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3991 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3993 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3994 * config/nds32/nds32.md (pipeline_model): Add n7.
3995 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3996 * config/nds32/pipelines.md: Include n7 settings.
3998 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3999 Chung-Ju Wu <jasonwucj@gmail.com>
4001 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4002 * config/nds32/nds32-e8.md: New file.
4003 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4004 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4006 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4007 * config/nds32/nds32.md (pipeline_model): Add e8.
4008 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4009 * config/nds32/pipelines.md: Include e8 settings.
4011 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4012 Chung-Ju Wu <jasonwucj@gmail.com>
4014 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4015 * config/nds32/nds32-n8.md: New file.
4016 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4017 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4019 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4020 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4021 * config/nds32/nds32.md (pipeline_model): Add n8.
4022 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4023 * config/nds32/pipelines.md: Include n8 settings.
4025 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4026 Chung-Ju Wu <jasonwucj@gmail.com>
4028 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4029 * config/nds32/nds32-n9-2r1w.md: New file.
4030 * config/nds32/nds32-n9-3r2w.md: New file.
4031 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4032 nds32_register_ports): New or modify for cpu n9.
4033 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4035 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4036 * config/nds32/nds32-utils.c: New file.
4037 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4038 TARGET_MUL_SLOW): Define.
4039 * config/nds32/nds32.md (pipeline_model): New attribute.
4040 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4041 New options that support cpu n9.
4042 * config/nds32/pipelines.md: Include n9 settings.
4043 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4045 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4047 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4048 information if necessary.
4049 (output_cond_branch_compare_zero): Likewise.
4050 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4051 (nds32_target_alignment): Refine for alignment.
4052 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4053 (FUNCTION_BOUNDARY): Modify.
4054 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4056 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4058 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4060 * config/nds32/constants.md (unspec_volatile_element): Add values for
4061 TLB operation and data prefetch.
4062 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4063 functions for TLB operation and data prefetch.
4064 * config/nds32/nds32-intrinsic.md: Likewise.
4065 * config/nds32/nds32_intrinsic.h: Likewise.
4066 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4067 (nds32_print_operand): Likewise.
4068 * config/nds32/nds32.h (nds32_builtins): Likewise.
4070 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4071 Andrew Pinski <pinsika@gcc.gnu.org>
4074 * match.pd: Use constant_boolean_node of correct type instead of
4075 boolean_true_node or boolean_false_node for simplifying
4076 pointer comparisons to zero.
4078 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4080 PR tree-optimization/80021
4081 * tree.c (verify_type_variant): Make error call in verify_variant_match
4082 translatable and remove final full stop.
4084 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4086 * config/nds32/constants.md (unspec_volatile_element): Add
4087 UNSPEC_VOLATILE_EH_RETURN.
4088 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4089 nds32_output_stack_pop): Support dwarf exception handling process.
4090 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4091 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4092 exception handling process.
4093 (nds32_compute_stack_frame): Likewise.
4094 (nds32_return_addr_rtx): Likewise.
4095 (nds32_initial_elimination_offset): Likewise.
4096 (nds32_expand_prologue): Likewise.
4097 (nds32_expand_epilogue): Likewise.
4098 (nds32_dynamic_chain_address): New function.
4099 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4101 (DYNAMIC_CHAIN_ADDRESS): Define.
4102 (EH_RETURN_DATA_REGNO): Define.
4103 (EH_RETURN_STACKADJ_RTX): Define.
4104 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4105 patterns for dwarf exception handling.
4107 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4109 * config/nds32/nds32.h: Clean up obsolete macros.
4111 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4113 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4114 Add enum values for particular instructions.
4115 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4116 particular intrinsic functions.
4117 * config/nds32/nds32-intrinsic.md: Likewise.
4118 * config/nds32/nds32_intrinsic.h: Likewise.
4119 * config/nds32/nds32.h (nds32_builtins): Likewise.
4120 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4121 (btst, ave): New patterns for particular instructions.
4123 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4125 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4126 Add enum values for atomic load/store and memory sync.
4127 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4129 * config/nds32/nds32-intrinsic.md: Likewise.
4130 * config/nds32/nds32_intrinsic.h: Likewise.
4131 * config/nds32/nds32.h (nds32_builtins): Likewise.
4133 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4135 PR tree-optimization/85257
4136 * fold-const.c (native_encode_vector): If not all elts could fit
4137 and off is -1, return 0 rather than offset.
4138 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4139 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4140 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4141 adjust buffer in native_interpret_expr call.
4143 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4145 * config/nds32/constants.md (unspec_volatile_element): Add cache
4146 control enum values.
4147 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4148 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4149 * config/nds32/nds32.c (nds32_cctl_names): New.
4150 (nds32_print_operand): Handle cache control register names.
4151 * config/nds32/nds32.h (nds32_builtins): New enum values.
4152 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4154 * config/nds32/nds32.md (type): Add mmu.
4155 * config/nds32/pipelines.md (simple_insn): Add mmu.
4157 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4159 * config/nds32/nds32.md (type): Remove call.
4160 * config/nds32/pipelines.md (simple_insn): Likewise.
4162 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4164 * config/nds32/constants.md (unspec_volatile_element): Add
4165 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4166 UNSPEC_VOLATILE_FMFCFG.
4167 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4168 description for fmfcfg and fmfcsr.
4169 (bdesc_1arg): Add fmtcsr.
4170 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4171 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4172 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4173 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4174 unspec_fmfcfg): New patterns.
4175 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4176 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4177 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4178 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4179 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4180 __nds32__fmfcfg): Define.
4182 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4184 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4185 intrinsic register names.
4186 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4187 intrinsic register enum values and macros.
4189 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4191 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4192 for load/store addressing form.
4193 (nds32_print_operand_address): Likewise.
4195 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4198 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4199 based on LABEL_REF. Remove useless assertion.
4200 (pic_address_needs_scratch): Fix formatting.
4201 (sparc_legitimize_pic_address): Minor tweaks.
4202 (sparc_delegitimize_address): Adjust assertion accordingly.
4203 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4204 into symbolic_operand.
4205 (movsi_high_pic_label_ref): Likewise.
4206 (movsi_lo_sum_pic_label_ref): Likewise.
4207 (movdi_pic_label_ref): Likewise.
4208 (movdi_high_pic_label_ref): Likewise.
4209 (movdi_lo_sum_pic_label_ref): Likewise.
4211 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4213 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4214 custom LIB_SPEC setup.
4216 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4217 Kito Cheng <kito.cheng@gmail.com>
4219 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4220 * config/riscv/freebsd.h: New.
4222 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4224 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4225 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4228 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4229 Kito Cheng <kito.cheng@gmail.com>
4231 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4232 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4233 * config/nds32/nds32-protos.h (nds32_output_call,
4234 nds32_output_return): Declare.
4235 * config/nds32/nds32.md: Refine all the call and return patterns.
4237 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4240 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4241 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4243 PR rtl-optimization/84872
4244 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4245 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4248 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4250 * expr.c (copy_blkmode_to_reg): Revert 254862.
4251 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4253 2018-04-06 Richard Biener <rguenther@suse.de>
4256 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4257 after seeing a component reference with an adjacent field. Treat
4258 refs to arrays at struct end of external decls similar to
4259 refs to unconstrained commons.
4261 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4264 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4265 look through SAVE_EXPRs with non-side-effects argument. Adjust
4267 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4270 2018-04-06 Richard Biener <rguenther@suse.de>
4273 * alias.c (find_base_term): New wrapper around find_base_term
4274 unwinding CSELIB_VAL_PTR changes.
4275 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4278 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4280 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4282 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4283 constant definitions.
4284 ("nop"): lr 0,0 -> nopr r0
4285 ("nop_lr0", "nop_lr1"): New insn definitions.
4287 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4289 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4290 NDS32_V3PUSH_AVAILABLE_P macro.
4292 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4293 Chung-Ju Wu <jasonwucj@gmail.com>
4295 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4296 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4297 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4298 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4299 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4300 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4301 * config/nds32/constraints.md: New constraints and checking for hard
4302 float configuration.
4303 * config/nds32/iterators.md: New mode iterator and attribute for hard
4304 float configuration.
4305 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4307 * config/nds32/nds32-fpu.md: New file.
4308 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4309 deal with hard float code generation.
4310 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4312 (abi_type, float_reg_number): New enum type.
4313 * config/nds32/nds32-predicates.c: New predicates for hard float.
4314 * config/nds32/nds32-protos.h: Declare functions for hard float.
4315 * config/nds32/nds32.c: Implementation for hard float configuration.
4316 * config/nds32/nds32.h: Definitions for hard float configuration.
4317 * config/nds32/nds32.md: Include hard float machine description and
4318 modify patterns for hard float configuration.
4319 * config/nds32/nds32.opt: New options for hard float configuration.
4320 * config/nds32/predicates.md: New predicates for hard float
4323 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4325 * common/config/nds32/nds32-common.c
4326 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4328 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4331 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4332 CONSTRUCTOR_ELT (ctor, ...)->value.
4334 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4337 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4339 2018-04-05 Tom de Vries <tom@codesourcery.com>
4342 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4345 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4346 Kito Cheng <kito.cheng@gmail.com>
4348 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4349 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4350 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4351 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4353 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4354 Kito Cheng <kito.cheng@gmail.com>
4356 * config/nds32/constraints.md (Ufe): New memory constraint.
4357 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4358 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4359 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4361 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4362 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4364 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4366 * config/nds32/nds32.md: Use optimize_size in the condition for
4367 alu-shift instructions.
4369 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4371 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4373 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4375 * config/nds32/nds32.md (negsi2): Refine pattern.
4377 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4378 Chung-Ju Wu <jasonwucj@gmail.com>
4380 * config/nds32/iterators.md (shift_rotate): New code iterator.
4381 (shift): New code attribute.
4382 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4383 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4384 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4385 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4386 bit-wise operations.
4387 (andsi3, *andsi3): Ditto.
4388 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4389 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4390 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4391 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4392 nds32_ior_operand, nds32_xor_operand): New predicates.
4394 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4396 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4397 (addsi3, subsi3): ... this.
4399 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4401 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4403 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4405 * config/nds32/nds32.md: Adjust indention.
4407 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4409 * config/nds32/nds32.md (feature): New attribute.
4411 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4413 * config/nds32/nds32.md (subtype): New attribute.
4415 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4418 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4419 expansion to perform a bitwise AND of the argument followed by a
4420 boolean negation of the result.
4422 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4424 PR rtl-optimization/84878
4425 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4426 the basic block. Assert the use reference is not artificial and that
4427 it has an associated insn.
4429 2018-04-04 Michael Matz <matz@suse.de>
4431 * builtins.c (compute_objsize): Pass correct operand
4432 to array_at_struct_end_p.
4434 2018-04-04 Richard Biener <rguenther@suse.de>
4437 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4438 from contexts for DINFO_LEVEL_TERSE and below.
4440 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4442 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4443 resiter_operand condition.
4444 * config/nds32/nds32.md (*move<mode>): Ditto.
4446 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4447 Monk Chiang <sh.chiang04@gmail.com>
4449 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4451 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4453 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4455 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4456 Kito Cheng <kito.cheng@gmail.com>
4458 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4459 nds32_cond_code_str, output_cond_branch,
4460 output_cond_branch_compare_zero, nds32_expand_cbranch,
4461 nds32_expand_cstore, nds32_expand_movcc,
4462 nds32_output_cbranchsi4_equality_zero,
4463 nds32_output_cbranchsi4_equality_reg,
4464 nds32_output_cbranchsi4_equality_reg_or_const_int,
4465 nds32_output_cbranchsi4_greater_less_zero: New functions.
4466 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4467 nds32_expand_cstore, nds32_expand_movcc,
4468 nds32_output_cbranchsi4_equality_zero,
4469 nds32_output_cbranchsi4_equality_reg,
4470 nds32_output_cbranchsi4_equality_reg_or_const_int,
4471 nds32_output_cbranchsi4_greater_less_zero): Declare.
4472 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4473 nds32_rimm11s_operand): New predicates.
4474 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4475 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4478 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4480 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4481 * config/nds32/nds32.md: Ditto.
4482 * config/nds32/pipelines.md: Ditto.
4484 2018-04-04 Richard Biener <rguenther@suse.de>
4486 PR tree-optimization/85168
4487 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4488 propagating abnormals.
4490 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4492 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4494 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4495 Kito Cheng <kito.cheng@gmail.com>
4497 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4498 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4499 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4500 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4501 * config/nds32/nds32.md (sibcall_internal): New.
4502 (sibcall_register): Remove.
4503 (sibcall_immediate): Remove.
4504 (sibcall_value_internal): New.
4505 (sibcall_value_register): Remove.
4506 (sibcall_value_immediate): Remove.
4507 * config/nds32/predicates.md (nds32_general_register_operand): New.
4508 (nds32_call_address_operand): New.
4510 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4512 PR rtl-optimization/85167
4513 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4514 bb_defs if *split_p, instead preinitialize it to NULL.
4516 PR tree-optimization/85156
4517 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4518 evaluating the argument multiple times.
4520 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4522 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4524 (_mm_cvtpd_ps): Likewise.
4525 (_mm_cvttpd_epi32): Likewise.
4526 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4527 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4528 vector, pixel, and bool following altivec.h include.
4530 2018-04-03 Martin Sebor <msebor@redhat.com>
4532 * doc/extend.texi (Common Function Attributes): Clarify.
4533 (const attribute): Likewise.
4534 (pure attribute): Likewise.
4536 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4539 * config/i386/i386.c (ix86_expand_vector_set): Use
4540 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4542 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4544 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4545 instructions when changing rounding bits to preserve precision bits
4546 in the x87 control word.
4548 2018-04-03 Martin Liska <mliska@suse.cz>
4550 PR tree-optimization/82491
4551 * rtl.h (strip_offset_and_add): Replace += suboffset with
4552 poly_uint64 () + suboffset.
4554 2018-03-29 Martin Liska <mliska@suse.cz>
4555 Martin Jambor <mjambor@suse.cz>
4558 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4559 param_type is not an integral or pointer type.
4561 2018-04-03 Richard Biener <rguenther@suse.de>
4563 * sese.h (recompute_all_dominators): Remove.
4565 2018-04-02 Martin Sebor <msebor@redhat.com>
4567 * doc/invoke.texi (-Wrestrict): Fix typos.
4569 2018-04-02 Jim Wilson <jimw@sifive.com>
4571 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4572 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4573 (<optab>di3, <optab>si3_extend): Likewise.
4574 (<optab>si3_mask, <optab>si3_mask_1): New.
4575 (<optab>di3_mask, <optab>di3_mask_1): New.
4576 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4577 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4578 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4580 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4582 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4585 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4587 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4588 (nds32_canonicalize_comparison): New function.
4590 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4591 Kito Cheng <kito.cheng@gmail.com>
4592 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4594 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4595 * config/nds32/constants.md (unspec_volatile_element): Add
4596 UNSPEC_VOLATILE_RELAX_GROUP.
4597 * config/nds32/nds32-relax-opt.c: New file.
4598 * config/nds32/nds32-predicates.c
4599 (nds32_symbol_load_store_p): New function.
4600 * config/nds32/nds32-protos.h
4601 (nds32_symbol_load_store_p): Declare function.
4602 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4603 * config/nds32/nds32.c
4604 (nds32_register_pass): New function to register pass.
4605 (nds32_register_passes): New function to register passes.
4606 * config/nds32/nds32.md (relax_group): New pattern.
4607 * config/nds32/nds32.opt (mrelax-hint): New option.
4608 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4610 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4612 * config/nds32/t-nds32: Modify files dependency.
4614 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4616 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4617 (PROFILE_HOOK): Define its implementation.
4619 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4621 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4622 type and 32-bit size.
4624 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4627 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4628 (V_128_256): New mode iterator.
4629 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4630 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4631 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4633 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4634 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4636 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4639 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4640 NaN inputs correctly.
4642 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4645 * config/rs6000/vsx.md (??r): New mode attribute.
4646 (*vsx_mov<mode>_64bit): Use it.
4647 (*vsx_mov<mode>_32bit): Likewise.
4649 2018-03-30 Martin Sebor <msebor@redhat.com>
4651 PR tree-optimization/84818
4652 * builtins.c (check_access): Use warning_n.
4654 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4657 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4659 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4662 2018-03-30 Julia Koval <julia.koval@intel.com>
4665 * x86-tune.def (movx, partial_reg_dependency): Enable for
4668 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4671 * lra-constraints.c (process_alt_operands): Move setting
4672 this_alternative_matches below.
4674 2018-03-29 Martin Liska <mliska@suse.cz>
4677 * doc/invoke.texi: Document how LTO works with debug info.
4678 Describe auto-load support of binutils. Mention 'x86-64'
4679 as valid option value of -march option.
4681 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4683 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4686 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4687 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4688 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4691 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4694 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4695 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4696 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4697 for RS6000_BTM_POWERPC64.
4698 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4699 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4700 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4705 2018-03-28 Carl Love <cel@us.ibm.com>
4708 2017-09-27 Carl Love <cel@us.ibm.com>
4710 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4711 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4712 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4715 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4717 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4718 instead of __vector bool.
4719 (_mm_max_pu8): Likewise.
4720 (_mm_min_pi16): Likewise.
4722 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4725 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4726 (DIVWEUO): Likewise.
4728 (DIVDEUO): Likewise.
4729 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4730 DIVWEUO and DIVDEUO.
4731 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4732 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4733 (div_extend): Likewise.
4734 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4736 (__builtin_divweuo): Likewise.
4737 (__builtin_divdeo): Likewise.
4738 (__builtin_divdeuo): Likewise.
4740 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4743 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4744 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4746 PR tree-optimization/82004
4747 * gimple-match-head.c (optimize_pow_to_exp): New function.
4748 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4749 Don't fold to exp if optimize_pow_to_exp is false.
4751 2018-03-28 Martin Liska <mliska@suse.cz>
4754 * calls.c (initialize_argument_information): Fix trailing space.
4755 * common.opt: Fix typo and provide better explanation for
4756 -fsanitize-coverage option.
4757 * config/i386/i386.opt: Fix typo.
4759 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4760 Martin Liska <mliska@suse.cz>
4763 * gimplify.c (asan_poison_variable): Don't do the check for
4764 gimplify_omp_ctxp here.
4765 (gimplify_decl_expr): Do it here.
4766 (gimplify_target_expr): Likewise.
4768 2018-03-28 Martin Liska <mliska@suse.cz>
4771 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4772 chkp_type_bounds_count if MPX is not enabled.
4774 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4776 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4778 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4781 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4782 function to create the function decl for complex long double
4783 multiply and divide for -mabi=ieeelongdouble.
4784 (init_float128_ieee): Call it.
4786 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4789 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4790 -fcf-protection=branch -mibt.
4791 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4793 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4796 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4798 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4801 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4802 extern array declarations.
4804 2018-03-27 Richard Biener <rguenther@suse.de>
4807 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4808 explicit single_use checks.
4810 2018-03-27 Richard Biener <rguenther@suse.de>
4812 PR tree-optimization/85082
4813 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4816 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4818 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4819 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4820 Turn on fasynchronous-unwind-tables and funwind-tables.
4822 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4825 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4826 (*bmi_blsr_<mode>_ccz): Ditto.
4828 2018-03-26 Tom de Vries <tom@codesourcery.com>
4830 PR tree-optimization/85063
4831 * omp-general.c (offloading_function_p): New function. Factor out
4833 * omp-offload.c (pass_omp_target_link::gate): ... here.
4834 * omp-general.h (offloading_function_p): Declare.
4835 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4836 with attribute omp declare target for offloading functions.
4838 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4840 PR tree-optimization/84005
4841 * tree-data-ref.h (get_base_for_alignment): Declare.
4842 * tree-data-ref.c (get_base_for_alignment_1): New function.
4843 (get_base_for_alignment): Likewise.
4844 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4845 get_base_for_alignment to find a suitable base object, instead
4846 of always using drb->base_address.
4848 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4851 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4852 known size by default.
4854 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4857 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4858 and non BLKmode operands.
4860 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4863 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4864 Clean up attributes.
4866 2018-03-23 Richard Biener <rguenther@suse.de>
4869 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4870 we are going to emit early debug for LTO.
4872 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4875 * function.c (match_asm_constraints_1): Don't optimize if input
4876 doesn't satisfy general_operand predicate for output's mode.
4879 * alias.c (write_dependence_p): Don't require for x_canonicalized
4880 non-VOIDmode if x has VOIDmode.
4883 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4884 just don't try to optimize it rather than assert it never happens.
4886 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4888 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4889 macro expansions for definition of ST_INTERNAL_<mode> and
4890 LD_INTERNAL_<mode> builtins.
4891 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4893 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4895 (altivec_expand_st_builtin): Likewise.
4896 (altivec_expand_builtin): Remove calls to deleted functions.
4897 (rs6000_address_for_altivec): Delete this function.
4898 * config/rs6000/vector.md: Remove expands for
4899 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4901 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4904 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4905 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4906 re-computing once computed.
4907 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4908 (arm_init_machine_status): Initialize
4909 machine->static_chain_stack_bytes.
4911 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4914 * doc/extend.texi: Add four new prototypes for vec_ld.
4915 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4916 definitions for more logical presentation.
4917 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4918 entries for V1TI variants of __builtin_altivec_ld builtin.
4919 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4920 handling of V1TI variant of LVX icode pattern.
4921 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4922 (rs6000_gimple_fold_builtin): Likewise.
4923 (altivec_init_builtins): Add code to define
4924 __builtin_altivec_lvx_v1ti function.
4926 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4929 * function.c (match_asm_constraints_1): Don't do the optimization
4930 if input isn't a REG, SUBREG, MEM or constant.
4932 2018-03-22 Tom de Vries <tom@codesourcery.com>
4934 PR tree-optimization/84956
4935 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4936 bb_has_abnormal_pred.
4938 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4941 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4942 DECL_INITIAL (decl) to decl at the end.
4943 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4946 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4948 * doc/extend.texi (__builtin_tgmath): Document when complex
4949 integer types are treated as _Complex _Float64.
4951 2018-03-21 Tom de Vries <tom@codesourcery.com>
4953 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4955 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4957 PR tree-optimization/84960
4958 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4959 if it is ENTRY block, move them into single succ of ENTRY in that case.
4961 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4963 PR tree-optimization/84811
4964 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4965 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4966 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4967 (known_subrange_p): Update accordingly. Cast each value involved
4968 in the size comparison, rather than casting the result of the
4971 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4973 PR tree-optimization/84982
4974 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4975 by flipping the least significant bit rather than all bits from
4976 bitpos to bitpos + bitsize - 1.
4978 2018-03-21 Nathan Sidwell <nathan@acm.org>
4980 * doc/extend.texi (Deprecated Features): Remove mention of
4981 long-deleted deprecations.
4983 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4986 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4987 * configure: Regenerate.
4989 2018-03-21 Tom de Vries <tom@codesourcery.com>
4991 PR tree-optimization/83126
4992 * tree-parloops.c (num_phis): New function.
4993 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4995 2018-03-21 Nathan Sidwell <nathan@acm.org>
4997 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4998 mention anon-struct/union members and trailing attributes.
5000 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5002 PR tree-optimization/84969
5003 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5004 builtin memset partitions if they set different rhs values.
5006 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5008 PR rtl-optimization/84989
5009 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5010 VEC_DUPLICATE with scalar result mode.
5012 2018-03-21 Martin Liska <mliska@suse.cz>
5015 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5016 not intended return statement.
5018 2018-03-21 Martin Liska <mliska@suse.cz>
5021 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5022 (chkp_find_bound_slots_1): Limit number of iterations.
5024 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5027 * Minor grammar fixes for x86 options.
5029 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5032 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5033 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5035 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5038 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5039 (altivec_lvx_<mode>_1op): Likewise.
5040 (altivec_stvx_<mode>_2op): Likewise.
5041 (altivec_stvx_<mode>_1op): Likewise.
5042 (altivec_lvx_<VM2:mode>): New define_expand.
5043 (altivec_stvx_<VM2:mode>): Likewise.
5044 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5045 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5046 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5047 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5048 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5049 (rs6000_gen_lvx): Likewise.
5050 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5051 (altivec_expand_stv_builtin): Likewise.
5052 (altivec_expand_builtin): Likewise.
5053 * config/rs6000/vector.md: Likewise.
5055 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5058 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5061 2018-03-20 Richard Biener <rguenther@suse.de>
5064 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5065 sign-conversions as zero, fall back to standard scalar_stmt
5068 2018-03-20 Martin Liska <mliska@suse.cz>
5071 * predict.c (rebuild_frequencies): Handle case when we have
5072 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5074 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5077 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5078 flag_section_anchors.
5079 * varasm.c (use_blocks_for_decl_p): Remove hack for
5080 dw2_force_const_mem.
5083 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5085 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5086 be created, use lowpart_subreg of operands[0] rather than operands[0]
5088 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5089 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5090 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5091 and n constraint instead of aarch64_shift_imm_di and Usd.
5092 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5093 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5095 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5098 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5099 to favor GPR over NEON registers.
5100 (<shift>di3_neon): Likewise.
5102 2018-03-20 Tom de Vries <tom@codesourcery.com>
5105 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5106 (nvptx_process_pars): Emit bar.sync asap and alap.
5108 2018-03-20 Tom de Vries <tom@codesourcery.com>
5111 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5112 seen_label if seen_label is already set.
5114 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5117 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5118 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5119 Use 1U instead of 1. Formatting fixes.
5122 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5123 instead of TREE_TYPE (s1) for the return value.
5125 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5127 PR tree-optimization/84946
5128 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5129 bitsize + bitsize in poly_uint64 rather than poly_int64.
5132 * dwarf2asm.c: Include fold-const.c.
5133 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5134 of decl rather than decl itself.
5136 PR rtl-optimization/84643
5137 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5139 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5142 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5143 calling assemble_variable.
5145 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5148 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5149 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5151 2018-03-19 Jim Wilson <jimw@sifive.com>
5154 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5155 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5156 (riscv_first_stack_step): Likewise.
5157 (riscv_option_override): Use STACK_BOUNDARY instead of
5159 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5161 (BIGGEST_ALIGNMENT): Set to 128.
5162 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5163 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5166 2018-03-19 Richard Biener <rguenther@suse.de>
5168 PR tree-optimization/84933
5169 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5170 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5172 2018-03-19 Richard Biener <rguenther@suse.de>
5174 PR tree-optimization/84859
5175 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5176 (cond_if_else_store_replacement): Perform sinking operation on
5177 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5178 Generalize what a BB with a single eligible store is.
5180 2018-03-19 Richard Biener <rguenther@suse.de>
5182 PR tree-optimization/84929
5183 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5184 chrec_is_positive against non-chrec arg.
5186 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5189 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5191 2018-03-18 Martin Liska <mliska@suse.cz>
5193 PR rtl-optimization/84635
5194 * regrename.c (build_def_use): Use matches_mode only when
5197 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5199 PR tree-optimization/84913
5200 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5201 vectorize chains of COND_EXPRs.
5203 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5205 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5207 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5209 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5211 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5213 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5215 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5216 Kito Cheng <kito.cheng@gmail.com>
5218 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5219 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5220 (nds32_adjust_reg_alloc_order): New function.
5221 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5223 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5225 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5226 nds32_print_operand, nds32_print_operand_address): Use
5227 HOST_WIDE_INT_PRINT_DEC instead.
5229 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5231 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5233 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5236 * config/i386/i386.c (initial_ix86_tune_features,
5237 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5239 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5240 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5241 rather than 1u << ix86_tune. Formatting fix.
5242 (ix86_option_override_internal): Change ix86_arch_mask from
5243 unsigned int to unsigned HOST_WIDE_INT, initialize to
5244 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5245 (ix86_function_specific_restore): Likewise.
5247 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5250 * postreload.c (reload_combine_recognize_pattern): Perform
5251 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5252 truncate_int_for_mode the result for the destination's mode.
5255 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5256 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5258 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5261 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5264 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5265 regno_allocno_class_array and sorted_pseudos.
5266 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5267 insns where regno is used.
5269 2018-03-16 Martin Liska <mliska@suse.cz>
5272 * multiple_target.c (create_dispatcher_calls): Redirect
5273 reference in the symbol table.
5275 2018-03-16 Martin Liska <mliska@suse.cz>
5278 * multiple_target.c (create_dispatcher_calls): Redirect also
5281 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5285 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5286 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5287 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5289 2018-03-16 Julia Koval <julia.koval@intel.com>
5291 * doc/invoke.texi (Skylake Server): Add CLWB.
5292 Cannonlake): Remove CLWB.
5294 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5296 PR tree-optimization/84841
5297 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5299 (FLOAT_ONE_CONST_TYPE): Define.
5300 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5301 (sort_by_operand_rank): Put entries with higher constant_type last
5302 rather than first to match comments.
5304 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5306 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5309 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5312 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5313 check and use address of target always.
5315 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5318 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5319 (indirect_thunk_bnd_needed): Likewise.
5320 (indirect_thunks_used): Likewise.
5321 (indirect_thunks_bnd_used): Likewise.
5322 (indirect_return_needed): New.
5323 (indirect_return_bnd_needed): Likewise.
5324 (output_indirect_thunk_function): Add a bool argument for
5326 (output_indirect_thunk_function): Don't generate alias for
5327 function return thunk.
5328 (ix86_code_end): Call output_indirect_thunk_function to generate
5329 function return thunks.
5330 (ix86_output_function_return): Set indirect_return_bnd_needed
5331 and indirect_return_needed instead of indirect_thunk_bnd_needed
5332 and indirect_thunk_needed.
5334 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5336 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5337 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5338 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5340 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5341 Paul Hua <paul.hua.gm@gmail.com>
5344 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5346 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5348 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5349 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5352 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5355 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5356 instead of GET_MODE_SIZE when comparing Units.
5358 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5361 * varasm.c (hash_section): Return an unchangeble hash value
5362 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5363 Return !aarch64_can_use_per_function_literal_pools_p ().
5365 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5368 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5369 rather than address of cmode as last argument to prepare_cmp_insn.
5371 2018-03-15 Julia Koval <julia.koval@intel.com>
5373 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5374 F_AVX512VNNI, F_AVX512BITALG): New.
5376 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5379 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5380 insn for floating-point loads and stores.
5382 2018-03-14 Carl Love <cel@us.ibm.com>
5384 * config/rs6000/rs6000-c.c: Add macro definitions for
5385 ALTIVEC_BUILTIN_VEC_PERMXOR.
5386 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5387 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5388 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5389 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5391 * config/doc/extend.texi: Add prototypes for vec_permxor.
5393 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5396 * diagnostic-show-locus.c (class layout_point): Convert m_line
5397 from int to linenum_type.
5398 (line_span::comparator): Use linenum "compare" function when
5399 comparing line numbers.
5400 (test_line_span): New function.
5401 (layout_range::contains_point): Convert param "row" from int to
5403 (layout_range::intersects_line_p): Likewise.
5404 (layout::will_show_line_p): Likewise.
5405 (layout::print_source_line): Likewise.
5406 (layout::should_print_annotation_line_p): Likewise.
5407 (layout::print_annotation_line): Likewise.
5408 (layout::print_leading_fixits): Likewise.
5409 (layout::annotation_line_showed_range_p): Likewise.
5410 (struct line_corrections): Likewise for field m_row.
5411 (line_corrections::line_corrections): Likewise for param "row".
5412 (layout::print_trailing_fixits): Likewise.
5413 (layout::get_state_at_point): Likewise.
5414 (layout::get_x_bound_for_row): Likewise.
5415 (layout::print_line): Likewise.
5416 (diagnostic_show_locus): Likewise for locals "last_line" and
5418 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5419 * input.c (selftest::test_linenum_comparisons): New function.
5420 (selftest::input_c_tests): Call it.
5421 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5422 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5423 * selftest.h (ASSERT_GT): New macro.
5424 (ASSERT_GT_AT): New macro.
5425 (ASSERT_LT): New macro.
5426 (ASSERT_LT_AT): New macro.
5428 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5430 PR rtl-optimization/84780
5431 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5433 2018-03-14 Martin Liska <mliska@suse.cz>
5435 * tree.c (record_node_allocation_statistics): Use
5436 get_stats_node_kind.
5437 (get_stats_node_kind): New function extracted from
5438 record_node_allocation_statistics.
5439 (free_node): Use get_stats_node_kind.
5441 2018-03-14 Richard Biener <rguenther@suse.de>
5443 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5444 that the value-set of ANTIC_IN doesn't grow.
5447 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5449 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5450 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5452 2018-03-14 Julia Koval <julia.koval@intel.com>
5454 * config.gcc (icelake-client, icelake-server): New.
5456 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5457 (initial_ix86_arch_features): Ditto.
5458 (PTA_SKYLAKE): Add SGX.
5459 (PTA_ICELAKE): Remove.
5460 (PTA_ICELAKE_CLIENT): New.
5461 (PTA_ICELAKE_SERVER): New.
5462 (ix86_option_override_internal): Split up icelake on icelake client and
5464 (get_builtin_code_for_version): Ditto.
5465 (fold_builtin_cpu): Ditto.
5466 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5467 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5468 * config/i386/i386.h (processor_type): Ditto.
5469 * doc/invoke.texi: Ditto.
5471 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5474 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5475 INTEGER_CST offset, add it together with bitpos / 8 and
5476 sign extend based on POINTER_SIZE.
5480 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5483 * config/i386/constraints.md (Yc): New register constraint.
5484 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5485 Use Yc constraint for alternative 2 of operand 0. Remove
5486 preferred_for_speed attribute.
5488 2018-03-14 Richard Biener <rguenther@suse.de>
5490 PR tree-optimization/84830
5491 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5492 with the old one to avoid oscillations.
5494 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5497 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5499 (assign_by_spills): Return a flag of reload assignment failure.
5500 Do not process the reload assignment failures. Do not spill other
5501 reload pseudos if they has the same reg class. Update n if
5503 (lra_assign): Add a return arg. Set up from the result of
5504 assign_by_spills call.
5505 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5506 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5507 usage_insns if it is not NULL.
5508 (spill_hard_reg_in_range): New function.
5509 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5510 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5511 function prototypes.
5512 (lra_assign): Change prototype.
5513 * lra.c (lra): Add code to deal with fails by splitting hard reg
5516 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5518 * config/riscv/riscv.opt (mrelax): New option.
5519 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5520 "norelax" when riscv_mrelax is disabled.
5521 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5523 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5526 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5527 reassociation for int modes.
5529 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5531 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5532 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5534 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5535 * config/aarch64/aarch64-sve.md
5536 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5537 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5538 (*extend<mode><Vwide>2): Rename to...
5539 (aarch64_sve_extend<mode><Vwide>2): ...this.
5540 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5541 renaming the old pattern to...
5542 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5544 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5545 define_expand, renaming the old pattern to...
5546 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5547 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5548 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5549 account when deciding which SVE instruction the optab should use.
5550 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5552 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5554 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5555 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5556 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5557 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5558 (tlsdesc_small_<mode>): Turn a define_expand and use
5559 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5560 (tlsdesc_small_advsimd_<mode>): ...this.
5561 (tlsdesc_small_sve_<mode>): New pattern.
5563 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5565 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5566 (UNSPEC_UMUL_HIGHPART): New constants.
5567 (MUL_HIGHPART): New int iteraor.
5568 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5569 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5571 (*<su>mul<mode>3_highpart): New define_insn.
5573 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5576 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5579 2018-03-13 Martin Liska <mliska@suse.cz>
5582 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5584 (sem_item_optimizer::~sem_item_optimizer): Release it.
5585 (sem_item_optimizer::merge_classes): Register variable aliases.
5586 (sem_item_optimizer::fixup_pt_set): New function.
5587 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5588 * ipa-icf.h: Declare new variables and functions.
5590 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5593 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5594 integer_pow2p@2 and test integer_pow2p in condition.
5595 (A < 0 ? C : 0): Similarly for @1.
5598 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5599 characters starting at p contain '\0' character, don't look beyond
5603 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5604 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5607 * reg-stack.c (change_stack): Change update_end var from int to
5608 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5609 also call set_block_for_insn on the newly added insns and rescan.
5612 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5613 on the last operand.
5616 * tree.c (stabilize_reference_1): Return save_expr (e) for
5617 STATEMENT_LIST even if it doesn't have side-effects.
5619 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5621 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5623 2018-03-12 Renlin Li <renlin.li@arm.com>
5625 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5626 aarch64_output_scalar_simd_mov_immediate.
5628 2018-03-12 Martin Sebor <msebor@redhat.com>
5630 PR tree-optimization/83456
5631 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5632 for perfectly overlapping calls to memcpy.
5633 (gimple_fold_builtin_memory_chk): Same.
5634 (gimple_fold_builtin_strcpy): Handle no-warning.
5635 (gimple_fold_builtin_stxcpy_chk): Same.
5636 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5638 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5640 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5641 parameter. Use it for SFmode.
5642 (rs6000_function_arg_advance_1): Adjust.
5643 (rs6000_function_arg): Adjust.
5644 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5646 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5648 PR rtl-optimization/84169
5649 PR rtl-optimization/84780
5650 * combine.c (can_combine_p): Check for a 2-insn combination whether
5651 the destination register is used between the two insns, too.
5653 2018-03-12 Richard Biener <rguenther@suse.de>
5655 PR tree-optimization/84803
5656 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5657 for refs DR analysis didn't process.
5659 2018-03-12 Richard Biener <rguenther@suse.de>
5661 PR tree-optimization/84777
5662 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5663 force-vectorize loops ignore whether we are optimizing for size.
5665 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5667 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5668 (TARGET_MD_ASM_ADJUST): Define.
5670 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5671 Kito Cheng <kito.cheng@gmail.com>
5672 Chung-Ju Wu <jasonwucj@gmail.com>
5674 * config/nds32/nds32.c (nds32_compute_stack_frame,
5675 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5676 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5677 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5678 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5679 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5680 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5681 * config/nds32/nds32.md (prologue, epilogue): Use macro
5682 NDS32_V3PUSH_AVAILABLE_P to do checking.
5684 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5687 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5688 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5689 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5690 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5691 addition of most attributes on !orig_type_die or the attribute not
5692 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5694 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5695 Chung-Ju Wu <jasonwucj@gmail.com>
5697 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5699 * config/nds32/nds32.opt (mvh): New option.
5701 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5702 Chung-Ju Wu <jasonwucj@gmail.com>
5704 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5706 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5707 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5710 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5711 Chung-Ju Wu <jasonwucj@gmail.com>
5713 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5715 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5716 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5718 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5719 Kito Cheng <kito.cheng@gmail.com>
5720 Chung-Ju Wu <jasonwucj@gmail.com>
5722 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5723 UNSPEC_FFMISM and UNSPEC_FLMISM.
5724 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5725 for ffb, ffmism and flmism.
5726 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5727 (unspec_ffmism): Ditto.
5728 (unspec_flmism): Ditto.
5729 (nds32_expand_builtin_impl): Check if string extension is available.
5730 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5731 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5733 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5736 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5739 * lra-assigns.c (assign_by_spills): Return a flag of reload
5740 assignment failure. Do not process the reload assignment
5741 failures. Do not spill other reload pseudos if they has the same
5743 (lra_assign): Add a return arg. Set up from the result of
5744 assign_by_spills call.
5745 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5746 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5747 usage_insns if it is not NULL.
5748 (spill_hard_reg_in_range): New function.
5749 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5750 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5751 function prototypes.
5752 (lra_assign): Change prototype.
5753 * lra.c (lra): Add code to deal with fails by splitting hard reg
5756 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5759 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5761 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5764 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5765 (dw_val_node): Add val_symbolic_view.
5766 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5767 (symview_upper_bound): New.
5768 (new_line_info_table): Initialize symviews_since_reset.
5769 (dwarf2out_source_line): Count symviews_since_reset and set
5770 symview_upper_bound.
5771 (dw_val_equal_p): Handle symview.
5772 (add_AT_symview): New.
5773 (print_dw_val): Handle symview.
5774 (attr_checksum, attr_checksum_ordered): Likewise.
5775 (same_dw_val_p, size_of_die): Likewise.
5776 (value_format, output_die): Likewise.
5777 (add_high_low_attributes): Use add_AT_symview for entry_view.
5778 (dwarf2out_finish): Reset symview_upper_bound, clear
5781 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5784 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5785 Add strict argument and use it.
5786 (rs6000_split_multireg_move): Update for new strict argument.
5787 (mem_operand_gpr): Disallow all non-offsettable addresses.
5788 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5790 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5793 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5794 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5795 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5798 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5799 decl, use remap_type if we want to use the type.
5801 2018-03-09 Martin Sebor <msebor@redhat.com>
5803 PR tree-optimization/84526
5804 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5806 (builtin_access::generic_overlap): Be prepared to handle non-array
5809 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5811 PR rtl-optimization/84682
5812 * lra-constraints.c (process_address_1): Check is_address flag
5813 for address constraints.
5814 (process_alt_operands): Likewise.
5815 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5816 preprocess_constraints.
5817 * recog.h (preprocess_constraints): Add oploc parameter.
5819 * recog.c (preprocess_constraints): Test address_operand for
5820 CT_ADDRESS constraints.
5822 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5825 * lra-assigns.c (assign_by_spills): Return a flag of reload
5826 assignment failure. Do not process the reload assignment
5827 failures. Do not spill other reload pseudos if they has the same
5829 (lra_assign): Add a return arg. Set up from the result of
5830 assign_by_spills call.
5831 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5832 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5833 usage_insns if it is not NULL.
5834 (spill_hard_reg_in_range): New function.
5835 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5836 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5837 function prototypes.
5838 (lra_assign): Change prototype.
5839 * lra.c (lra): Add code to deal with fails by splitting hard reg
5842 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5845 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5846 Accept complain bool parameter. Only emit errors if it is true.
5847 (arm_parse_cpu_option_name): Likewise.
5848 (arm_target_thumb_only): Adjust callers of the above.
5849 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5850 prototype to take a default true bool parameter.
5851 (arm_parse_arch_option_name): Likewise.
5853 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5854 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5858 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5859 * configure: Regenerate.
5860 * configure.ac ("linker --version-script option"): New.
5861 ("linker soname option"): New.
5863 2018-03-09 Richard Biener <rguenther@suse.de>
5865 PR tree-optimization/84775
5866 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5867 immediate uses of predicate stmts and mark them modified.
5870 PR tree-optimization/84178
5871 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5873 (version_loop_for_if_conversion): Delay update_ssa call.
5874 (tree_if_conversion): Delay update_ssa until after predicate
5877 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5880 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5881 when the function accesses prior frames.
5883 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5886 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5887 gen_llsym, otherwise call maybe_gen_llsym.
5890 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5891 has ',' character inside of it.
5893 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5896 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5897 as clobbering CC_REGNUM.
5899 2018-03-08 Richard Biener <rguenther@suse.de>
5902 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5903 (follow_copies_to_constant): Do not follow SSA names registered
5906 2018-03-08 Richard Biener <rguenther@suse.de>
5908 PR tree-optimization/84178
5909 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5911 (version_loop_for_if_conversion): Delay update_ssa call.
5912 (tree_if_conversion): Delay update_ssa until after predicate
5915 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5917 PR tree-optimization/84178
5918 * tree-if-conv.c (release_bb_predicate): Remove the
5919 the assertion that the stmts have NULL use_ops.
5920 Discard the statements, asserting that they haven't
5921 yet been added to a BB.
5923 2018-03-08 Richard Biener <rguenther@suse.de>
5925 PR tree-optimization/84746
5926 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5927 (phi_translate): Pass in destination ANTIC_OUT set.
5928 (phi_translate_1): Likewise. For a simplified result lookup
5929 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5930 (phi_translate_set): Adjust.
5931 (do_pre_regular_insertion): Likewise.
5932 (do_pre_partial_partial_insertion): Likewise.
5934 2018-03-08 Martin Liska <mliska@suse.cz>
5936 PR gcov-profile/84735
5937 * doc/gcov.texi: Document usage of profile files.
5938 * gcov-io.h: Document changes in the format.
5940 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5944 * dwarf2out.c (struct dw_line_info_table): Update comments for
5946 (FORCE_RESET_NEXT_VIEW): New.
5947 (FORCE_RESETTING_VIEW_P): New.
5948 (RESETTING_VIEW_P): Check for -1 too.
5949 (ZERO_VIEW_P): Likewise.
5950 (new_line_info_table): Force-reset next view.
5951 (dwarf2out_begin_function): Likewise.
5952 (dwarf2out_source_line): Simplify zero_view_p initialization.
5953 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5954 view directly. Omit view when omitting .loc at line 0.
5956 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5958 PR tree-optimization/84740
5959 * tree-switch-conversion.c (process_switch): Call build_constructors
5960 only if info.phi_count is non-zero.
5962 PR tree-optimization/84739
5963 * tree-tailcall.c (find_tail_calls): Check call arguments against
5964 DECL_ARGUMENTS (current_function_decl) rather than
5965 DECL_ARGUMENTS (func) when checking for tail recursion.
5967 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5969 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5970 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5971 Volker Reichelt's entry and add entries for people that perform
5972 GCC fuzzy testing and report numerous bugs.
5974 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5977 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5978 readonly data in sdata, if that is disabled.
5979 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5980 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5981 -mreadonly-in-sdata option.
5983 2018-03-07 Martin Sebor <msebor@redhat.com>
5985 PR tree-optimization/84468
5986 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5987 basic block when looking for nul assignment.
5989 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5992 * except.h (output_function_exception_table): Adjust prototype.
5993 * except.c (output_function_exception_table): Remove FNNAME parameter
5994 and add SECTION parameter. Ouput one part of the table at a time.
5995 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5996 the first part of the exception table and emit unwind directives.
5997 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5998 (i386_pe_seh_cold_init): Likewise.
5999 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6000 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6001 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6002 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6003 * config/i386/winnt.c: Include except.h.
6004 (struct seh_frame_state): Add reg_offset, after_prologue and
6005 in_cold_section fields.
6006 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6007 (i386_pe_seh_cold_init): New function.
6008 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6009 to seh->in_cold_section.
6010 (seh_emit_push): Record the offset of the push.
6011 (seh_emit_save): Record the offet of the save.
6012 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6013 Test seh->after_prologue to disregard the epilogue.
6014 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6015 (i386_pe_end_cold_function): New function.
6017 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6020 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6021 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6024 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6026 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6027 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6030 * multiple_target.c: Include tree-inline.h and intl.h.
6031 (expand_target_clones): Diagnose and fail if node->definition and
6032 !tree_versionable_function_p (node->decl).
6034 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6036 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6038 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6039 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6040 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6042 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6045 * combine.c (try_combine): Use reg_or_subregno instead of handling
6046 just paradoxical SUBREGs and REGs.
6048 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6050 * config/arc/arc.c (arc_finalize_pic): Remove function.
6051 (arc_must_save_register): We use single base PIC register, remove
6052 checks to save/restore the PIC register.
6053 (arc_expand_prologue): Likewise.
6054 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6056 (arc_verify_short): Likewise.
6057 (arc_attr_type): Likewise.
6058 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6059 (walk_stores): Likewise.
6060 (arc_address_cost): Make it static.
6061 (arc_verify_short): Likewise.
6062 (branch_dest): Likewise.
6063 (arc_attr_type): Likewise.
6064 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6065 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6066 (arc_final_prescan_insn): Remove inserting the nops due to
6067 hardware hazards. It is done in reorg step.
6068 (insn_length_variant_t): Remove.
6069 (insn_length_parameters_t): Likewise.
6070 (arc_insn_length_parameters): Likewise.
6071 (arc_get_insn_variants): Likewise.
6072 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6074 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6077 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6080 PR tree-optimization/84687
6081 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6083 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6085 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6087 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6088 Rename to ppc_speculation_barrier.
6089 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6090 __builtin_ppc_speculation_barrier.
6092 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6095 * combine.c (combine_simplify_rtx): Don't try to simplify if
6096 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6099 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6101 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6102 to 32 bytes when compiling for POWER9.
6104 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6107 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6108 regparm >= 3 with no arg reg available also for calls with
6109 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6112 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6114 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6116 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6119 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6121 2018-03-05 Richard Biener <rguenther@suse.de>
6123 PR tree-optimization/84486
6124 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6125 When inserting a __builtin_assume_aligned call set the LHS
6126 SSA name alignment info accordingly.
6128 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6130 PR tree-optimization/84114
6131 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6132 Avoid reassociation of FLOAT_MODE addition.
6134 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6136 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6137 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6138 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6139 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6140 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6141 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6142 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6144 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6145 __builtin_ia32_wbinvd): New builtins.
6146 (SPECIAL_ARGS2): New.
6147 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6148 (SPECIAL_ARGS2): New.
6149 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6150 (ix86_valid_target_attribute_inner_p): Ditto.
6151 (ix86_init_mmx_sse_builtins): Add special_args2.
6152 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6153 TARGET_WBNOINVD_P): New.
6154 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6155 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6156 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6157 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6158 * config/i386/pconfigintrin.h: New file.
6159 * config/i386/wbnoinvdintrin.h: Ditto.
6160 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6161 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6163 2018-03-05 Richard Biener <rguenther@suse.de>
6165 PR tree-optimization/84670
6166 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6168 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6169 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6170 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6171 doesn't grow if all successors (recursively) were visited at least
6174 2018-03-05 Richard Biener <rguenther@suse.de>
6176 PR tree-optimization/84650
6177 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6178 if executed in the loop pipeline.
6180 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6182 * doc/configfiles.texi (Configuration Files): Move info about
6183 conditionalizing $target-protos.h to...
6184 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6185 differs from $target-protos.h.
6187 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6188 Chung-Ju Wu <jasonwucj@gmail.com>
6190 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6191 * config/nds32/nds32-multiple.md (setmemsi): Define.
6192 * config/nds32/nds32-memory-manipulation.c
6193 (nds32_gen_dup_4_byte_to_word_value): New.
6194 (emit_setmem_word_loop): New.
6195 (emit_setmem_byte_loop): New.
6196 (nds32_expand_setmem_loop): New.
6197 (nds32_expand_setmem_loop_v3m): New.
6198 (nds32_expand_setmem_unroll): New.
6199 (nds32_expand_setmem): New.
6201 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6202 Chung-Ju Wu <jasonwucj@gmail.com>
6204 * config/nds32/nds32-memory-manipulation.c
6205 (nds32_emit_load_store): New.
6206 (nds32_emit_post_inc_load_store): New.
6207 (nds32_emit_mem_move): New.
6208 (nds32_emit_mem_move_block): New.
6209 (nds32_expand_movmemsi_loop_unknown_size): New.
6210 (nds32_expand_movmemsi_loop_known_size): New.
6211 (nds32_expand_movmemsi_loop): New.
6212 (nds32_expand_movmemsi_unroll): New.
6213 (nds32_expand_movmemqi): Rename ...
6214 (nds32_expand_movmemsi): ... to this.
6215 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6216 (movmemsi): ... to this.
6217 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6218 (nds32_expand_movmemsi): ... to this.
6220 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6221 Monk Chiang <sh.chiang04@gmail.com>
6222 Chung-Ju Wu <jasonwucj@gmail.com>
6224 * config/nds32/nds32-protos.h
6225 (nds32_expand_load_multiple): New arguments.
6226 (nds32_expand_store_multiple): Ditto.
6227 (nds32_valid_multiple_load_store): Rename ...
6228 (nds32_valid_multiple_load_store_p): ... to this.
6229 * config/nds32/nds32-memory-manipulation.c
6230 (nds32_expand_load_multiple): Refine implementation.
6231 (nds32_expand_store_multiple): Ditto.
6232 * config/nds32/nds32-multiple.md
6233 (load_multiple): Update nds32_expand_load_multiple interface.
6234 (store_multiple): Update nds32_expand_store_multiple interface.
6235 * config/nds32/nds32-predicates.c
6236 (nds32_valid_multiple_load_store): Rename ...
6237 (nds32_valid_multiple_load_store_p): ... to this and refine
6239 * config/nds32/predicates.md
6240 (nds32_load_multiple_and_update_address_operation): New predicate.
6241 (nds32_store_multiple_and_update_address_operation): New predicate.
6243 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6244 Chung-Ju Wu <jasonwucj@gmail.com>
6246 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6247 (combo): New attribute.
6248 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6250 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6252 * config/nds32/nds32.opt: Change -mcmodel= default value.
6254 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6255 Monk Chiang <sh.chiang04@gmail.com>
6256 Chung-Ju Wu <jasonwucj@gmail.com>
6258 * config/nds32/constants.md (unspec_element): New enum.
6259 * config/nds32/constraints.md (Umw): New constraint.
6260 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6261 * config/nds32/nds32-intrinsic.md: Likewise.
6262 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6263 (nds32_valid_smw_lwm_base_p): New.
6264 (nds32_output_smw_single_word): New.
6265 (nds32_output_lmw_single_word): New.
6266 (nds32_expand_unaligned_load): New.
6267 (nds32_expand_unaligned_store): New.
6268 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6269 (nds32_output_smw_single_word): Declare.
6270 (nds32_output_lmw_single_word): Declare.
6271 (nds32_expand_unaligned_load): Declare.
6272 (nds32_expand_unaligned_store): Declare.
6273 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6274 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6275 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6276 NDS32_BUILTIN_UASTORE_DW.
6277 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6280 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6281 Kito Cheng <kito.cheng@gmail.com>
6282 Chung-Ju Wu <jasonwucj@gmail.com>
6284 * config/nds32/nds32-intrinsic.c
6285 (nds32_expand_builtin_null_ftype_reg): Delete.
6286 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6287 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6288 (nds32_read_argument): New.
6289 (nds32_legitimize_target): Ditto.
6290 (nds32_legitimize_argument): Ditto.
6291 (nds32_check_constant_argument): Ditto.
6292 (nds32_expand_unop_builtin): Ditto.
6293 (nds32_expand_unopimm_builtin): Ditto.
6294 (nds32_expand_binop_builtin): Ditto.
6295 (nds32_builtin_decl_impl): Ditto.
6296 (builtin_description): Ditto.
6297 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6298 (nds32_init_builtins_impl): Ditto.
6299 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6300 (nds32_builtin_decl): New.
6301 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6302 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6304 2018-03-02 Jeff Law <law@redhat.com>
6306 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6307 (redundant_insn, fill_simple_delay_slots): Likewise.
6308 (fill_slots_from_thread): Likewise.
6309 * resource.c (mark_referenced_resources): Likewise.
6310 (mark_set_resources, find_dead_or_set_registers): Likewise.
6312 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6314 * substring-locations.h (format_warning_va): Formatting fix for
6316 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6318 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6319 * substring-locations.c: Include intl.h.
6320 (format_warning_va): Turned into small wrapper around
6321 format_warning_n_va, renamed to ...
6322 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6323 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6325 (format_warning_at_substring_n): New function.
6326 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6327 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6328 format_warning_at_substring with just a shorter name instead of
6329 const function pointer.
6330 (fmtwarn_n): New function.
6331 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6332 appropriate, get rid of all the fmtstr temporaries, move conditionals
6333 with G_() wrapped string literals directly into fmtwarn arguments,
6334 cast dir.len to (int), formatting fixes.
6336 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6338 * doc/invoke.texi: Remove "Cilk Plus" references.
6340 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6341 Richard Biener <rguenther@suse.de>
6344 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6345 for error or warning attributes if CALL_FROM_THUNK_P is set.
6348 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6351 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6352 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6355 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6356 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6358 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6359 instead of -1U in last predictors element's probability member.
6361 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6364 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6365 arguments if they are comparable.
6367 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6369 PR tree-optimization/84634
6370 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6371 masks and masked_loop_p with a single loop_masks, making sure it's
6372 null for bb vectorization.
6374 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6376 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6377 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6378 loop->force_vectorize to check whether there is no alias.
6380 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6383 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6385 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6387 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6388 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6389 instead of a loop around prev_real_insn.
6390 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6394 * config/i386/i386.c (ix86_print_operand): Use conditional
6395 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6398 2018-03-02 Richard Biener <rguenther@suse.de>
6400 PR tree-optimization/84427
6401 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6402 (bitmap_set_subtract_values): Rewrite to handle multiple
6405 (prune_clobbered_mems): Likewise.
6406 (phi_translate): Take edge instead of pred/phiblock.
6407 (phi_translate_1): Likewise.
6408 (phi_translate_set): Likewise. Insert all translated
6409 exprs for a value into the set, keeping possibly multiple
6410 expressions per value.
6411 (compute_antic_aux): Adjust for phi_translate changes.
6412 When intersecting union the expressions and prune those
6413 not in the final value set, keeping possibly multiple
6414 expressions per value. Do not use value-insertion
6415 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6416 all expressions. Add verification that the value-sets
6417 only shrink during iteration.
6418 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6419 (do_pre_regular_insertion): Likewise.
6420 (do_pre_partial_partial_insertion): Likewise.
6422 2018-03-02 Richard Biener <rguenther@suse.de>
6425 * config/darwin.c (saved_debug_info_level): New static global.
6426 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6427 (darwin_asm_lto_end): Restore debug info generation settings.
6429 2018-03-01 Martin Liska <mliska@suse.cz>
6432 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6435 2018-03-01 Richard Biener <rguenther@suse.de>
6438 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6439 type annotation with LTO when debug was disabled at compile-time.
6441 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6443 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6445 (mips_final_postscan_insn): Likewise.
6447 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6449 PR rtl-optimization/84528
6450 * alias.c (init_alias_target): Add commentary.
6451 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6452 a unique base value if the frame pointer is not eliminated
6453 to the stack pointer.
6455 2018-03-01 Tom de Vries <tom@codesourcery.com>
6457 PR rtl-optimization/83327
6458 * lra-int.h (hard_regs_spilled_into): Declare.
6459 * lra.c (hard_regs_spilled_into): Define.
6460 (init_reg_info): Init hard_regs_spilled_into.
6461 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6462 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6463 (process_bb_lives): Handle hard_regs_spilled_into.
6464 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6465 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6467 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6469 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6470 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6471 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6472 * config/rs6000/aix72.h: New file.
6474 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6476 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6477 instead of warning_at with conditional singular and plural messages
6481 * stor-layout.c (update_alignment_for_field): For
6482 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6483 && !DECL_PACKED (field), do the alignment update, just use
6484 only desired_align instead of MAX (type_align, desired_align)
6486 (place_field): Don't do known_align < desired_align handling
6487 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6488 is non-NULL, instead do it after rli->prev_field handling and
6489 only if not within a bitfield word. For DECL_PACKED (field)
6490 use type_align of BITS_PER_UNIT.
6492 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6494 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6495 superfluous parentheses and trailing spaces.
6497 2018-02-28 Richard Biener <rguenther@suse.de>
6499 PR tree-optimization/84584
6500 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6501 SCoPs with fake exit edge.
6503 2018-02-28 Martin Liska <mliska@suse.cz>
6506 * timevar.c (timer::print): Fix format to properly print 100%
6509 2018-02-28 Richard Biener <rguenther@suse.de>
6512 * genmatch.c (capture_info::walk_match): Do not mark
6513 captured expressions without operands as expr_p given
6514 they act more like predicates and should be subject to
6515 "lost tail" side-effect preserving.
6517 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6519 PR rtl-optimization/81611
6520 * auto-inc-dec.c (attempt_change): Move dead note from
6521 mem_insn if it's the next use of regno
6522 (find_address): Take address use of reg holding
6523 non-incremented value. Add parm to limit search to the named
6525 (merge_in_block): Attempt to use a mem insn that is the next
6526 use of the original regno.
6528 2018-02-27 Martin Sebor <msebor@redhat.com>
6531 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6532 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6534 2018-02-27 Martin Sebor <msebor@redhat.com>
6536 PR translation/84207
6537 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6538 n argument to unsigned HOST_WIDE_INT.
6539 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6540 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6541 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6542 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6544 2018-02-27 Richard Biener <rguenther@suse.de>
6546 PR tree-optimization/84512
6547 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6548 Do not use the estimate returned from record_stmt_cost for
6549 the scalar iteration cost but sum properly using add_stmt_cost.
6551 2018-02-27 Richard Biener <rguenther@suse.de>
6553 PR tree-optimization/84466
6554 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6555 Adjust last change to less strictly validate use operands.
6557 2018-02-27 Martin Liska <mliska@suse.cz>
6559 PR gcov-profile/84548
6560 * gcov.c (process_file): Allow partial overlap and consider it
6561 also as group functions.
6562 (output_lines): Properly calculate range of lines for a group.
6564 2018-02-27 Martin Liska <mliska@suse.cz>
6566 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6567 'ggc' suffixes. Change first column width.
6568 (timer::print): Fix formatting of the column.
6570 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6572 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6573 preserve inline entry blocks for the sake of debug inline
6574 entry point markers alone.
6575 (remove_unused_locals): Suggest in comments a better place to
6576 force the preservation of inline entry blocks that are
6577 otherwise unused, but do not preserve them.
6579 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6581 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6583 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6586 * config/i386/constraints.md (Bs): Replace
6587 ix86_indirect_branch_register with
6588 TARGET_INDIRECT_BRANCH_REGISTER.
6590 * config/i386/i386.md (indirect_jump): Likewise.
6591 (tablejump): Likewise.
6592 (*sibcall_memory): Likewise.
6593 (*sibcall_value_memory): Likewise.
6594 Peepholes of indirect call and jump via memory: Likewise.
6595 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6596 (*sibcall_value_GOT_32): Likewise.
6597 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6598 (GOT_memory_operand): Likewise.
6599 (call_insn_operand): Likewise.
6600 (sibcall_insn_operand): Likewise.
6601 (GOT32_symbol_operand): Likewise.
6602 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6604 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6606 PR rtl-optimization/83496
6607 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6608 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6609 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6610 redundant insn, if any.
6611 (relax_delay_slots): Likewise.
6612 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6614 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6616 PR tree-optimization/83965
6617 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6618 that grouped statements are part of a reduction chain. Return
6619 true if the statement is not marked as a reduction itself but
6621 (vect_recog_dot_prod_pattern): Don't check whether the statement
6622 is part of a group here.
6623 (vect_recog_sad_pattern): Likewise.
6624 (vect_recog_widen_sum_pattern): Likewise.
6626 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6629 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6631 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6634 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6636 (ix86_output_indirect_function_return): New prototype.
6637 (ix86_split_simple_return_pop_internal): Likewise.
6638 * config/i386/i386.c (indirect_return_via_cx): New.
6639 (indirect_return_via_cx_bnd): Likewise.
6640 (indirect_thunk_name): Handle return va CX_REG.
6641 (output_indirect_thunk_function): Create alias for
6642 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6643 (ix86_output_indirect_jmp): Remove the bool argument.
6644 (ix86_output_indirect_function_return): New function.
6645 (ix86_split_simple_return_pop_internal): Likewise.
6646 * config/i386/i386.md (*indirect_jump): Don't pass false
6647 to ix86_output_indirect_jmp.
6648 (*tablejump_1): Likewise.
6649 (simple_return_pop_internal): Change it to define_insn_and_split.
6650 Call ix86_split_simple_return_pop_internal to split it for
6652 (simple_return_indirect_internal): Call
6653 ix86_output_indirect_function_return instead of
6654 ix86_output_indirect_jmp.
6656 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6659 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6660 memset and value initialization afterwards.
6662 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6664 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6666 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6669 * common/config/aarch64/aarch64-common.c
6670 (aarch_option_optimization_table[]): Switch
6671 off fomit-frame-pointer
6673 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6674 Chung-Ju Wu <jasonwucj@gmail.com>
6676 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6678 (store_multiple): Ditto.
6680 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6682 * config.gcc: Add --with-cpu support for nds32 target.
6683 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6684 * config/nds32/nds32.opt: Add -mcpu= option.
6686 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6688 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6689 isel=yes): Warn for these deprecated options.
6691 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6693 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6694 ISA_2_5_MASKS_EMBEDDED.
6696 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6698 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6699 p->max as pointers rather than using iterative_hash_expr.
6701 2018-02-23 Carl Love <cel@us.ibm.com>
6703 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6704 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6706 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6707 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6708 P8V_BUILTIN_VEC_VUNSIGNED2.
6710 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6713 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6714 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6716 * lra-constraints.c (curr_insn_transform): Set up
6717 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6719 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6720 * lra-eliminations.c (spill_pseudos): Ditto.
6721 (process_insn_for_elimination): Ditto.
6722 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6723 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6724 LRA_NON_CLOBBERED_ALT.
6726 2018-02-22 Martin Sebor <msebor@redhat.com>
6728 PR tree-optimization/84480
6729 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6730 to maybe_diag_stxncpy_trunc. Call it.
6731 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6732 from gimple_fold_builtin_strcpy. Print inlining stack.
6733 (handle_builtin_stxncpy): Print inlining stack.
6734 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6736 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6739 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6740 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6741 and -fcheck-pointer-bounds are used together.
6742 (indirect_thunk_prefix): New enum.
6743 (indirect_thunk_need_prefix): New function.
6744 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6745 "_nt" instead of "_bnd" for NOTRACK prefix.
6746 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6747 (output_indirect_thunk_function): Likewise.
6749 (ix86_code_end): Update output_indirect_thunk_function calls.
6750 (ix86_output_indirect_branch_via_reg): Replace
6751 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6752 (ix86_output_indirect_branch_via_push): Likewise.
6753 (ix86_output_function_return): Likewise.
6754 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6755 incompatible with -fcf-protection=branch and
6756 -fcheck-pointer-bounds.
6758 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6761 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6762 Change gcc_assert call to output_operand_lossage.
6764 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6766 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6768 2018-02-22 DJ Delorie <dj@redhat.com>
6769 Sebastian Perta <sebastian.perta@renesas.com>
6770 Oleg Endo <olegendo@gcc.gnu.org>
6772 * config/rx/rx.c (rx_rtx_costs): New function.
6773 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6775 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6777 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6779 2018-02-22 Martin Liska <mliska@suse.cz>
6782 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6783 Add "native" as a possible value.
6785 2018-02-22 Martin Liska <mliska@suse.cz>
6788 * config/i386/i386.c (ix86_option_override_internal):
6789 Add "native" as a possible value for -march and -mtune.
6791 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6794 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6795 to all type variants.
6797 PR tree-optimization/84503
6798 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6799 width as info->bitpos + info->bitsize - start.
6800 (merged_store_group::merge_overlapping): Simplify width computation.
6801 (check_no_overlap): New function.
6802 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6803 start + width and last_order of the group, fail if check_no_overlap
6805 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6806 to group if check_no_overlap fails.
6808 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6810 * config/rs6000/altivec.md: Delete contraint arguments to
6811 define_expand, define_split, and define_peephole2, and in
6812 define_insn_and_split if always unused.
6813 * config/rs6000/darwin.md: Ditto.
6814 * config/rs6000/dfp.md: Ditto.
6815 * config/rs6000/rs6000.md: Ditto.
6816 * config/rs6000/sync.md: Ditto.
6817 * config/rs6000/vector.md: Ditto.
6818 * config/rs6000/vsx.md: Ditto.
6820 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6822 * config/rs6000/altivec.md: Write output control strings as braced
6823 blocks instead of double-quoted strings.
6824 * config/rs6000/darwin.md: Ditto.
6825 * config/rs6000/rs6000.md: Ditto.
6826 * config/rs6000/vector.md: Ditto.
6827 * config/rs6000/vsx.md: Ditto.
6829 2018-02-21 Jason Merrill <jason@redhat.com>
6831 PR c++/84314 - ICE with templates and fastcall attribute.
6832 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6834 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6836 * ipa-cp.c (determine_versionability): Fix comment typos.
6838 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6841 * ipa-cp.c (determine_versionability): Do not version functions caling
6844 2018-02-21 Martin Liska <mliska@suse.cz>
6847 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6848 Add "native" as a possible value.
6849 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6850 the macro when native cpu detection is available.
6852 2018-02-21 Martin Liska <mliska@suse.cz>
6855 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6856 Add "native" as a possible value.
6857 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6858 when native cpu detection is available.
6860 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6861 Martin Sebor <msebor@redhat.com>
6863 PR tree-optimization/84478
6864 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6866 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6867 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6868 support which is conservatively correct, for 2 only stay conservative
6869 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6870 argument to the 2 argument get_range_strlen, adjust 6 arg
6871 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6873 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6874 (gimple_fold_builtin_strlen): Pass true as last argument to
6877 2018-02-20 Martin Sebor <msebor@redhat.com>
6880 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6881 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6882 (builtin_memref::builtin_memref): Factor out parts into
6883 set_base_and_offset and call it.
6885 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6888 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6889 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6890 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6891 search at the associated MODE_INT.
6893 2018-02-20 Jeff Law <law@redhat.com>
6896 PR tree-optimization/81592
6898 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6899 for range data rather than using global data.
6900 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6901 range data rather than using global data.
6902 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6903 pass it to children as needed.
6904 (struct directive::fmtresult): Similarly.
6905 (struct directive::set_width): Similarly.
6906 (struct directive::set_precision): Similarly.
6907 (format_integer, format_directive, parse_directive): Similarly.
6908 (format_none): Accept unnamed vr_values parameter.
6909 (format_percent, format_floating, format_character): Similarly.
6910 (format_string, format_plain): Similarly.
6911 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6912 the EVRP range analyzer for range data rather than using global data.
6913 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6914 gimple-ssa-evrp-analyze.h
6915 (class sprintf_dom_walker): Add after_dom_children member function.
6916 Add evrp_range_analyzer member.
6917 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6918 range analyzer as needed.
6919 (sprintf_dom_walker::after_dom_children): New member function.
6920 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6922 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6923 (evrp_range_analyzer::pop_to_marker): Likewise.
6925 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6927 PR tree-optimization/84419
6928 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6929 with the required type if its current type is compatible but
6932 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6935 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6936 after vectorization.
6938 2018-02-20 Martin Liska <mliska@suse.cz>
6941 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6942 possible values if we don't have a hint.
6944 2018-02-20 Martin Liska <mliska@suse.cz>
6948 * final.c (shorten_branches): Build align_tab array with one
6950 * opts.c (finish_options): Add alignment option limit check.
6951 (MAX_CODE_ALIGN): Likewise.
6952 (MAX_CODE_ALIGN_VALUE): Likewise.
6953 * doc/invoke.texi: Document maximum allowed option value for
6954 all -falign-* options.
6956 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6959 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6960 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6961 * var-tracking.c (emit_note_insn_var_location): Remove all references
6962 to NOTE_INSN_CALL_ARG_LOCATION.
6963 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6964 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6965 Use copy_rtx_if_shared.
6966 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6967 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6968 (dwarf2out_var_location): Remove handling of
6969 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6971 * final.c (final_scan_insn): Remove all references to
6972 NOTE_INSN_CALL_ARG_LOCATION.
6973 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6974 before dumping final insns.
6975 * except.c (emit_note_eh_region_end): Remove all references to
6976 NOTE_INSN_CALL_ARG_LOCATION.
6977 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6978 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6979 * config/arc/arc.c (hwloop_optimize): Likewise.
6980 * config/arm/arm.c (create_fix_barrier): Likewise.
6981 * config/s390/s390.c (s390_chunkify_start): Likewise.
6982 * config/sh/sh.c (find_barrier): Likewise.
6983 * config/i386/i386.c (rest_of_insert_endbranch,
6984 ix86_seh_fixup_eh_fallthru): Likewise.
6985 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6986 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6987 * config/frv/frv.c (frv_function_prologue): Likewise.
6988 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6990 (note_outside_basic_block_p): Remove all references to
6991 NOTE_INSN_CALL_ARG_LOCATION.
6992 * gengtype.c (adjust_field_rtx_def): Likewise.
6993 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6995 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6996 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6999 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7002 PR tree-optimization/84452
7003 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7004 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7007 2018-02-19 Martin Liska <mliska@suse.cz>
7010 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7012 2018-02-19 Martin Liska <mliska@suse.cz>
7013 Richard Sandiford <richard.sandiford@linaro.org>
7015 PR tree-optimization/82491
7016 * gimple-fold.c (get_base_constructor): Make earlier bail out
7019 2018-02-19 Carl Love <cel@us.ibm.com>
7021 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7022 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7024 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7025 P8V_BUILTIN_VEC_NEG.
7027 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7029 * config/rl78/rl78.md (movdf): New define expand.
7031 2018-02-19 Martin Liska <mliska@suse.cz>
7034 * doc/invoke.texi: Fix typo.
7035 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7037 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7039 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7040 handle rs6000_single_float and rs6000_double_float specially for
7043 2018-02-16 Jeff Law <law@redhat.com>
7045 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7048 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7051 * ipa-inline.c (inline_small_functions): Fix a typo.
7053 2018-02-16 Nathan Sidwell <nathan@acm.org>
7055 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7057 2018-02-16 Carl Love <cel@us.ibm.com>
7059 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7060 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7061 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7062 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7063 expansion to P8V_BUILTIN_VEC_FLOAT2.
7065 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7067 PR rtl-optimization/70023
7068 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7069 src_regno into account.
7071 2018-02-16 Carl Love <cel@us.ibm.com>
7073 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7074 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7075 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7076 * config/rs6000/rs6000.c: Remove case statements for
7077 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7078 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7079 and P9V_BUILTIN_VEC_VINSERT4B.
7080 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7081 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7082 * config/rs6000/vsx.md:
7083 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7086 2018-02-16 Carl Love <cel@us.ibm.com>
7088 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7090 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7092 * config/rs6000/rs6000-c.c: Add the definitions for
7093 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7094 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7095 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7096 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7097 definition for insert4b and define insn *insert3b_internal.
7098 * doc/extend.texi: Add documentation for vec_extract4b.
7100 2018-02-16 Nathan Sidwell <nathan@acm.org>
7102 * doc/extend.texi (Backwards Compatibility): Mention friend
7103 injection. Note for-scope is deprecated.
7104 * doc/invoke.texi (-ffriend-injection): Deprecate.
7106 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7108 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7109 that moved to I2, also allow destinations that are a paradoxical
7110 subreg (instead of a normal reg).
7112 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7115 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7118 2018-02-16 Richard Biener <rguenther@suse.de>
7120 PR tree-optimization/84037
7121 PR tree-optimization/84016
7123 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7124 Adjust vec_construct for the fact we need additional higher latency
7125 128bit inserts for AVX256 and AVX512 vector builds.
7126 (ix86_add_stmt_cost): Scale vector construction cost for
7129 2018-02-16 Richard Biener <rguenther@suse.de>
7131 PR tree-optimization/84417
7132 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7133 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7134 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7136 2018-02-16 Martin Liska <mliska@suse.cz>
7139 * internal-fn.def (ASAN_CHECK): Set proper flags.
7140 (ASAN_MARK): Likewise.
7142 2018-02-16 Julia Koval <julia.koval@intel.com>
7144 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7145 from PTA_CANNONLAKE.
7147 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7150 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7151 Use ++iter rather than iter++ for std::list iterators.
7152 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7153 defer deleting them until all nodes in the forest are processed. Do
7154 free even leaf nodes. Change to_process into auto_vec.
7157 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7158 * vec.h (vec_default_construct): Use memset instead of placement new
7159 if BROKEN_VALUE_INITIALIZATION is defined.
7160 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7161 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7164 PR rtl-optimization/83723
7165 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7166 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7167 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7169 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7171 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7173 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7175 PR rtl-optimization/81443
7176 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7177 from inner REGs to paradoxical SUBREGs.
7179 2018-02-16 Richard Biener <rguenther@suse.de>
7181 PR tree-optimization/84399
7182 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7183 For operands we can analyze at their definition make sure we can
7184 analyze them at each use as well.
7186 2018-02-16 Richard Biener <rguenther@suse.de>
7188 PR tree-optimization/84190
7189 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7190 volatile accesses if the decl isn't volatile.
7192 2018-02-15 Jason Merrill <jason@redhat.com>
7194 PR c++/84314 - ICE with templates and fastcall attribute.
7195 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7196 TYPE_CANONICAL on an existing type.
7198 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7200 PR tree-optimization/84383
7201 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7202 dstoff nor call operand_equal_p if dstbase is NULL.
7204 PR tree-optimization/84334
7205 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7206 also a CONSTANT_CLASS_P, punt.
7208 2018-02-14 Jim Wilson <jimw@sifive.com>
7210 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7211 first SMALL_OPERAND check. New local min_second_step. Move assert
7212 to where locals are set. Add TARGET_RVC support.
7213 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7215 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7217 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7219 2018-02-14 Martin Sebor <msebor@redhat.com>
7221 PR tree-optimization/83698
7222 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7223 arrays constrain the offset range to their bounds.
7224 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7225 (builtin_access::overlap): Avoid setting the size of overlap if it's
7227 (maybe_diag_overlap): Also consider arrays when deciding what values
7228 of offsets to include in diagnostics.
7230 2018-02-14 Martin Sebor <msebor@redhat.com>
7233 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7234 that correspond to the kind of a declaration.
7236 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7239 * config/pa/pa.md: Load address of PIC label using the linkage table
7240 if the label is nonlocal.
7242 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7244 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7245 warning message if user requests -maltivec=be.
7246 * doc/invoke.texi: Document deprecation of -maltivec=be.
7248 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7251 * config/rs6000/rs6000-c.c: Update definitions for
7252 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7253 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7255 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7258 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7259 add _get_ssp intrinsics. Remove argument from
7260 __builtin_ia32_rdssp[d|q].
7261 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7262 * config/i386/i386-builtin.def: Remove argument from
7263 __builtin_ia32_rdssp[d|q].
7264 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7265 ix86_expand_special_args_builtin for _rdssp[d|q].
7266 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7267 Clear register before usage.
7268 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7269 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7271 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7273 PR tree-optimization/84357
7274 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7275 operand 1 of an ARRAY_REF too.
7277 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7280 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7281 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7283 (set_of_reg): New struct.
7284 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7285 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7286 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7288 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7289 Split into bitclr, bitset, bitinvert patterns if appropriate.
7290 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7291 use rx_fuse_in_memory_bitop.
7292 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7293 to named insn, correct maximum insn length.
7295 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7298 * machmode.def: Define a complex mode for PARTIAL_INT.
7299 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7301 * doc/rtl.texi: Document CSPImode.
7302 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7304 (msp430_hard_regno_nregs_with_padding): Likewise.
7306 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7309 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7311 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7313 PR rtl-optimization/84169
7314 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7315 we generated a parallel as new i3 and we split that to new i2 and i3
7316 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7317 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7318 those to i2, not i1. Partially rewrite this scan code.
7320 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7323 * stor-layout.c (place_field): For variable length fields, adjust
7324 offset_align afterwards not just based on the field's alignment,
7325 but also on the size.
7328 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7329 of exps and logs in the use_exp2 case.
7331 2018-02-13 Jeff Law <law@redhat.com>
7333 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7336 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7339 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7343 * common.opt (gas-loc-support, gas-locview-support): New.
7344 (ginline-points, ginternal-reset-location-views): New.
7345 * doc/invoke.texi: Document them. Use @itemx where intended.
7346 (gvariable-location-views): Adjust.
7347 * target.def (reset_location_view): New.
7348 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7349 (TARGET_RESET_LOCATION_VIEW): New.
7350 * doc/tm.texi: Rebuilt.
7351 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7352 (dwarf2out_default_as_locview_support): New.
7353 (output_asm_line_debug_info): Use option variables.
7354 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7355 (output_loc_list): Likewise.
7356 (add_high_low_attributes): Check option variables.
7357 Don't output entry view attribute in strict mode.
7358 (gen_inlined_subroutine_die): Check option variables.
7359 (dwarf2out_inline_entry): Likewise.
7360 (init_sections_and_labels): Likewise.
7361 (dwarf2out_early_finish): Likewise.
7362 (maybe_reset_location_view): New, from...
7363 (dwarf2out_var_location): ... here. Call it.
7364 * debug.h (dwarf2out_default_as_loc_support): Declare.
7365 (dwarf2out_default_as_locview_support): Declare.
7366 * hooks.c (hook_int_rtx_insn_0): New.
7367 * hooks.h (hook_int_rtx_insn_0): Declare.
7368 * toplev.c (process_options): Take -gas-loc-support and
7369 -gas-locview-support from dwarf2out. Enable
7370 -gvariable-location-views by default only with locview
7371 assembler support. Enable -ginternal-reset-location-views by
7372 default only if the target defines the corresponding hook.
7373 Enable -ginline-points by default if location views are
7374 enabled; force it disabled if statement frontiers are
7376 * tree-inline.c (expand_call_inline): Check option variables.
7377 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7379 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7381 PR tree-optimization/84321
7382 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7383 handling. Also check whether the anti-range contains any values
7384 that satisfy the mask; switch to a VR_RANGE if not.
7386 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7389 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7391 2018-02-13 Martin Jambor <mjambor@suse.cz>
7394 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7395 of call statements, also set location of a load to a temporary.
7397 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7399 * config/rl78/rl78.c (add_vector_labels): New function.
7400 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7401 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7402 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7403 which checks that no arguments are passed.
7404 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7405 * doc/extend.texi: Documentation for the new attribute.
7407 2018-02-13 Andreas Schwab <schwab@suse.de>
7409 * config/riscv/linux.h (CPP_SPEC): Define.
7411 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7414 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7415 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7416 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7417 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7418 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7419 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7420 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7421 temporarily for AES and PCLMUL builtins.
7423 PR tree-optimization/84339
7424 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7425 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7429 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7430 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7431 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7433 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7435 * omp-simd-clone.h: New file.
7436 * omp-simd-clone.c: Include omp-simd-clone.h.
7437 (expand_simd_clones): No longer static.
7438 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7439 cgraph.h and omp-simd-clone.h.
7440 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7441 (vect_recog_widen_shift_pattern): Formatting fix.
7442 (vect_pattern_recog_1): Don't check optab for calls.
7445 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7446 operands[2] into a REG before using gen_lowpart on it.
7448 2018-02-12 Jeff Law <law@redhat.com>
7451 * config/sh/sh.c (find_barrier): Consider a sibling call
7454 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7455 successfully back substituting a reg.
7457 2018-02-12 Richard Biener <rguenther@suse.de>
7459 PR tree-optimization/84037
7460 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7461 parameter, move visited init to caller.
7462 (vect_slp_analyze_operations): Separate cost from validity
7463 check, initialize visited once for all instances.
7464 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7466 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7468 (vect_model_promotion_demotion_cost): Likewise.
7469 (vectorizable_bswap): Guard cost modeling with !slp_node
7470 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7472 (vectorizable_call): Likewise.
7473 (vectorizable_conversion): Likewise.
7474 (vectorizable_assignment): Likewise.
7475 (vectorizable_shift): Likewise.
7476 (vectorizable_operation): Likewise.
7477 (vectorizable_store): Likewise.
7478 (vectorizable_load): Likewise.
7479 (vectorizable_condition): Likewise.
7480 (vectorizable_comparison): Likewise.
7482 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7485 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7486 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7487 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7489 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7492 * params.def (inline-min-speedup): Increase from 8 to 15.
7493 (max-inline-insns-auto): Decrease from 40 to 30.
7494 * ipa-split.c (consider_split): Add some buffer for function to
7495 be considered inlining candidate.
7496 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7499 2018-02-12 Richard Biener <rguenther@suse.de>
7501 PR tree-optimization/84037
7502 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7503 matched stmts if we cannot swap the non-matched ones.
7505 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7507 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7508 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7509 _mm_maskz_scalef_round_ss): New intrinsics.
7510 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7511 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7512 __builtin_ia32_scalefss_round): Remove.
7513 (__builtin_ia32_scalefsd_mask_round,
7514 __builtin_ia32_scalefss_mask_round): New intrinsics.
7515 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7516 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7517 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7518 "<round_constraint>")): Changed to ...
7519 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7520 "<round_scalar_constraint>")): ... this.
7521 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7522 %0, %1, %2<round_op3>}"): Changed to ...
7523 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7524 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7525 %2<round_scalar_mask_op3>}"): ... this.
7526 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7528 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7530 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7531 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7532 (_mm_maskz_sqrt_round_ss): New intrinsics.
7533 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7534 (__builtin_ia32_sqrtsd_mask_round)
7535 (__builtin_ia32_sqrtss_mask_round): New builtins.
7536 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7537 (__builtin_ia32_sqrtss_round): Remove.
7538 (__builtin_ia32_sqrtsd_mask_round)
7539 (__builtin_ia32_sqrtss_mask_round): New builtins.
7540 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7541 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7542 ((match_operand:VF_128 1 "vector_operand"
7543 "xBm,<round_constraint>")): Changed to ...
7544 ((match_operand:VF_128 1 "vector_operand"
7545 "xBm,<round_scalar_constraint>")): ... this.
7546 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7547 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7548 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7549 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7550 %<iptr>1<round_scalar_mask_op3>}): ... this.
7551 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7552 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7554 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7557 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7558 Cast vec_cmpeq result to correct type.
7559 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7560 Cast vec_cmpgt result to correct type.
7562 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7564 * final.c (final_scan_insn_1): Renamed from...
7565 (final_scan_insn): ... this. New wrapper, to recover
7566 seen from the outermost call in recursive ones.
7567 * config/sparc/sparc.c (output_return): Drop seen from call.
7568 (output_sibcall): Likewise.
7569 * config/visium/visium.c (output_branch): Likewise.
7571 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7573 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7576 2018-02-10 Alan Modra <amodra@gmail.com>
7579 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7580 Specify LR as an input.
7582 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7585 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7586 remove_member_access_dummy_vars): New functions.
7587 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7588 lower_omp_1, execute_lower_omp): Use them.
7590 PR rtl-optimization/84308
7591 * shrink-wrap.c (spread_components): Release todo vector.
7593 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7595 PR rtl-optimization/57193
7596 * ira-color.c (struct allocno_color_data): Add member
7597 conflict_allocno_hard_prefs.
7598 (update_conflict_allocno_hard_prefs): New.
7599 (bucket_allocno_compare_func): Add a preference based on
7600 conflict_allocno_hard_prefs.
7601 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7602 (color_allocnos): Remove a dead code. Initiate
7603 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7605 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7608 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7609 constraint from =wa to wa. Avoid a subreg on the output operand,
7610 instead use a pseudo and subreg it in a move.
7611 (p9_xxbrd_<mode>): Changed to ...
7612 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7613 (p9_xxbrd_v2df): New expander.
7614 (p9_xxbrw_<mode>): Changed to ...
7615 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7616 (p9_xxbrw_v4sf): New expander.
7618 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7620 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7622 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7625 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7626 multiply in 32-bit mode.
7627 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7628 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7631 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7633 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7634 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7635 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7636 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7638 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7641 * dwarf2out.c (is_trivial_indirect_ref): New function.
7642 (dwarf2out_late_global_decl): Do not generate a location
7643 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7644 and that are not defined in the current unit.
7646 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7648 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7649 instead of a libcall for UNORDERED.
7651 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7654 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7655 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7657 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7660 * config/s390/s390.c (s390_set_current_function): Invoke
7661 s390_indirect_branch_settings also if fndecl didn't change.
7663 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7665 * config/rs6000/rs6000.md (blockage): Set length to zero.
7667 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7669 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7671 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7674 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7675 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7679 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7680 PARALLEL incoming that failed vt_get_decl_and_offset check.
7683 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7684 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7686 (get_variable_section): For decls in named .bss* sections pass true as
7687 second argument to bss_initializer_p.
7689 2018-02-09 Marek Polacek <polacek@redhat.com>
7690 Jakub Jelinek <jakub@redhat.com>
7693 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7694 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7695 Sync some changes from cxx_fold_indirect_ref.
7697 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7699 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7701 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7702 (BLOCK_INLINE_ENTRY_LABEL): New.
7703 (dwarf2out_var_location): Disregard inline entry markers.
7704 (inline_entry_data): New struct.
7705 (inline_entry_data_hasher): New hashtable type.
7706 (inline_entry_data_hasher::hash): New.
7707 (inline_entry_data_hasher::equal): New.
7708 (inline_entry_data_table): New variable.
7709 (add_high_low_attributes): Add DW_AT_entry_pc and
7710 DW_AT_GNU_entry_view attributes if a pending entry is found
7711 in inline_entry_data_table. Add old entry_pc attribute only
7712 if debug nonbinding markers are disabled.
7713 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7714 markers are enabled.
7715 (block_within_block_p, dwarf2out_inline_entry): New.
7716 (dwarf2out_finish): Check that no entries remained in
7717 inline_entry_data_table.
7718 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7719 (final_scan_insn, notice_source_line): Likewise.
7720 (rest_of_clean_state): Skip inline entry markers.
7721 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7723 * gimple.c (gimple_build_debug_inline_entry): New.
7724 * gimple.h (enum gimple_debug_subcode): Add
7725 GIMPLE_DEBUG_INLINE_ENTRY.
7726 (gimple_build_debug_inline_entry): Declare.
7727 (gimple_debug_inline_entry_p): New.
7728 (gimple_debug_nonbind_marker_p): Adjust.
7729 * insn-notes.def (INLINE_ENTRY): New.
7730 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7731 inline entry marker notes.
7732 (print_insn): Likewise.
7733 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7734 (INSN_DEBUG_MARKER_KIND): Likewise.
7735 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7736 * tree-inline.c (expand_call_inline): Build and insert
7737 debug_inline_entry stmt.
7738 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7739 inline entry blocks early, if nonbind markers are enabled.
7740 (dump_scope_block): Dump fragment info.
7741 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7742 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7743 (gimple_build_debug_inline_entry): New.
7744 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7745 Enable/disable inline entry points too.
7746 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7747 (DEBUG_INSN): Describe inline entry markers.
7749 * common.opt (gvariable-location-views): New.
7750 (gvariable-location-views=incompat5): New.
7751 * config.in: Rebuilt.
7752 * configure: Rebuilt.
7753 * configure.ac: Test assembler for view support.
7754 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7755 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7756 * dwarf2out.c (var_loc_view): New typedef.
7757 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7758 (dwarf2out_locviews_in_attribute): New.
7759 (dwarf2out_locviews_in_loclist): New.
7760 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7761 (enum dw_line_info_opcode): Add LI_adv_address.
7762 (struct dw_line_info_table): Add view.
7763 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7764 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7765 (zero_view_p): New variable.
7766 (ZERO_VIEW_P): New macro.
7767 (output_asm_line_debug_info): New.
7768 (struct var_loc_node): Add view.
7769 (add_AT_view_list, AT_loc_list): New.
7770 (add_var_loc_to_decl): Add view param. Test it against last.
7771 (new_loc_list): Add view params. Record them.
7772 (AT_loc_list_ptr): Handle loc and view lists.
7773 (view_list_to_loc_list_val_node): New.
7774 (print_dw_val): Handle dw_val_class_view_list.
7775 (size_of_die): Likewise.
7776 (value_format): Likewise.
7777 (loc_list_has_views): New.
7778 (gen_llsym): Set vl_symbol too.
7779 (maybe_gen_llsym, skip_loc_list_entry): New.
7780 (dwarf2out_maybe_output_loclist_view_pair): New.
7781 (output_loc_list): Output view list or entries too.
7782 (output_view_list_offset): New.
7783 (output_die): Handle dw_val_class_view_list.
7784 (output_dwarf_version): New.
7785 (output_compilation_unit_header): Use it.
7786 (output_skeleton_debug_sections): Likewise.
7787 (output_rnglists, output_line_info): Likewise.
7788 (output_pubnames, output_aranges): Update version comments.
7789 (output_one_line_info_table): Output view numbers in asm comments.
7790 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7791 Call maybe_gen_llsym.
7792 (loc_list_from_tree_1): Adjust.
7793 (add_AT_location_description): Create view list attribute if
7794 needed, check it's absent otherwise.
7795 (convert_cfa_to_fb_loc_list): Adjust.
7796 (maybe_emit_file): Call output_asm_line_debug_info for test.
7797 (dwarf2out_var_location): Reset views as needed. Precompute
7798 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7799 attribute. Set view.
7800 (new_line_info_table): Reset next view.
7801 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7802 (dwarf2out_source_line): Likewise. Output view resets and labels to
7803 the assembler, or select appropriate line info opcodes.
7804 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7805 (optimize_string_length): Catch it. Adjust.
7806 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7807 dw_val_class_view_list, and remove it if no longer needed.
7808 (hash_loc_list): Hash view numbers.
7809 (loc_list_hasher::equal): Compare them.
7810 (optimize_location_lists): Check whether a view list symbol is
7811 needed, and whether the locview attribute is present, and
7812 whether they match. Remove the locview attribute if no longer
7814 (index_location_lists): Call skip_loc_list_entry for test.
7815 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7816 Use output_dwarf_version.
7817 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7818 (struct dw_val_node): Add val_view_list.
7819 * final.c (SEEN_NEXT_VIEW): New.
7820 (set_next_view_needed): New.
7821 (clear_next_view_needed): New.
7822 (maybe_output_next_view): New.
7823 (final_start_function): Rename to...
7824 (final_start_function_1): ... this. Take pointer to FIRST,
7825 add SEEN parameter. Emit param bindings in the initial view.
7826 (final_start_function): Reintroduce SEEN-less interface.
7827 (final): Rename to...
7828 (final_1): ... this. Take SEEN parameter. Output final pending
7829 next view at the end.
7830 (final): Reintroduce seen-less interface.
7831 (final_scan_insn): Output pending next view before switching
7832 sections or ending a block. Mark the next view as needed when
7833 outputting variable locations. Notify debug backend of section
7834 changes, and of location view changes.
7835 (rest_of_handle_final): Adjust.
7836 * toplev.c (process_options): Autodetect value for debug variable
7837 location views option. Warn on incompat5 without -gdwarf-5.
7838 * doc/invoke.texi (gvariable-location-views): New.
7839 (gvariable-location-views=incompat5): New.
7840 (gno-variable-location-views): New.
7842 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7844 PR tree-optimization/84136
7845 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7846 that the result of find_edge is non-NULL.
7848 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7851 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7852 storing integer register in SImode. Fix cost of 256 and 512
7853 byte aligned SSE register store.
7855 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7857 * config/i386/i386.c (ix86_multiplication_cost): Fix
7858 multiplication cost for TARGET_AVX512DQ.
7860 2018-02-08 Marek Polacek <polacek@redhat.com>
7862 PR tree-optimization/84238
7863 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7866 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7868 PR tree-optimization/84265
7869 * tree-vect-stmts.c (vectorizable_store): Don't treat
7870 VMAT_CONTIGUOUS accesses as grouped.
7871 (vectorizable_load): Likewise.
7873 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7875 PR tree-optimization/81635
7876 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7877 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7878 (test_round_for_mask): New functions.
7879 (wide_int_cc_tests): Call test_round_for_mask.
7880 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7881 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7882 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7883 range returned by get_range_info.
7885 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7888 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7889 * symtab.c: Include builtins.h
7890 (symtab_node::output_to_lto_symbol_table_p): Move here
7891 from lto-streamer-out.c:output_symbol_p.
7892 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7893 (output_symbol_p): Move all logic to symtab.c
7894 (produce_symtab): Update.
7896 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7898 * config/s390/s390-opts.h (enum indirect_branch): Define.
7899 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7900 (s390_indirect_branch_via_thunk)
7901 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7902 (enum s390_indirect_branch_type): Define.
7903 * config/s390/s390.c (struct s390_frame_layout, struct
7904 machine_function): Remove.
7905 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7906 (indirect_branch_table_label_no, indirect_branch_table_name):
7908 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7909 (enum s390_indirect_branch_option): Define.
7910 (s390_return_addr_from_memory): New function.
7911 (s390_handle_string_attribute): New function.
7912 (s390_attribute_table): Add new attribute handler.
7913 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7914 (s390_indirect_branch_via_thunk): New function.
7915 (s390_indirect_branch_via_inline_thunk): New function.
7916 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7917 sibling call optimization for non z10 compiles.
7918 (s390_emit_call): Force indirect branch target to be a single
7919 register. Add r1 clobber for non-z10 compiles.
7920 (s390_emit_epilogue): Emit return jump via return_use expander.
7921 (s390_reorg): Handle JUMP_INSNs as execute targets.
7922 (s390_option_override_internal): Perform validity checks for the
7923 new command line options.
7924 (s390_indirect_branch_attrvalue): New function.
7925 (s390_indirect_branch_settings): New function.
7926 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7927 (s390_output_indirect_thunk_function): New function.
7928 (s390_code_end): Implement target hook.
7929 (s390_case_values_threshold): Implement target hook.
7930 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7932 * config/s390/s390.h (struct s390_frame_layout)
7933 (struct machine_function): Move here from s390.c.
7934 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7935 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7936 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7937 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7938 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7939 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7940 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7941 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7942 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7943 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7944 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7945 (mnemonic attribute): Add values which aren't recognized
7947 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7948 pattern for branch conversion. Fix mnemonic attribute.
7949 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7950 indirect branch via thunk if requested.
7951 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7952 ("*indirect_jump"): Disable for branch conversion using out of
7954 ("indirect_jump_via_thunk<mode>_z10")
7955 ("indirect_jump_via_thunk<mode>")
7956 ("indirect_jump_via_inlinethunk<mode>_z10")
7957 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7958 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7959 ("casesi_jump_via_inlinethunk<mode>_z10")
7960 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7961 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7962 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7963 ("*indirect2_jump"): Disable for branch conversion.
7964 ("casesi_jump"): Turn into expander and expand patterns for branch
7966 ("return_use"): New expander.
7967 ("*return"): Emit return via thunk and rename it to ...
7968 ("*return<mode>"): ... this one.
7969 * config/s390/s390.opt: Add new options and and enum for the
7972 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7974 * lra-constraints.c (match_reload): Unconditionally use
7975 gen_lowpart_SUBREG, rather than selecting between that
7976 and equivalent gen_rtx_SUBREG code.
7978 2018-02-08 Richard Biener <rguenther@suse.de>
7980 PR tree-optimization/84233
7981 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7982 changed flag instead of boguously re-using phi_inserted.
7984 2018-02-08 Martin Jambor <mjambor@suse.cz>
7986 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7987 static local variables.
7989 2018-02-08 Richard Biener <rguenther@suse.de>
7991 PR tree-optimization/84278
7992 * tree-vect-stmts.c (vectorizable_store): When looking for
7993 smaller vector types to perform grouped strided loads/stores
7994 make sure the mode is supported by the target.
7995 (vectorizable_load): Likewise.
7997 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7999 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8000 Increase LDP/STP opportunities by adding adjacent callee-saves.
8002 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8004 PR rtl-optimization/84068
8005 PR rtl-optimization/83459
8006 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8008 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8010 PR tree-optimization/84224
8011 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8012 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8015 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8018 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8019 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8020 count, remove one USE.
8022 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8024 * doc/install.texi (Configuration): Document the
8025 --with-long-double-format={ibm,ieee} PowerPC configuration
8029 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8030 Convert from define_expand to be define_insn_and_split. Rework
8031 float/double/_Float128 conversions to QI/HI/SImode to work with
8032 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8033 conversions to QI/HImode types did a store and then a load to
8034 truncate the value. For conversions to VSX registers, don't split
8035 the insn, instead emit the code directly. Use the code iterator
8036 any_fix to combine signed and unsigned conversions.
8037 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8038 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8039 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8040 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8041 (fix_<mode>di2_hw): Likewise.
8042 (fixuns_<mode>di2_hw): Likewise.
8043 (fix_<mode>si2_hw): Likewise.
8044 (fixuns_<mode>si2_hw): Likewise.
8045 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8046 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8047 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8048 fix<uns>_trunc<SFDF:mode>si2_p8.
8049 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8051 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8052 (fix<uns>_<mode>_mem): Likewise.
8053 (fctiw<u>z_<mode>_mem): Likewise.
8054 (fix<uns>_<mode>_mem): Likewise.
8055 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8056 the register allocator from doing a direct move to the GPRs to do
8057 a store, and instead use the ISA 3.0 store byte/half-word from
8058 vector register instruction. For IEEE 128-bit floating point,
8059 also optimize stores of 32-bit ints.
8060 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8062 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8064 * genextract.c (push_pathstr_operand): New function to support
8066 (walk_rtx): Call push_pathstr_operand.
8067 (print_path): Support [a-zA-Z].
8069 2018-02-07 Richard Biener <rguenther@suse.de>
8071 PR tree-optimization/84037
8072 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8073 (cse_and_gimplify_to_preheader): Declare.
8074 (vect_get_place_in_interleaving_chain): Likewise.
8075 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8077 (_loop_vec_info::~_loop_vec_info): Delete it.
8078 (cse_and_gimplify_to_preheader): New function.
8079 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8080 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8081 (vectorizable_load): Likewise. For grouped stores always base
8082 the IV on the first element.
8083 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8084 condition before gimplifying.
8086 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8088 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8089 *DIV_EXPR and *MOD_EXPR.
8091 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8094 * config/i386/i386.c (ix86_option_override_internal): Mask out
8095 the CF_SET bit when checking -fcf-protection.
8097 2018-02-07 Tom de Vries <tom@codesourcery.com>
8100 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8103 2018-02-07 Richard Biener <rguenther@suse.de>
8105 PR tree-optimization/84204
8106 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8109 PR tree-optimization/84205
8110 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8111 special-case isl_ast_op_zdiv_r.
8113 PR tree-optimization/84223
8114 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8115 Only add conditions from within the region.
8116 (gather_bbs::after_dom_children): Adjust.
8118 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8121 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8122 * config/avr/avr.md: Only post-reload split REG-REG moves if
8123 either register is GENERAL_REG_P.
8125 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8127 PR tree-optimization/84235
8128 * tree-ssa-scopedtables.c
8129 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8130 if the subtraction is performed in floating point type where NaNs are
8131 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8132 build 1. Formatting fix.
8134 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8137 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8138 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8139 and skip it regardless of bb boundaries. Use CALL_P macro,
8140 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8141 unnecessarily, formatting fix.
8143 2018-02-06 Michael Collison <michael.collison@arm.com>
8145 * config/arm/thumb2.md:
8146 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8147 (*thumb_mov_notscc): Ditto.
8149 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8152 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8153 unsigned_fix, not "s".
8155 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8157 * configure.ac (gcc_fn_eh_frame_ro): New function.
8158 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8159 correct .eh_frame permissions.
8160 * configure: Regenerate.
8162 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8164 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8167 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8169 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8170 Display warning message for -mno-speculate-indirect-jumps.
8172 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8174 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8176 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8178 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8180 PR tree-optimization/84225
8181 * tree-eh.c (find_trapping_overflow): Only call
8182 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8184 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8187 * config/i386/i386.c: Reimplement the check of possible options
8188 -mibt/-mshstk conbination. Change error messages.
8189 * doc/invoke.texi: Fix a typo: remove extra '='.
8191 2018-02-06 Marek Polacek <polacek@redhat.com>
8193 PR tree-optimization/84228
8194 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8196 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8199 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8200 emitted arch directives.
8201 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8202 __ARM_FEATURE_COPROC before changing architectures.
8204 2018-02-06 Richard Biener <rguenther@suse.de>
8206 * config/i386/i386.c (print_reg): Fix typo.
8207 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8209 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8211 * configure: Regenerate.
8213 2018-02-05 Martin Sebor <msebor@redhat.com>
8215 PR tree-optimization/83369
8216 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8219 2018-02-05 Martin Liska <mliska@suse.cz>
8221 * doc/invoke.texi: Cherry-pick upstream r323995.
8223 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8225 * ira.c (ira_init_register_move_cost): Adjust comment.
8227 2018-02-05 Martin Liska <mliska@suse.cz>
8229 PR gcov-profile/84137
8230 * doc/gcov.texi: Fix typo in documentation.
8232 2018-02-05 Martin Liska <mliska@suse.cz>
8234 PR gcov-profile/83879
8235 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8236 using dlopen functionality.
8238 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8240 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8241 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8242 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8243 _mm_maskz_range_round_ss): New intrinsics.
8244 (__builtin_ia32_rangesd128_round)
8245 (__builtin_ia32_rangess128_round): Remove.
8246 (__builtin_ia32_rangesd128_mask_round,
8247 __builtin_ia32_rangess128_mask_round): New builtins.
8248 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8249 __builtin_ia32_rangess128_round): Remove.
8250 (__builtin_ia32_rangesd128_mask_round,
8251 __builtin_ia32_rangess128_mask_round): New builtins.
8252 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8253 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8254 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8255 "<round_saeonly_constraint>")): Changed to ...
8256 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8257 "<round_saeonly_scalar_constraint>")): ... this.
8258 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8259 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8260 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8261 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8262 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8264 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8266 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8268 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8269 Remove all values except native, 8540 and 8548.
8271 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8273 * config/i386/i386.c (ix86_output_function_return): Pass
8274 INVALID_REGNUM, instead of -1, as invalid register number to
8275 indirect_thunk_name and output_indirect_thunk.
8277 2018-02-02 Julia Koval <julia.koval@intel.com>
8279 * config.gcc: Add -march=icelake.
8280 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8281 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8282 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8283 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8284 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8285 (processor_target_table): Add icelake.
8286 (ix86_option_override_internal): Handle new PTAs.
8287 (get_builtin_code_for_version): Handle icelake.
8288 (M_INTEL_COREI7_ICELAKE): New.
8289 (fold_builtin_cpu): Handle icelake.
8290 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8291 * doc/invoke.texi: Add -march=icelake.
8293 2018-02-02 Julia Koval <julia.koval@intel.com>
8295 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8296 to wide_int_bitmask.
8297 * wide-int-bitmask.h: New.
8299 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8302 * config/i386/i386.md: Replace Pmode with word_mode in
8303 builtin_setjmp_setup and builtin_longjmp to support x32.
8305 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8309 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8311 (rs6000_supported_cpu_names): New static variable.
8312 (linux_cpu_translation_table): Likewise.
8313 (elf_platform) <cpu>: Define new static variable and use it.
8314 Translate kernel AT_PLATFORM name to canonical name if needed.
8315 Error if platform name is unknown.
8317 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8320 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8322 2018-02-01 Jeff Law <law@redhat.com>
8325 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8326 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8327 the scratch if RELEASE_VIA_POP is false.
8328 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8329 If we have to save a temporary register, decrement SIZE appropriately.
8330 Pass new arguments to release_scratch_register_on_entry.
8331 (ix86_adjust_stack_and_probe): Likewise.
8332 (ix86_emit_probe_stack_range): Pass new arguments to
8333 release_scratch_register_on_entry.
8335 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8337 PR rtl-optimization/84157
8338 * combine.c (change_zero_ext): Use REG_P predicate in
8339 front of HARD_REGISTER_P predicate.
8341 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8343 * config/avr/avr.c (avr_option_override): Move disabling of
8344 -fdelete-null-pointer-checks to...
8345 * common/config/avr/avr-common.c (avr_option_optimization_table):
8348 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8350 PR tree-optimization/81635
8351 * tree-data-ref.c (split_constant_offset_1): For types that
8352 wrap on overflow, try to use range info to prove that wrapping
8355 2018-02-01 Renlin Li <renlin.li@arm.com>
8358 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8360 (aarch64_register_move_cost): Likewise.
8361 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8363 (REG_CLASS_NAMES): Likewise.
8364 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8365 TAILCALL_ADDR_REGS. Remove IP registers.
8366 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8368 2018-02-01 Richard Biener <rguenther@suse.de>
8370 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8371 for specifying RPO order and allow NULL for that.
8372 * domwalk.c (dom_walker::dom_walker): Likewise.
8373 (dom_walker::walk): Handle NULL RPO order.
8374 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8376 (rewrite_update_dom_walker): Likewise.
8377 (mark_def_dom_walker): Likewise.
8379 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8381 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8382 (aarch64_maybe_expand_sve_subreg_move): Declare.
8383 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8384 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8386 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8387 that are semantically a reverse operation.
8388 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8389 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8390 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8392 (aarch64_can_change_mode_class): For big-endian, forbid changes
8393 between two SVE modes if they have different element sizes.
8395 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8397 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8398 the TImode handling for big-endian targets.
8400 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8402 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8403 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8405 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8406 Remove BSWAP handing for big-endian targets and use the form of
8407 LD1RQ appropariate for the mode.
8409 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8411 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8412 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8415 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8418 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8419 check for operands that need to go through aarch64_sve_reload_be.
8421 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8423 PR tree-optimization/81661
8424 PR tree-optimization/84117
8425 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8426 * tree-eh.c: Include gimplify.h.
8427 (find_trapping_overflow, replace_trapping_overflow,
8428 rewrite_to_non_trapping_overflow): New functions.
8429 * tree-vect-loop.c: Include tree-eh.h.
8430 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8431 * tree-data-ref.c: Include tree-eh.h.
8432 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8434 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8436 PR rtl-optimization/84123
8437 * combine.c (change_zero_ext): Check if hard register satisfies
8438 can_change_dest_mode before calling gen_lowpart_SUBREG.
8440 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8443 * ira.c (ira_init_register_move_cost): Remove assert.
8445 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8447 PR rtl-optimization/84071
8448 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8449 * doc/tm.texi: Regenerate.
8451 2018-01-31 Richard Biener <rguenther@suse.de>
8453 PR tree-optimization/84132
8454 * tree-data-ref.c (analyze_miv_subscript): Properly
8455 check whether evolution_function_is_affine_multivariate_p
8456 before calling gcd_of_steps_may_divide_p.
8458 2018-01-31 Julia Koval <julia.koval@intel.com>
8461 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8462 * config/i386/i386.md (rdpid_rex64) New.
8463 (rdpid): Make 32bit only.
8465 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8468 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8469 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8471 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8474 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8476 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8478 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8480 PR rtl-optimization/84071
8481 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8482 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8484 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8486 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8487 (arc_attribute_table): Add 'aux' attribute.
8488 (arc_in_small_data_p): Consider aux like variables.
8489 (arc_is_aux_reg_p): New function.
8490 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8491 (arc_get_aux_arg): New function.
8492 (prepare_move_operands): Handle aux-register access.
8493 (arc_handle_aux_attribute): New function.
8494 * doc/extend.texi (ARC Variable attributes): Add subsection.
8496 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8498 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8499 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8500 (arc_attribute_table): Add 'uncached' attribute.
8501 (arc_print_operand): Print '.di' flag for uncached memory
8503 (arc_in_small_data_p): Do not consider for small data the uncached
8505 (arc_is_uncached_mem_p): New function.
8506 * config/arc/predicates.md (compact_store_memory_operand): Check
8507 for uncached memory accesses.
8508 (nonvol_nonimm_operand): Likewise.
8509 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8511 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8514 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8515 falign-loops=): Add Optimization flag.
8517 2018-01-30 Jeff Law <law@redhat.com>
8520 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8521 INT_REGISTERS_SAVED. Check it prior to calling
8522 get_scratch_register_on_entry.
8523 (ix86_adjust_stack_and_probe): Similarly.
8524 (ix86_emit_probe_stack_range): Similarly.
8525 (ix86_expand_prologue): Corresponding changes.
8527 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8530 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8531 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8533 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8536 * lra-constraints.c (curr_insn_transform): Process AND in the
8539 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8541 PR rtl-optimization/83986
8542 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8543 dependence against last_pending_memory_flush in addition to
8546 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8548 PR tree-optimization/81611
8549 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8552 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8555 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8558 2018-01-30 Richard Biener <rguenther@suse.de>
8559 Jakub Jelinek <jakub@redhat.com>
8561 PR tree-optimization/84111
8562 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8563 inner loops added during recursion, as they don't have up-to-date
8566 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8569 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8570 (can_inline_edge_by_limits_p): ... here.
8571 (can_early_inline_edge_p, check_callers,
8572 update_caller_keys, update_callee_keys, recursive_inlining,
8573 add_new_edges_to_heap, speculation_useful_p,
8574 inline_small_functions,
8575 inline_small_functions, flatten_function,
8576 inline_to_all_callers_1): Update.
8578 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8580 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8583 2018-01-30 Richard Biener <rguenther@suse.de>
8585 PR tree-optimization/83008
8586 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8587 invariant and constant vector uses in stmts when they need
8590 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8593 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8594 * configure: Regenerate.
8596 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8598 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8600 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8601 Use gen_rtx_REG rather than gen_lowpart.
8603 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8605 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8606 rather than 0 when creating partial subregs.
8608 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8610 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8613 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8616 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8617 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8618 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8619 flags. This restores the settings used before the 2017-07-24.
8620 Turning off pre increment/decrement/modify allows IVOPTS to
8621 optimize DF/SF loops where the index is an int.
8623 2018-01-29 Richard Biener <rguenther@suse.de>
8624 Kelvin Nilsen <kelvin@gcc.gnu.org>
8627 * tree-vect-stmts.c (vectorizable_call): Don't call
8628 targetm.vectorize_builtin_md_vectorized_function if callee is
8631 2018-01-22 Carl Love <cel@us.ibm.com>
8633 * doc/extend.tex: Fix typo in second arg in
8634 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8636 2018-01-29 Richard Biener <rguenther@suse.de>
8638 PR tree-optimization/84086
8639 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8640 (flush_ssaname_freelist): When SSA names were released reset
8641 the SCEV hash table.
8643 2018-01-29 Richard Biener <rguenther@suse.de>
8645 PR tree-optimization/84057
8646 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8647 removed paths when removing edges.
8649 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8651 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8652 -mfunction-return=@var{choice}.
8654 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8657 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8659 (layout::print_source_line): Likewise.
8660 (test_get_line_width_without_trailing_whitespace): Add test cases.
8662 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8665 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8668 2018-01-26 Jim Wilson <jimw@sifive.com>
8670 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8672 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8675 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8677 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8678 and CMP + SUB-immediate -> SUBS.
8680 2018-01-26 Martin Sebor <msebor@redhat.com>
8682 PR tree-optimization/83896
8683 * tree-ssa-strlen.c (get_string_len): Rename...
8684 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8685 Avoid assuming length is constant.
8686 (handle_char_store): Use HOST_WIDE_INT for string length.
8688 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8691 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8692 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8694 2018-01-26 Richard Biener <rguenther@suse.de>
8696 PR rtl-optimization/84003
8697 * dse.c (record_store): Only record redundant stores when
8698 the earlier store aliases at least all accesses the later one does.
8700 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8702 PR rtl-optimization/83985
8703 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8704 REG_CFA_RESTORE insns.
8705 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8706 REG_CFA_RESTORE insns here.
8709 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8710 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8712 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8714 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8715 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8716 (arc_init): Likewise.
8717 (arc_override_options): Likewise.
8718 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8720 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8722 * config/arc/arc.h (TARGET_DBNZ): Define.
8723 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8724 properly set the tune attribute.
8725 (dbnz): Use TARGET_DBNZ guard.
8726 * config/arc/arc.opt (mtune): Add core3 option.
8728 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8730 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8731 recognize new pic like addresses.
8732 (arc_delegitimize_address): Clean up.
8734 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8736 * config/arc/arc-arches.def: Option mrf16 valid for all
8738 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8739 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8740 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8741 * config/arc/arc-tables.opt: Regenerate.
8742 * config/arc/arc.c (arc_conditional_register_usage): Handle
8743 reduced register file case.
8744 (arc_file_start): Set must have build attributes.
8745 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8747 * config/arc/arc.opt (mrf16): Add new option.
8748 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8749 * config/arc/genmultilib.awk: Handle new mrf16 option.
8750 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8751 * config/arc/t-multilib: Regenerate.
8752 * doc/invoke.texi (ARC Options): Document mrf16 option.
8754 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8756 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8757 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8758 (arc_attribute_table): Add 'secure_call' attribute.
8759 (arc_print_operand): Print secure call operand.
8760 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8762 (arc_is_secure_call_p): New function. * config/arc/arc.md
8763 (call_i): Add support for sjli instruction.
8764 (call_value_i): Likewise.
8765 * config/arc/constraints.md (Csc): New constraint.
8767 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8768 John Eric Martin <John.Martin@emmicro-us.com>
8770 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8771 * config/arc/arc.c (_arc_jli_section): New struct.
8772 (arc_jli_section): New type.
8773 (rc_jli_sections): New static variable.
8774 (arc_handle_jli_attribute): New function.
8775 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8776 (arc_file_end): New function.
8777 (TARGET_ASM_FILE_END): Define.
8778 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8779 (arc_add_jli_section): New function.
8780 (jli_call_scan): Likewise.
8781 (arc_reorg): Call jli_call_scan.
8782 (arc_output_addsi): Remove 'S' from printing asm operand.
8783 (arc_is_jli_call_p): New function.
8784 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8786 (movhi_insn): Likewise.
8787 (movsi_insn): Likewise.
8788 (movsi_set_cc_insn): Likewise.
8789 (loadqi_update): Likewise.
8790 (load_zeroextendqisi_update): Likewise.
8791 (load_signextendqisi_update): Likewise.
8792 (loadhi_update): Likewise.
8793 (load_zeroextendhisi_update): Likewise.
8794 (load_signextendhisi_update): Likewise.
8795 (loadsi_update): Likewise.
8796 (loadsf_update): Likewise.
8797 (movsicc_insn): Likewise.
8798 (bset_insn): Likewise.
8799 (bxor_insn): Likewise.
8800 (bclr_insn): Likewise.
8801 (bmsk_insn): Likewise.
8802 (bicsi3_insn): Likewise.
8803 (cmpsi_cc_c_insn): Likewise.
8804 (movsi_ne): Likewise.
8805 (movsi_cond_exec): Likewise.
8806 (clrsbsi2): Likewise.
8817 (call_i): Remove 'S' asm letter, add jli instruction.
8818 (call_value_i): Likewise.
8819 * config/arc/arc.op (mjli-always): New option.
8820 * config/arc/constraints.md (Cji): New constraint.
8821 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8823 (subsf3_fpx): Likewise.
8824 (mulsf3_fpx): Likewise.
8825 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8827 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8829 * doc/invoke.texi (ARC): Document mjli-always option.
8831 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8833 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8834 avoid addition with 0 and use incw and decw where possible.
8836 2018-01-26 Richard Biener <rguenther@suse.de>
8838 PR tree-optimization/81082
8839 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8840 association if it requires casting to unsigned.
8841 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8842 from fold_plusminus_mult_expr to catch important cases late when
8843 range info is available.
8845 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8847 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8848 * configure.ac (hidden_linkonce): New test.
8849 * configure: Regenerate.
8850 * config.in: Regenerate.
8852 2018-01-26 Julia Koval <julia.koval@intel.com>
8854 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8855 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8856 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8857 _mm_mask_bitshuffle_epi64_mask): Fix type.
8858 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8859 USI_FTYPE_V4DI_V4DI_USI): Remove.
8860 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8861 __builtin_ia32_vpshufbitqmb256_mask,
8862 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8863 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8864 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8866 2018-01-26 Alan Modra <amodra@gmail.com>
8869 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8870 UNSPEC_VBPERMQ. Sort other unspecs.
8872 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8874 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8876 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8879 * predict.c (drop_profile): Do not push/pop cfun; update also
8881 (handle_missing_profiles): Fix logic looking for zero profiles.
8883 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8886 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8887 on functions with #pragma omp declare simd or functions with simd
8889 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8890 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8891 Remove trailing \n from warning_at calls.
8893 2018-01-25 Tom de Vries <tom@codesourcery.com>
8896 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8897 for neutered workers.
8899 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8902 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8903 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8905 2018-01-24 Jeff Law <law@redhat.com>
8908 * i386.c (get_probe_interval): Move to earlier point.
8909 (ix86_compute_frame_layout): If -fstack-clash-protection and
8910 the frame is larger than the probe interval, then use pushes
8911 to save registers rather than reg->mem moves.
8912 (ix86_expand_prologue): Remove conditional for int_registers_saved
8915 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8918 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8919 min/max for never referenced object.
8921 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8924 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8926 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8927 attributes from DECL_ATTRIBUTES (decl) without affecting
8928 DECL_ATTRIBUTES (current_function_decl).
8929 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8930 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8932 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8934 PR tree-optimization/83979
8935 * fold-const.c (fold_comparison): Use constant_boolean_node
8936 instead of boolean_{true,false}_node.
8938 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8940 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8943 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8945 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8946 Simplify the clause that sets the length attribute.
8947 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8948 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8949 clause that sets the length attribute.
8950 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8952 2018-01-24 Tom de Vries <tom@codesourcery.com>
8955 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8956 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8957 Add strict parameter.
8958 (prevent_branch_around_nothing): Insert dummy insn between branch to
8959 label and label with no ptx insn inbetween.
8960 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8962 2018-01-24 Tom de Vries <tom@codesourcery.com>
8965 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8966 for neutered threads in warp.
8967 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8969 2018-01-24 Richard Biener <rguenther@suse.de>
8971 PR tree-optimization/83176
8972 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8975 2018-01-24 Richard Biener <rguenther@suse.de>
8977 PR tree-optimization/82819
8978 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8979 code generating pluses that are no-ops in the target precision.
8981 2018-01-24 Richard Biener <rguenther@suse.de>
8984 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8986 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8988 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8989 to merge probabilities.
8990 * predict.c (probably_never_executed): Also mark as cold functions
8991 with global 0 profile and guessed local profile.
8992 * profile-count.c (profile_probability::combine_with_count): New
8994 * profile-count.h (profile_probability::operator*,
8995 profile_probability::operator*=, profile_probability::operator/,
8996 profile_probability::operator/=): Reduce precision to adjusted
8997 and set value to guessed on contradictory divisions.
8998 (profile_probability::combine_with_freq): Remove.
8999 (profile_probability::combine_wiht_count): Declare.
9000 (profile_count::force_nonzero):: Set to adjusted.
9001 (profile_count::probability_in):: Set quality to adjusted.
9002 * tree-ssa-tail-merge.c (replace_block_by): Use
9005 2018-01-23 Andrew Waterman <andrew@sifive.com>
9006 Jim Wilson <jimw@sifive.com>
9008 * config/riscv/riscv.c (riscv_stack_boundary): New.
9009 (riscv_option_override): Set riscv_stack_boundary. Handle
9010 riscv_preferred_stack_boundary_arg.
9011 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9012 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9013 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9014 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9015 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9016 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9018 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9021 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9022 of struct ix86_frame.
9023 (ix86_expand_epilogue): Likewise. Add a local variable for
9024 the reg_save_offset field in struct ix86_frame.
9026 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9028 PR tree-optimization/82604
9029 * tree-loop-distribution.c (enum partition_kind): New enum item
9030 PKIND_PARTIAL_MEMSET.
9031 (partition_builtin_p): Support above new enum item.
9032 (generate_code_for_partition): Ditto.
9033 (compute_access_range): Differentiate cases that equality can be
9034 proven at all loops, the innermost loops or no loops.
9035 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9036 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9037 (finalize_partitions, distribute_loop): Don't fuse partition of
9038 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9039 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9042 2018-01-23 Martin Liska <mliska@suse.cz>
9044 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9045 order to ignore the predictor.
9046 (PRED_POLYMORPHIC_CALL): Likewise.
9047 (PRED_RECURSIVE_CALL): Likewise.
9049 2018-01-23 Martin Liska <mliska@suse.cz>
9051 * tree-profile.c (tree_profiling): Print function header to
9052 aware reader which function we are working on.
9053 * value-prof.c (gimple_find_values_to_profile): Do not print
9054 not interesting value histograms.
9056 2018-01-23 Martin Liska <mliska@suse.cz>
9058 * profile-count.h (enum profile_quality): Add
9059 profile_uninitialized as the first value. Do not number values
9060 as they are zero based.
9061 (profile_count::verify): Update sanity check.
9062 (profile_probability::verify): Likewise.
9064 2018-01-23 Nathan Sidwell <nathan@acm.org>
9066 * doc/invoke.texi (ffor-scope): Deprecate.
9068 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9070 PR tree-optimization/83510
9071 * domwalk.c (set_all_edges_as_executable): New function.
9072 (dom_walker::dom_walker): Convert bool param
9073 "skip_unreachable_blocks" to enum reachability. Move setup of
9074 edge flags to set_all_edges_as_executable and only do it when
9075 reachability is REACHABLE_BLOCKS.
9076 * domwalk.h (enum dom_walker::reachability): New enum.
9077 (dom_walker::dom_walker): Convert bool param
9078 "skip_unreachable_blocks" to enum reachability.
9079 (set_all_edges_as_executable): New decl.
9080 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9081 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9083 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9084 but converting true to REACHABLE_BLOCKS.
9085 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9087 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9088 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9089 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9091 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9092 if check_all_array_refs will be called.
9094 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9096 * tree.c (selftest::test_location_wrappers): Add more test
9099 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9101 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9102 (selftest::test_bit_in_range): Likewise.
9104 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9107 * doc/sourcebuild.texi (vect_float): Say that the selector
9108 only describes the situation when -funsafe-math-optimizations is on.
9109 (vect_float_strict): Document.
9111 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9113 PR tree-optimization/83965
9114 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9115 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9116 instead of checking only for a reduction.
9117 (vect_recog_widen_sum_pattern): Likewise.
9119 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9121 * predict.c (probably_never_executed): Only use precise profile info.
9122 (compute_function_frequency): Skip after inlining hack since we now
9123 have quality checking.
9125 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9127 * profile-count.h (profile_probability::very_unlikely,
9128 profile_probability::unlikely, profile_probability::even): Set
9129 precision to guessed.
9131 2018-01-23 Richard Biener <rguenther@suse.de>
9133 PR tree-optimization/83963
9134 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9135 Properly terminate dominator walk when crossing the exit edge not
9136 when visiting its source block.
9138 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9141 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9142 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9144 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9146 PR tree-optimization/83957
9147 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9148 semicolon after for body surrounded by braces.
9150 PR tree-optimization/83081
9151 * profile-count.h (profile_probability::split): New method.
9152 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9153 Use profile_probability::split.
9154 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9155 when splitting a single conditional jump into 2.
9157 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9159 PR tree-optimization/69452
9160 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9163 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9165 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9166 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9167 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9169 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9171 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9172 * config/rl78/rl78.md (movdi): New define_expand.
9173 * config/rl78/rl78.c (rl78_split_movdi): New function.
9175 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9178 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9180 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9181 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9182 128-bit to produce an UNSPEC move to get the double word with the
9183 signbit and then a shift directly to do signbit.
9184 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9185 implementation with a new version that just does either a direct
9186 move or a regular move. Move memory interface to separate insns.
9187 Move insns so they are next to the expander.
9188 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9189 with signbit move. Split big and little endian case.
9190 (signbit<mode>2_dm_mem_le): Likewise.
9191 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9192 (signbit<mode>2_dm2): Likewise.
9194 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9196 * config/rl78/rl78.md (anddi3): New define_expand.
9198 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9200 * config/rl78/rl78.md (umindi3): New define_expand.
9202 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9204 * config/rl78/rl78.md (smindi3): New define_expand.
9206 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9208 * config/rl78/rl78.md (smaxdi3): New define_expand.
9210 2018-01-22 Carl Love <cel@us.ibm.com>
9212 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9213 LVX_V1TI): Add macro expansion.
9214 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9215 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9216 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9217 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9218 Change check to determine if the instruction is a byte reversing
9219 entry. Fix typo in comment.
9220 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9221 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9222 Add def_builtin calls for new builtins.
9223 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9224 Add define_insn expansion.
9226 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9228 * config/rl78/rl78.md (umaxdi3): New define_expand.
9230 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9232 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9233 for non-QImode registers.
9235 2018-01-22 Richard Biener <rguenther@suse.de>
9237 PR tree-optimization/83963
9238 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9239 including the loop exit block.
9240 (scop_detection::merge_sese): Likewise.
9241 (scop_detection::add_scop): Do it here instead.
9243 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9245 * doc/sourcebuild.texi (arm_softfloat): Document.
9247 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9250 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9251 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9252 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9254 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9255 David Edelsohn <dje.gcc@gmail.com>
9258 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9259 Change "crset eq" to "crset 2".
9260 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9261 (*call_indirect_aix<mode>_nospec): Likewise.
9262 (*call_value_indirect_aix<mode>_nospec): Likewise.
9263 (*call_indirect_elfv2<mode>_nospec): Likewise.
9264 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9265 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9266 change assembly output from . to $.
9267 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9268 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9269 (*tablejump<mode>_internal1_nospec): Likewise.
9271 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9274 * config/sh/sh_optimize_sett_clrt.cc:
9275 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9277 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9279 PR tree-optimization/83940
9280 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9281 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9282 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9283 use it to pass back the definition type.
9284 (vect_check_store_rhs): Likewise rhs_dt_out.
9285 (vect_build_gather_load_calls): Add a mask_dt argument and use
9286 it instead of a call to vect_is_simple_use.
9287 (vectorizable_store): Update calls to vect_check_load_store_mask
9288 and vect_check_store_rhs. Use the dt returned by the latter instead
9289 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9290 instead of calls to vect_is_simple_use. Pass the scalar rather
9291 than the vector operand to vect_is_simple_use when handling
9292 second and subsequent copies of an rhs value.
9293 (vectorizable_load): Update calls to vect_check_load_store_mask
9294 and vect_build_gather_load_calls. Use the cached mask_dt and
9295 gs_info.offset_dt instead of calls to vect_is_simple_use.
9297 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9300 * tree-emutls.c: Include gimplify.h.
9301 (lower_emutls_2): New function.
9302 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9303 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9304 it before further processing.
9307 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9308 UINTVAL (trueop1) instead of INTVAL (op1).
9310 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9314 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9315 INCOMING_FRAME_SP_OFFSET if not defined.
9316 (scan_trace): Add ENTRY argument. If true and
9317 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9318 emit a note to adjust the CFA offset.
9319 (create_cfi_notes): Adjust scan_trace callers.
9320 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9321 INCOMING_FRAME_SP_OFFSET in the CIE.
9322 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9323 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9325 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9326 * doc/tm.texi: Regenerated.
9328 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9330 PR rtl-optimization/83147
9331 * lra-constraints.c (remove_inheritance_pseudos): Use
9332 lra_substitute_pseudo_within_insn.
9334 2018-01-19 Tom de Vries <tom@codesourcery.com>
9335 Cesar Philippidis <cesar@codesourcery.com>
9338 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9340 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9343 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9344 spaces for function labels.
9346 2018-01-19 Martin Liska <mliska@suse.cz>
9348 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9349 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9350 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9351 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9352 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9353 (PRED_CONST_RETURN): Change from 69 to 65.
9354 (PRED_NULL_RETURN): Change from 91 to 71.
9355 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9356 (PRED_LOOP_GUARD): Change from 66 to 73.
9358 2018-01-19 Martin Liska <mliska@suse.cz>
9360 * predict.c (predict_insn_def): Add new assert.
9361 (struct branch_predictor): Change type to signed integer.
9362 (test_prediction_value_range): Amend test to cover
9364 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9365 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9366 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9367 (PRED_LOOP_IV_COMPARE): Likewise.
9368 * predict.h (PROB_UNINITIALIZED): Define new constant.
9370 2018-01-19 Martin Liska <mliska@suse.cz>
9372 * predict.c (dump_prediction): Add new format for
9373 analyze_brprob.py script which is enabled with -details
9375 * profile-count.h (precise_p): New function.
9377 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9379 PR tree-optimization/83922
9380 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9381 there are no statements that need masking.
9382 (vect_active_double_reduction_p): New function.
9383 (vect_analyze_loop_operations): Use it when handling phis that
9384 are not in the loop header.
9386 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9388 PR tree-optimization/83914
9389 * tree-vect-loop.c (vectorizable_induction): Don't convert
9390 init_expr or apply the peeling adjustment for inductions
9391 that are nested within the vectorized loop.
9393 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9395 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9398 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9402 * function.h (gimplify_parameters): Add gimple_seq * argument.
9403 * function.c: Include gimple.h and options.h.
9404 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9405 for the added local temporaries if needed.
9406 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9407 if there are any parameter cleanups, wrap whole body into a
9408 try/finally with the cleanups.
9410 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9413 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9414 Use GET_MODE_CLASS for scalar floating point.
9416 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9420 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9421 Fix call of call_cgraph_insertion_hooks.
9423 2018-01-18 Martin Sebor <msebor@redhat.com>
9425 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9427 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9430 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9433 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9436 * common.opt: (-ffile-prefix-map): New option.
9437 * opts.c (common_handle_option): Defer it.
9438 * opts-global.c (handle_common_deferred_options): Handle it.
9439 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9440 * file-prefix-map.h: New file.
9441 (remap_debug_filename, add_debug_prefix_map): ...here.
9442 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9443 * final.c (debug_prefix_map, add_debug_prefix_map
9444 remap_debug_filename): Move to...
9445 * file-prefix-map.c: New file.
9446 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9447 generalize, get rid of alloca(), use strrchr() instead of strchr().
9448 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9449 Implement in terms of add_prefix_map().
9450 (remap_macro_filename, remap_debug_filename): Implement in term of
9452 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9453 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9454 * dbxout.c: Include file-prefix-map.h.
9455 * varasm.c: Likewise.
9456 * vmsdbgout.c: Likewise.
9457 * xcoffout.c: Likewise.
9458 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9459 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9460 * doc/invoke.texi (-ffile-prefix-map): Document.
9461 (-fdebug-prefix-map): Update description.
9463 2018-01-18 Martin Liska <mliska@suse.cz>
9465 * config/i386/i386.c (indirect_thunk_name): Document that also
9467 (output_indirect_thunk): Document why both instructions
9468 (pause and lfence) are generated.
9470 2018-01-18 Richard Biener <rguenther@suse.de>
9472 PR tree-optimization/83887
9473 * graphite-scop-detection.c
9474 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9475 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9476 (scop_detection::merge_sese): Re-implement with a flood-fill
9477 algorithm that properly finds a SESE region if it exists.
9479 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9482 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9483 pointer_diff optimizations use view_convert instead of convert.
9485 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9487 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9488 Generate different code for -mno-speculate-indirect-jumps.
9489 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9490 (*call_indirect_aix<mode>): Disable for
9491 -mno-speculate-indirect-jumps.
9492 (*call_indirect_aix<mode>_nospec): New define_insn.
9493 (*call_value_indirect_aix<mode>): Disable for
9494 -mno-speculate-indirect-jumps.
9495 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9496 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9497 -mno-speculate-indirect-jumps.
9498 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9500 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9502 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9503 long double type, set the flags for noting the default long double
9504 type, even if we don't pass or return a long double type.
9506 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9509 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9512 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9514 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9515 support for merge[hl].
9516 (fold_mergehl_helper): New helper function.
9517 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9518 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9519 (altivec_vmrglw_direct): Add xxmrglw insn.
9521 2018-01-17 Andrew Waterman <andrew@sifive.com>
9523 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9524 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9526 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9529 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9530 call the lto_location_cache before reading the
9531 DECL_SOURCE_LOCATION of the types.
9533 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9534 Richard Sandiford <richard.sandiford@linaro.org>
9536 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9537 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9538 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9539 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9540 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9542 * config/aarch64/constraints.md (aarch64_movti_operand):
9544 * config/aarch64/predicates.md (Uti): Add new constraint.
9546 2018-01-17 Carl Love <cel@us.ibm.com>
9548 * config/rs6000/vsx.md (define_expand xl_len_r,
9549 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9550 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9552 (define_expand, define_insn): Move the shift left from the
9553 define_insn to the define_expand for lxvl and stxvl instructions.
9554 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9555 and XL_LEN_R definitions to PURE.
9557 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9559 * config/i386/i386.c (indirect_thunk_name): Declare regno
9560 as unsigned int. Compare regno with INVALID_REGNUM.
9561 (output_indirect_thunk): Ditto.
9562 (output_indirect_thunk_function): Ditto.
9563 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9564 in the call to output_indirect_thunk_function.
9566 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9569 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9570 rather than the size of inner_type to determine the stack slot size
9571 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9573 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9576 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9579 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9581 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9582 endian Linux systems to optionally enable multilibs for selecting
9583 the long double type if the user configured an explicit type.
9584 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9585 have no long double multilibs if not defined.
9586 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9587 warn if the user used -mabi={ieee,ibm}longdouble and we built
9588 multilibs for long double.
9589 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9590 appropriate multilib option.
9591 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9593 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9594 for building long double multilibs.
9595 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9597 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9599 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9602 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9604 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9607 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9610 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9613 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9615 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9616 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9618 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9620 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9621 different rtl trees depending on TARGET_64BIT.
9622 (rs6000_gen_lvx): Likewise.
9624 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9626 * config/visium/visium.md (nop): Tweak comment.
9627 (hazard_nop): Likewise.
9629 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9631 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9632 -mspeculate-indirect-jumps.
9633 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9634 for -mno-speculate-indirect-jumps.
9635 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9636 (*call_value_indirect_elfv2<mode>): Disable for
9637 -mno-speculate-indirect-jumps.
9638 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9639 (indirect_jump): Emit different RTL for
9640 -mno-speculate-indirect-jumps.
9641 (*indirect_jump<mode>): Disable for
9642 -mno-speculate-indirect-jumps.
9643 (*indirect_jump<mode>_nospec): New define_insn.
9644 (tablejump): Emit different RTL for
9645 -mno-speculate-indirect-jumps.
9646 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9647 (tablejumpsi_nospec): New define_expand.
9648 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9649 (tablejumpdi_nospec): New define_expand.
9650 (*tablejump<mode>_internal1): Disable for
9651 -mno-speculate-indirect-jumps.
9652 (*tablejump<mode>_internal1_nospec): New define_insn.
9653 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9656 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9658 * caller-save.c (insert_save): Drop unnecessary parameter. All
9661 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9662 Richard Biener <rguenth@suse.de>
9665 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9666 return early, inline manually is_gimple_sizepos. Make sure if we
9667 call gimplify_expr we don't end up with a gimple constant.
9668 * tree.c (variably_modified_type_p): Don't return true for
9669 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9670 * gimplify.h (is_gimple_sizepos): Remove.
9672 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9674 PR tree-optimization/83857
9675 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9676 vectorizable_live_operation for pure SLP statements.
9677 (vectorizable_live_operation): Handle PHIs.
9679 2018-01-16 Richard Biener <rguenther@suse.de>
9681 PR tree-optimization/83867
9682 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9683 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9685 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9688 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9689 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9690 If off is not INTEGER_CST, issue a may not be aligned warning
9691 rather than isn't aligned. Use isn%'t rather than isn't.
9692 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9694 <case MULT_EXPR>: Improve the case when bottom and one of the
9695 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9696 operand, in that case check if the other operand is multiple of
9697 bottom divided by the INTEGER_CST operand.
9699 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9702 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9703 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9704 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9705 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9706 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9707 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9708 (pa_function_arg_size): New function.
9710 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9712 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9713 in a separate statement.
9715 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9717 PR tree-optimization/83847
9718 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9719 group gathers and scatters.
9721 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9723 PR rtl-optimization/86620
9724 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9726 PR rtl-optimization/83213
9727 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9728 to last if both are JUMP_INSNs.
9730 PR tree-optimization/83843
9731 * gimple-ssa-store-merging.c
9732 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9733 store_immediate_info for bswap/nop orig_stores.
9735 2018-01-15 Andrew Waterman <andrew@sifive.com>
9737 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9739 <UDIV>: Increase cost if !TARGET_DIV.
9741 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9743 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9744 (define_attr "cr_logical_3op"): New.
9745 (cceq_ior_compare): Adjust.
9746 (cceq_ior_compare_complement): Adjust.
9747 (*cceq_rev_compare): Adjust.
9748 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9749 (is_cracked_insn): Adjust.
9750 (insn_must_be_first_in_group): Adjust.
9751 * config/rs6000/40x.md: Adjust.
9752 * config/rs6000/440.md: Adjust.
9753 * config/rs6000/476.md: Adjust.
9754 * config/rs6000/601.md: Adjust.
9755 * config/rs6000/603.md: Adjust.
9756 * config/rs6000/6xx.md: Adjust.
9757 * config/rs6000/7450.md: Adjust.
9758 * config/rs6000/7xx.md: Adjust.
9759 * config/rs6000/8540.md: Adjust.
9760 * config/rs6000/cell.md: Adjust.
9761 * config/rs6000/e300c2c3.md: Adjust.
9762 * config/rs6000/e500mc.md: Adjust.
9763 * config/rs6000/e500mc64.md: Adjust.
9764 * config/rs6000/e5500.md: Adjust.
9765 * config/rs6000/e6500.md: Adjust.
9766 * config/rs6000/mpc.md: Adjust.
9767 * config/rs6000/power4.md: Adjust.
9768 * config/rs6000/power5.md: Adjust.
9769 * config/rs6000/power6.md: Adjust.
9770 * config/rs6000/power7.md: Adjust.
9771 * config/rs6000/power8.md: Adjust.
9772 * config/rs6000/power9.md: Adjust.
9773 * config/rs6000/rs64.md: Adjust.
9774 * config/rs6000/titan.md: Adjust.
9776 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9778 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9779 ix86_indirect_branch_register logic.
9781 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9783 * config/i386/constraints.md (Bs): Update
9784 ix86_indirect_branch_register check. Don't check
9785 ix86_indirect_branch_register with GOT_memory_operand.
9787 * config/i386/predicates.md (GOT_memory_operand): Don't check
9788 ix86_indirect_branch_register here.
9789 (GOT32_symbol_operand): Likewise.
9791 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9793 * config/i386/predicates.md (constant_call_address_operand):
9794 Rewrite ix86_indirect_branch_register logic.
9795 (sibcall_insn_operand): Likewise.
9797 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9799 * config/i386/constraints.md (Bs): Replace
9800 ix86_indirect_branch_thunk_register with
9801 ix86_indirect_branch_register.
9803 * config/i386/i386.md (indirect_jump): Likewise.
9804 (tablejump): Likewise.
9805 (*sibcall_memory): Likewise.
9806 (*sibcall_value_memory): Likewise.
9807 Peepholes of indirect call and jump via memory: Likewise.
9808 * config/i386/i386.opt: Likewise.
9809 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9810 (GOT_memory_operand): Likewise.
9811 (call_insn_operand): Likewise.
9812 (sibcall_insn_operand): Likewise.
9813 (GOT32_symbol_operand): Likewise.
9815 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9818 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9819 type rather than type addr's type points to.
9820 (expand_omp_atomic_mutex): Likewise.
9821 (expand_omp_atomic): Likewise.
9823 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9826 * config/i386/i386.c (output_indirect_thunk_function): Use
9827 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9828 for __x86_return_thunk.
9830 2018-01-15 Richard Biener <rguenther@suse.de>
9833 * expmed.c (extract_bit_field_1): Fix typo.
9835 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9838 * config/arm/iterators.md (VF): New mode iterator.
9839 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9840 Remove integer-related logic from pattern.
9841 (neon_vabd<mode>_3): Likewise.
9843 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9846 * common.opt (fstrict-overflow): No longer an alias.
9847 (fwrapv-pointer): New option.
9848 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9849 also for pointer types based on flag_wrapv_pointer.
9850 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9851 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9852 opts->x_flag_wrapv got set.
9853 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9854 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9855 POINTER_TYPE_OVERFLOW_UNDEFINED.
9856 * match.pd: Likewise in address comparison pattern.
9857 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9859 2018-01-15 Richard Biener <rguenther@suse.de>
9862 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9863 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9864 Reset type names to their identifier if their TYPE_DECL doesn't
9865 have linkage (and thus is used for ODR and devirt).
9866 (save_debug_info_for_decl): Remove.
9867 (save_debug_info_for_type): Likewise.
9868 (add_tree_to_fld_list): Adjust.
9869 * tree-pretty-print.c (dump_generic_node): Make dumping of
9870 type names more robust.
9872 2018-01-15 Richard Biener <rguenther@suse.de>
9874 * BASE-VER: Bump to 8.0.1.
9876 2018-01-14 Martin Sebor <msebor@redhat.com>
9879 * builtins.c (check_access): Avoid warning when the no-warning bit
9882 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9884 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9885 * ira-color (allocno_hard_regs_compare): Likewise.
9887 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9890 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9891 Use .pushsection/.popsection.
9893 2018-01-14 Martin Sebor <msebor@redhat.com>
9896 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9898 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9900 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9901 entry from extra_headers.
9902 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9903 extra_headers, make the list bitwise identical to the i?86-*-* one.
9905 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9907 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9908 -mcmodel=large with -mindirect-branch=thunk,
9909 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9910 -mfunction-return=thunk-extern.
9911 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9912 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9913 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9915 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9917 * config/i386/i386.c (print_reg): Print the name of the full
9918 integer register without '%'.
9919 (ix86_print_operand): Handle 'V'.
9920 * doc/extend.texi: Document 'V' modifier.
9922 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9924 * config/i386/constraints.md (Bs): Disallow memory operand for
9925 -mindirect-branch-register.
9927 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9928 (GOT_memory_operand): Likewise.
9929 (call_insn_operand): Likewise.
9930 (sibcall_insn_operand): Likewise.
9931 (GOT32_symbol_operand): Likewise.
9932 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9933 for -mindirect-branch-register.
9934 (tablejump): Likewise.
9935 (*sibcall_memory): Likewise.
9936 (*sibcall_value_memory): Likewise.
9937 Disallow peepholes of indirect call and jump via memory for
9938 -mindirect-branch-register.
9939 (*call_pop): Replace m with Bw.
9940 (*call_value_pop): Likewise.
9941 (*sibcall_pop_memory): Replace m with Bs.
9942 * config/i386/i386.opt (mindirect-branch-register): New option.
9943 * doc/invoke.texi: Document -mindirect-branch-register option.
9945 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9947 * config/i386/i386-protos.h (ix86_output_function_return): New.
9948 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9949 set function_return_type.
9950 (indirect_thunk_name): Add ret_p to indicate thunk for function
9952 (output_indirect_thunk_function): Pass false to
9953 indirect_thunk_name.
9954 (ix86_output_indirect_branch_via_reg): Likewise.
9955 (ix86_output_indirect_branch_via_push): Likewise.
9956 (output_indirect_thunk_function): Create alias for function
9957 return thunk if regno < 0.
9958 (ix86_output_function_return): New function.
9959 (ix86_handle_fndecl_attribute): Handle function_return.
9960 (ix86_attribute_table): Add function_return.
9961 * config/i386/i386.h (machine_function): Add
9962 function_return_type.
9963 * config/i386/i386.md (simple_return_internal): Use
9964 ix86_output_function_return.
9965 (simple_return_internal_long): Likewise.
9966 * config/i386/i386.opt (mfunction-return=): New option.
9967 (indirect_branch): Mention -mfunction-return=.
9968 * doc/extend.texi: Document function_return function attribute.
9969 * doc/invoke.texi: Document -mfunction-return= option.
9971 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9973 * config/i386/i386-opts.h (indirect_branch): New.
9974 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9975 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9976 with local indirect jump when converting indirect call and jump.
9977 (ix86_set_indirect_branch_type): New.
9978 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9979 (indirectlabelno): New.
9980 (indirect_thunk_needed): Likewise.
9981 (indirect_thunk_bnd_needed): Likewise.
9982 (indirect_thunks_used): Likewise.
9983 (indirect_thunks_bnd_used): Likewise.
9984 (INDIRECT_LABEL): Likewise.
9985 (indirect_thunk_name): Likewise.
9986 (output_indirect_thunk): Likewise.
9987 (output_indirect_thunk_function): Likewise.
9988 (ix86_output_indirect_branch_via_reg): Likewise.
9989 (ix86_output_indirect_branch_via_push): Likewise.
9990 (ix86_output_indirect_branch): Likewise.
9991 (ix86_output_indirect_jmp): Likewise.
9992 (ix86_code_end): Call output_indirect_thunk_function if needed.
9993 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9995 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9996 (ix86_attribute_table): Add indirect_branch.
9997 * config/i386/i386.h (machine_function): Add indirect_branch_type
9998 and has_local_indirect_jump.
9999 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10001 (tablejump): Likewise.
10002 (*indirect_jump): Use ix86_output_indirect_jmp.
10003 (*tablejump_1): Likewise.
10004 (simple_return_indirect_internal): Likewise.
10005 * config/i386/i386.opt (mindirect-branch=): New option.
10006 (indirect_branch): New.
10009 (thunk-inline): Likewise.
10010 (thunk-extern): Likewise.
10011 * doc/extend.texi: Document indirect_branch function attribute.
10012 * doc/invoke.texi: Document -mindirect-branch= option.
10014 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10017 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10019 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10021 * ipa-inline.c (want_inline_small_function_p): Return false if
10022 inlining has already failed with CIF_FINAL_ERROR.
10023 (update_caller_keys): Call want_inline_small_function_p before
10025 (update_callee_keys): Likewise.
10027 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10029 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10031 (rs6000_quadword_masked_address_p): Likewise.
10032 (quad_aligned_load_p): Likewise.
10033 (quad_aligned_store_p): Likewise.
10034 (const_load_sequence_p): Add comment to describe the outer-most loop.
10035 (mimic_memory_attributes_and_flags): New function.
10036 (rs6000_gen_stvx): Likewise.
10037 (replace_swapped_aligned_store): Likewise.
10038 (rs6000_gen_lvx): Likewise.
10039 (replace_swapped_aligned_load): Likewise.
10040 (replace_swapped_load_constant): Capitalize argument name in
10041 comment describing this function.
10042 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10043 and stores that access quad-word aligned addresses and replace
10044 with stvx or lvx instructions when appropriate.
10045 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10046 New function prototype.
10047 (rs6000_quadword_masked_address_p): Likewise.
10048 (rs6000_gen_lvx): Likewise.
10049 (rs6000_gen_stvx): Likewise.
10050 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10051 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10052 when memory address is aligned.
10053 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10054 this split to select lvx instruction when memory address is aligned.
10055 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10056 instruction when memory address is aligned.
10057 (*vsx_le_perm_load_v16qi): Likewise.
10058 (four unnamed splitters): Modify to select the stvx instruction
10059 when memory is aligned.
10061 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10063 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10064 which appears in the queue multiple times.
10066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10067 Alan Hayward <alan.hayward@arm.com>
10068 David Sherwood <david.sherwood@arm.com>
10070 * tree-vectorizer.h (vec_lower_bound): New structure.
10071 (_loop_vec_info): Add check_nonzero and lower_bounds.
10072 (LOOP_VINFO_CHECK_NONZERO): New macro.
10073 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10074 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10075 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10076 fields. Make seg_len the distance travelled, not including the
10078 (dr_direction_indicator): Declare.
10079 (dr_zero_step_indicator): Likewise.
10080 (dr_known_forward_stride_p): Likewise.
10081 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10083 (runtime_alias_check_p): Allow runtime alias checks with
10085 (operator ==): Compare access_size and align.
10086 (prune_runtime_alias_test_list): Rework for new distinction between
10087 the access_size and seg_len.
10088 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10090 (get_segment_min_max): New function.
10091 (create_intersect_range_checks): Use it.
10092 (dr_step_indicator): New function.
10093 (dr_direction_indicator): Likewise.
10094 (dr_zero_step_indicator): Likewise.
10095 (dr_known_forward_stride_p): Likewise.
10096 * tree-loop-distribution.c (data_ref_segment_size): Return
10097 DR_STEP * (niters - 1).
10098 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10100 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10101 (vect_preserves_scalar_order_p): New function, split out from...
10102 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10103 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10104 (vect_vfa_access_size): New function.
10105 (vect_vfa_align): Likewise.
10106 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10107 (dump_lower_bound): New function.
10108 (vect_check_lower_bound): Likewise.
10109 (vect_small_gap_p): Likewise.
10110 (vectorizable_with_step_bound_p): Likewise.
10111 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10112 depencies if the vectorization factor is 1. Convert the checks
10113 for nonzero steps into checks on the bounds of DR_STEP. Try using
10114 a bunds check for variable steps if the minimum required step is
10115 relatively small. Update calls to the dr_with_seg_len
10116 constructor and to vect_compile_time_alias.
10117 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10119 (vect_loop_versioning): Call it.
10120 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10122 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10124 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10125 Alan Hayward <alan.hayward@arm.com>
10126 David Sherwood <david.sherwood@arm.com>
10128 * doc/sourcebuild.texi (vect_scatter_store): Document.
10129 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10131 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10133 * genopinit.c (main): Add supports_vec_scatter_store and
10134 supports_vec_scatter_store_cached to target_optabs.
10135 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10136 IFN_MASK_SCATTER_STORE.
10137 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10139 * internal-fn.h (internal_store_fn_p): Declare.
10140 (internal_fn_stored_value_index): Likewise.
10141 * internal-fn.c (scatter_store_direct): New macro.
10142 (expand_scatter_store_optab_fn): New function.
10143 (direct_scatter_store_optab_supported_p): New macro.
10144 (internal_store_fn_p): New function.
10145 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10146 IFN_MASK_SCATTER_STORE.
10147 (internal_fn_mask_index): Likewise.
10148 (internal_fn_stored_value_index): New function.
10149 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10150 for scatter stores.
10151 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10152 * optabs-query.c (supports_vec_scatter_store_p): New function.
10153 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10154 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10155 true for scatter stores.
10156 (vect_gather_scatter_fn_p): Handle scatter stores too.
10157 (vect_check_gather_scatter): Consider using scatter stores if
10158 supports_vec_scatter_store_p.
10159 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10160 scatter stores too.
10161 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10162 internal_fn_stored_value_index.
10163 (check_load_store_masking): Handle scatter stores too.
10164 (vect_get_store_rhs): Make public.
10165 (vectorizable_call): Use internal_store_fn_p.
10166 (vectorizable_store): Handle scatter store internal functions.
10167 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10168 when deciding whether the end of the group has been reached.
10169 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10170 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10171 (mask_scatter_store<mode>): New insns.
10173 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10174 Alan Hayward <alan.hayward@arm.com>
10175 David Sherwood <david.sherwood@arm.com>
10177 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10178 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10179 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10181 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10182 Use vect_truncate_gather_scatter_offset if we can't treat the
10183 operation as a normal gather load or scatter store.
10184 (get_group_load_store_type): Take the gather_scatter_info
10185 as argument. Try using a gather load or scatter store for
10186 single-element groups.
10187 (get_load_store_type): Update calls to get_group_load_store_type
10188 and vect_use_strided_gather_scatters_p.
10190 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10191 Alan Hayward <alan.hayward@arm.com>
10192 David Sherwood <david.sherwood@arm.com>
10194 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10195 optional tree argument.
10196 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10198 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10199 but continue to use the current value as a fallback.
10200 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10201 to compare the updates.
10202 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10203 (get_load_store_type): Use it when handling a strided access.
10204 (vect_get_strided_load_store_ops): New function.
10205 (vect_get_data_ptr_increment): Likewise.
10206 (vectorizable_load): Handle strided gather loads. Always pass
10207 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10209 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10210 Alan Hayward <alan.hayward@arm.com>
10211 David Sherwood <david.sherwood@arm.com>
10213 * doc/md.texi (gather_load@var{m}): Document.
10214 (mask_gather_load@var{m}): Likewise.
10215 * genopinit.c (main): Add supports_vec_gather_load and
10216 supports_vec_gather_load_cached to target_optabs.
10217 * optabs-tree.c (init_tree_optimization_optabs): Use
10218 ggc_cleared_alloc to allocate target_optabs.
10219 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10220 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10222 * internal-fn.h (internal_load_fn_p): Declare.
10223 (internal_gather_scatter_fn_p): Likewise.
10224 (internal_fn_mask_index): Likewise.
10225 (internal_gather_scatter_fn_supported_p): Likewise.
10226 * internal-fn.c (gather_load_direct): New macro.
10227 (expand_gather_load_optab_fn): New function.
10228 (direct_gather_load_optab_supported_p): New macro.
10229 (direct_internal_fn_optab): New function.
10230 (internal_load_fn_p): Likewise.
10231 (internal_gather_scatter_fn_p): Likewise.
10232 (internal_fn_mask_index): Likewise.
10233 (internal_gather_scatter_fn_supported_p): Likewise.
10234 * optabs-query.c (supports_at_least_one_mode_p): New function.
10235 (supports_vec_gather_load_p): Likewise.
10236 * optabs-query.h (supports_vec_gather_load_p): Declare.
10237 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10238 and memory_type field.
10239 (NUM_PATTERNS): Bump to 15.
10240 * tree-vect-data-refs.c: Include internal-fn.h.
10241 (vect_gather_scatter_fn_p): New function.
10242 (vect_describe_gather_scatter_call): Likewise.
10243 (vect_check_gather_scatter): Try using internal functions for
10244 gather loads. Recognize existing calls to a gather load function.
10245 (vect_analyze_data_refs): Consider using gather loads if
10246 supports_vec_gather_load_p.
10247 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10248 (vect_get_gather_scatter_offset_type): Likewise.
10249 (vect_convert_mask_for_vectype): Likewise.
10250 (vect_add_conversion_to_patterm): Likewise.
10251 (vect_try_gather_scatter_pattern): Likewise.
10252 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10253 (vect_vect_recog_func_ptrs): Add it.
10254 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10255 internal_fn_mask_index and internal_gather_scatter_fn_p.
10256 (check_load_store_masking): Take the gather_scatter_info as an
10257 argument and handle gather loads.
10258 (vect_get_gather_scatter_ops): New function.
10259 (vectorizable_call): Check internal_load_fn_p.
10260 (vectorizable_load): Likewise. Handle gather load internal
10262 (vectorizable_store): Update call to check_load_store_masking.
10263 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10264 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10265 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10266 (aarch64_gather_scale_operand_d): New predicates.
10267 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10268 (mask_gather_load<mode>): New insns.
10270 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10271 Alan Hayward <alan.hayward@arm.com>
10272 David Sherwood <david.sherwood@arm.com>
10274 * optabs.def (fold_left_plus_optab): New optab.
10275 * doc/md.texi (fold_left_plus_@var{m}): Document.
10276 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10277 * internal-fn.c (fold_left_direct): Define.
10278 (expand_fold_left_optab_fn): Likewise.
10279 (direct_fold_left_optab_supported_p): Likewise.
10280 * fold-const-call.c (fold_const_fold_left): New function.
10281 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10282 * tree-parloops.c (valid_reduction_p): New function.
10283 (gather_scalar_reductions): Use it.
10284 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10285 (vect_finish_replace_stmt): Declare.
10286 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10287 (needs_fold_left_reduction_p): New function, split out from...
10288 (vect_is_simple_reduction): ...here. Accept reductions that
10289 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10290 (vect_force_simple_reduction): Also store the reduction type in
10291 the assignment's STMT_VINFO_REDUC_TYPE.
10292 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10293 (merge_with_identity): New function.
10294 (vect_expand_fold_left): Likewise.
10295 (vectorize_fold_left_reduction): Likewise.
10296 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10297 scalar phi in place for it. Check for target support and reject
10298 cases that would reassociate the operation. Defer the transform
10299 phase to vectorize_fold_left_reduction.
10300 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10301 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10302 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10304 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10306 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10307 call to ifc_temp_var.
10309 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10310 Alan Hayward <alan.hayward@arm.com>
10311 David Sherwood <david.sherwood@arm.com>
10313 * target.def (legitimize_address_displacement): Take the original
10314 offset as a poly_int.
10315 * targhooks.h (default_legitimize_address_displacement): Update
10317 * targhooks.c (default_legitimize_address_displacement): Likewise.
10318 * doc/tm.texi: Regenerate.
10319 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10320 as an argument, moving assert of ad->disp == ad->disp_term to...
10321 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10322 Try calling targetm.legitimize_address_displacement before expanding
10323 the address rather than afterwards, and adjust for the new interface.
10324 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10325 Match the new hook interface. Handle SVE addresses.
10326 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10327 new hook interface.
10329 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10331 * Makefile.in (OBJS): Add early-remat.o.
10332 * target.def (select_early_remat_modes): New hook.
10333 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10334 * doc/tm.texi: Regenerate.
10335 * targhooks.h (default_select_early_remat_modes): Declare.
10336 * targhooks.c (default_select_early_remat_modes): New function.
10337 * timevar.def (TV_EARLY_REMAT): New timevar.
10338 * passes.def (pass_early_remat): New pass.
10339 * tree-pass.h (make_pass_early_remat): Declare.
10340 * early-remat.c: New file.
10341 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10343 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10345 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10346 Alan Hayward <alan.hayward@arm.com>
10347 David Sherwood <david.sherwood@arm.com>
10349 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10350 vfm1 with a bound_epilog parameter.
10351 (vect_do_peeling): Update calls accordingly, and move the prologue
10352 call earlier in the function. Treat the base bound_epilog as 0 for
10353 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10354 this base when peeling for gaps.
10355 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10356 with fully-masked loops.
10357 (vect_estimate_min_profitable_iters): Handle the single peeled
10358 iteration in that case.
10360 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10361 Alan Hayward <alan.hayward@arm.com>
10362 David Sherwood <david.sherwood@arm.com>
10364 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10365 single-element interleaving even if the size is not a power of 2.
10366 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10367 accesses for single-element interleaving if the group size is
10370 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10371 Alan Hayward <alan.hayward@arm.com>
10372 David Sherwood <david.sherwood@arm.com>
10374 * doc/md.texi (fold_extract_last_@var{m}): Document.
10375 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10376 * optabs.def (fold_extract_last_optab): New optab.
10377 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10378 * internal-fn.c (fold_extract_direct): New macro.
10379 (expand_fold_extract_optab_fn): Likewise.
10380 (direct_fold_extract_optab_supported_p): Likewise.
10381 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10382 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10383 EXTRACT_LAST_REDUCTION.
10384 (get_initial_def_for_reduction): Do not create an initial vector
10385 for EXTRACT_LAST_REDUCTION reductions.
10386 (vectorizable_reduction): Leave the scalar phi in place for
10387 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10388 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10389 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10390 transform phase to vectorizable_condition.
10391 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10393 (vect_finish_stmt_generation): ...here.
10394 (vect_finish_replace_stmt): New function.
10395 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10396 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10398 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10401 Alan Hayward <alan.hayward@arm.com>
10402 David Sherwood <david.sherwood@arm.com>
10404 * doc/md.texi (extract_last_@var{m}): Document.
10405 * optabs.def (extract_last_optab): New optab.
10406 * internal-fn.def (EXTRACT_LAST): New internal function.
10407 * internal-fn.c (cond_unary_direct): New macro.
10408 (expand_cond_unary_optab_fn): Likewise.
10409 (direct_cond_unary_optab_supported_p): Likewise.
10410 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10411 loops using EXTRACT_LAST.
10412 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10413 (extract_last_<mode>): ...this optab.
10414 (vec_extract<mode><Vel>): Update accordingly.
10416 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10417 Alan Hayward <alan.hayward@arm.com>
10418 David Sherwood <david.sherwood@arm.com>
10420 * target.def (empty_mask_is_expensive): New hook.
10421 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10422 * doc/tm.texi: Regenerate.
10423 * targhooks.h (default_empty_mask_is_expensive): Declare.
10424 * targhooks.c (default_empty_mask_is_expensive): New function.
10425 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10426 if the target says that empty masks are expensive.
10427 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10429 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10431 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10432 Alan Hayward <alan.hayward@arm.com>
10433 David Sherwood <david.sherwood@arm.com>
10435 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10436 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10437 (vect_use_loop_mask_for_alignment_p): New function.
10438 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10439 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10440 niters_skip argument. Make sure that the first niters_skip elements
10441 of the first iteration are inactive.
10442 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10443 Update call to vect_set_loop_masks_directly.
10444 (get_misalign_in_elems): New function, split out from...
10445 (vect_gen_prolog_loop_niters): ...here.
10446 (vect_update_init_of_dr): Take a code argument that specifies whether
10447 the adjustment should be added or subtracted.
10448 (vect_update_init_of_drs): Likewise.
10449 (vect_prepare_for_masked_peels): New function.
10450 (vect_do_peeling): Skip prologue peeling if we're using a mask
10451 instead. Update call to vect_update_inits_of_drs.
10452 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10454 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10455 alignment. Do not include the number of peeled iterations in
10456 the minimum threshold in that case.
10457 (vectorizable_induction): Adjust the start value down by
10458 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10459 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10460 Take the number of skipped iterations into account when calculating
10462 * tree-vect-stmts.c (vect_gen_while_not): New function.
10464 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10465 Alan Hayward <alan.hayward@arm.com>
10466 David Sherwood <david.sherwood@arm.com>
10468 * doc/sourcebuild.texi (vect_fully_masked): Document.
10469 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10470 default value to 0.
10471 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10473 (vect_analyze_loop_2): ...here. Don't check the vectorization
10474 factor against the number of loop iterations if the loop is
10477 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10478 Alan Hayward <alan.hayward@arm.com>
10479 David Sherwood <david.sherwood@arm.com>
10481 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10482 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10483 (dump_groups): Update accordingly.
10484 (iv_use::mem_type): New member variable.
10485 (address_p): New function.
10486 (record_use): Add a mem_type argument and initialize the new
10488 (record_group_use): Add a mem_type argument. Use address_p.
10489 Remove obsolete null checks of base_object. Update call to record_use.
10490 (find_interesting_uses_op): Update call to record_group_use.
10491 (find_interesting_uses_cond): Likewise.
10492 (find_interesting_uses_address): Likewise.
10493 (get_mem_type_for_internal_fn): New function.
10494 (find_address_like_use): Likewise.
10495 (find_interesting_uses_stmt): Try find_address_like_use before
10496 calling find_interesting_uses_op.
10497 (addr_offset_valid_p): Use the iv mem_type field as the type
10498 of the addressed memory.
10499 (add_autoinc_candidates): Likewise.
10500 (get_address_cost): Likewise.
10501 (split_small_address_groups_p): Use address_p.
10502 (split_address_groups): Likewise.
10503 (add_iv_candidate_for_use): Likewise.
10504 (autoinc_possible_for_pair): Likewise.
10505 (rewrite_groups): Likewise.
10506 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10507 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10508 (get_alias_ptr_type_for_ptr_address): New function.
10509 (rewrite_use_address): Rewrite address uses in calls that were
10510 identified by find_address_like_use.
10512 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10513 Alan Hayward <alan.hayward@arm.com>
10514 David Sherwood <david.sherwood@arm.com>
10516 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10518 * gimple-expr.h (is_gimple_addressable: Likewise.
10519 * gimple-expr.c (is_gimple_address): Likewise.
10520 * internal-fn.c (expand_call_mem_ref): New function.
10521 (expand_mask_load_optab_fn): Use it.
10522 (expand_mask_store_optab_fn): Likewise.
10524 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10525 Alan Hayward <alan.hayward@arm.com>
10526 David Sherwood <david.sherwood@arm.com>
10528 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10529 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10530 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10531 (cond_umax@var{mode}): Document.
10532 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10533 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10534 (cond_umin_optab, cond_umax_optab): New optabs.
10535 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10536 (COND_IOR, COND_XOR): New internal functions.
10537 * internal-fn.h (get_conditional_internal_fn): Declare.
10538 * internal-fn.c (cond_binary_direct): New macro.
10539 (expand_cond_binary_optab_fn): Likewise.
10540 (direct_cond_binary_optab_supported_p): Likewise.
10541 (get_conditional_internal_fn): New function.
10542 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10543 Cope with reduction statements that are vectorized as calls rather
10545 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10546 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10547 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10548 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10549 (UNSPEC_COND_EOR): New unspecs.
10550 (optab): Add mappings for them.
10551 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10552 (sve_int_op, sve_fp_op): New int attributes.
10554 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10555 Alan Hayward <alan.hayward@arm.com>
10556 David Sherwood <david.sherwood@arm.com>
10558 * optabs.def (while_ult_optab): New optab.
10559 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10560 * internal-fn.def (WHILE_ULT): New internal function.
10561 * internal-fn.h (direct_internal_fn_supported_p): New override
10562 that takes two types as argument.
10563 * internal-fn.c (while_direct): New macro.
10564 (expand_while_optab_fn): New function.
10565 (convert_optab_supported_p): Likewise.
10566 (direct_while_optab_supported_p): New macro.
10567 * wide-int.h (wi::udiv_ceil): New function.
10568 * tree-vectorizer.h (rgroup_masks): New structure.
10569 (vec_loop_masks): New typedef.
10570 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10571 and fully_masked_p.
10572 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10573 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10574 (vect_max_vf): New function.
10575 (slpeel_make_loop_iterate_ntimes): Delete.
10576 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10577 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10578 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10579 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10580 internal-fn.h, stor-layout.h and optabs-query.h.
10581 (vect_set_loop_mask): New function.
10582 (add_preheader_seq): Likewise.
10583 (add_header_seq): Likewise.
10584 (interleave_supported_p): Likewise.
10585 (vect_maybe_permute_loop_masks): Likewise.
10586 (vect_set_loop_masks_directly): Likewise.
10587 (vect_set_loop_condition_masked): Likewise.
10588 (vect_set_loop_condition_unmasked): New function, split out from
10589 slpeel_make_loop_iterate_ntimes.
10590 (slpeel_make_loop_iterate_ntimes): Rename to..
10591 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10592 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10593 (vect_do_peeling): Update call accordingly.
10594 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10596 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10597 mask_compare_type, can_fully_mask_p and fully_masked_p.
10598 (release_vec_loop_masks): New function.
10599 (_loop_vec_info): Use it to free the loop masks.
10600 (can_produce_all_loop_masks_p): New function.
10601 (vect_get_max_nscalars_per_iter): Likewise.
10602 (vect_verify_full_masking): Likewise.
10603 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10604 retries, and free the mask rgroups before retrying. Check loop-wide
10605 reasons for disallowing fully-masked loops. Make the final decision
10606 about whether use a fully-masked loop or not.
10607 (vect_estimate_min_profitable_iters): Do not assume that peeling
10608 for the number of iterations will be needed for fully-masked loops.
10609 (vectorizable_reduction): Disable fully-masked loops.
10610 (vectorizable_live_operation): Likewise.
10611 (vect_halve_mask_nunits): New function.
10612 (vect_double_mask_nunits): Likewise.
10613 (vect_record_loop_mask): Likewise.
10614 (vect_get_loop_mask): Likewise.
10615 (vect_transform_loop): Handle the case in which the final loop
10616 iteration might handle a partial vector. Call vect_set_loop_condition
10617 instead of slpeel_make_loop_iterate_ntimes.
10618 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10619 (check_load_store_masking): New function.
10620 (prepare_load_store_mask): Likewise.
10621 (vectorizable_store): Handle fully-masked loops.
10622 (vectorizable_load): Likewise.
10623 (supportable_widening_operation): Use vect_halve_mask_nunits for
10625 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10626 (vect_gen_while): New function.
10627 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10628 (aarch64_uqdec<mode>): New insn.
10630 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10631 Alan Hayward <alan.hayward@arm.com>
10632 David Sherwood <david.sherwood@arm.com>
10634 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10635 (reduc_xor_scal_optab): New optabs.
10636 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10637 (reduc_xor_scal_@var{m}): Document.
10638 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10639 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10640 internal functions.
10641 * fold-const-call.c (fold_const_call): Handle them.
10642 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10643 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10644 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10645 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10646 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10647 (UNSPEC_XORV): New unspecs.
10648 (optab): Add entries for them.
10649 (BITWISEV): New int iterator.
10650 (bit_reduc_op): New int attributes.
10652 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10653 Alan Hayward <alan.hayward@arm.com>
10654 David Sherwood <david.sherwood@arm.com>
10656 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10657 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10658 * optabs.def (vec_shl_insert_optab): New optab.
10659 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10660 (duplicate_and_interleave): Likewise.
10661 * tree-vect-loop.c: Include internal-fn.h.
10662 (neutral_op_for_slp_reduction): New function, split out from
10663 get_initial_defs_for_reduction.
10664 (get_initial_def_for_reduction): Handle option 2 for variable-length
10665 vectors by loading the neutral value into a vector and then shifting
10666 the initial value into element 0.
10667 (get_initial_defs_for_reduction): Replace the code argument with
10668 the neutral value calculated by neutral_op_for_slp_reduction.
10669 Use gimple_build_vector for constant-length vectors.
10670 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10671 but the first group_size elements have a neutral value.
10672 Use duplicate_and_interleave otherwise.
10673 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10674 Update call to get_initial_defs_for_reduction. Handle SLP
10675 reductions for variable-length vectors by creating one vector
10676 result for each scalar result, with the elements associated
10677 with other scalar results stubbed out with the neutral value.
10678 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10679 Require IFN_VEC_SHL_INSERT for double reductions on
10680 variable-length vectors, or SLP reductions that have
10681 a neutral value. Require can_duplicate_and_interleave_p
10682 support for variable-length unchained SLP reductions if there
10683 is no neutral value, such as for MIN/MAX reductions. Also require
10684 the number of vector elements to be a multiple of the number of
10685 SLP statements when doing variable-length unchained SLP reductions.
10686 Update call to vect_create_epilog_for_reduction.
10687 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10688 and remove initial values.
10689 (duplicate_and_interleave): Make public.
10690 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10691 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10693 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10694 Alan Hayward <alan.hayward@arm.com>
10695 David Sherwood <david.sherwood@arm.com>
10697 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10698 (can_duplicate_and_interleave_p): New function.
10699 (vect_get_and_check_slp_defs): Take the vector of statements
10700 rather than just the current one. Remove excess parentheses.
10701 Restriction rejectinon of vect_constant_def and vect_external_def
10702 for variable-length vectors to boolean types, or types for which
10703 can_duplicate_and_interleave_p is false.
10704 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10705 (duplicate_and_interleave): New function.
10706 (vect_get_constant_vectors): Use gimple_build_vector for
10707 constant-length vectors and suitable variable-length constant
10708 vectors. Use duplicate_and_interleave for other variable-length
10709 vectors. Don't defer the update when inserting new statements.
10711 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10712 Alan Hayward <alan.hayward@arm.com>
10713 David Sherwood <david.sherwood@arm.com>
10715 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10716 min_profitable_iters doesn't go negative.
10718 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10719 Alan Hayward <alan.hayward@arm.com>
10720 David Sherwood <david.sherwood@arm.com>
10722 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10723 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10724 * optabs.def (vec_mask_load_lanes_optab): New optab.
10725 (vec_mask_store_lanes_optab): Likewise.
10726 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10727 (MASK_STORE_LANES): Likewise.
10728 * internal-fn.c (mask_load_lanes_direct): New macro.
10729 (mask_store_lanes_direct): Likewise.
10730 (expand_mask_load_optab_fn): Handle masked operations.
10731 (expand_mask_load_lanes_optab_fn): New macro.
10732 (expand_mask_store_optab_fn): Handle masked operations.
10733 (expand_mask_store_lanes_optab_fn): New macro.
10734 (direct_mask_load_lanes_optab_supported_p): Likewise.
10735 (direct_mask_store_lanes_optab_supported_p): Likewise.
10736 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10738 (vect_load_lanes_supported): Likewise.
10739 * tree-vect-data-refs.c (strip_conversion): New function.
10740 (can_group_stmts_p): Likewise.
10741 (vect_analyze_data_ref_accesses): Use it instead of checking
10742 for a pair of assignments.
10743 (vect_store_lanes_supported): Take a masked_p parameter.
10744 (vect_load_lanes_supported): Likewise.
10745 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10746 vect_store_lanes_supported and vect_load_lanes_supported.
10747 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10748 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10749 parameter. Don't allow gaps for masked accesses.
10750 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10751 and vect_load_lanes_supported.
10752 (get_load_store_type): Take a masked_p parameter and update
10753 call to get_group_load_store_type.
10754 (vectorizable_store): Update call to get_load_store_type.
10755 Handle IFN_MASK_STORE_LANES.
10756 (vectorizable_load): Update call to get_load_store_type.
10757 Handle IFN_MASK_LOAD_LANES.
10759 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10760 Alan Hayward <alan.hayward@arm.com>
10761 David Sherwood <david.sherwood@arm.com>
10763 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10765 * config/aarch64/aarch64-protos.h
10766 (aarch64_sve_struct_memory_operand_p): Declare.
10767 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10768 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10769 (VPRED, vpred): Handle SVE structure modes.
10770 * config/aarch64/constraints.md (Utx): New constraint.
10771 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10772 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10773 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10774 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10775 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10776 structure modes. Split into pieces after RA.
10777 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10778 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10780 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10781 SVE structure modes.
10782 (aarch64_classify_address): Likewise.
10783 (sizetochar): Move earlier in file.
10784 (aarch64_print_operand): Handle SVE register lists.
10785 (aarch64_array_mode): New function.
10786 (aarch64_sve_struct_memory_operand_p): Likewise.
10787 (TARGET_ARRAY_MODE): Redefine.
10789 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10790 Alan Hayward <alan.hayward@arm.com>
10791 David Sherwood <david.sherwood@arm.com>
10793 * target.def (array_mode): New target hook.
10794 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10795 * doc/tm.texi: Regenerate.
10796 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10797 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10798 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10799 targetm.array_mode.
10800 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10803 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10804 Alan Hayward <alan.hayward@arm.com>
10805 David Sherwood <david.sherwood@arm.com>
10807 * fold-const.c (fold_binary_loc): Check the argument types
10808 rather than the result type when testing for a vector operation.
10810 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10812 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10813 * doc/tm.texi: Regenerate.
10815 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10816 Alan Hayward <alan.hayward@arm.com>
10817 David Sherwood <david.sherwood@arm.com>
10819 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10820 (sve): Document new AArch64 extension.
10821 * doc/md.texi (w): Extend the description of the AArch64
10822 constraint to include SVE vectors.
10823 (Upl, Upa): Document new AArch64 predicate constraints.
10824 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10826 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10827 (msve-vector-bits=): New option.
10828 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10829 SVE when these are disabled.
10830 (sve): New extension.
10831 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10832 modes. Adjust their number of units based on aarch64_sve_vg.
10833 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10834 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10835 aarch64_addr_query_type.
10836 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10837 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10838 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10839 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10840 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10841 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10842 (aarch64_simd_imm_zero_p): Delete.
10843 (aarch64_check_zero_based_sve_index_immediate): Declare.
10844 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10845 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10846 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10847 (aarch64_sve_float_mul_immediate_p): Likewise.
10848 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10849 rather than an rtx.
10850 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10851 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10852 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10853 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10854 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10855 (aarch64_regmode_natural_size): Likewise.
10856 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10857 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10859 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10860 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10861 for VG and the SVE predicate registers.
10862 (V_ALIASES): Add a "z"-prefixed alias.
10863 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10864 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10865 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10866 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10867 (REG_CLASS_NAMES): Add entries for them.
10868 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10869 and the predicate registers.
10870 (aarch64_sve_vg): Declare.
10871 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10872 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10873 (REGMODE_NATURAL_SIZE): Define.
10874 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10876 * config/aarch64/aarch64.c: Include cfgrtl.h.
10877 (simd_immediate_info): Add a constructor for series vectors,
10878 and an associated step field.
10879 (aarch64_sve_vg): New variable.
10880 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10881 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10882 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10883 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10884 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10885 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10886 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10887 (aarch64_get_mask_mode): New functions.
10888 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10889 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10890 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10891 predicate modes and predicate registers. Explicitly restrict
10892 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10893 to store a vector mode if it is recognized by
10894 aarch64_classify_vector_mode.
10895 (aarch64_regmode_natural_size): New function.
10896 (aarch64_hard_regno_caller_save_mode): Return the original mode
10898 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10899 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10900 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10901 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10903 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10904 does not overlap dest if the function is frame-related. Handle
10906 (aarch64_split_add_offset): New function.
10907 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10908 them aarch64_add_offset.
10909 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10910 and update call to aarch64_sub_sp.
10911 (aarch64_add_cfa_expression): New function.
10912 (aarch64_expand_prologue): Pass extra temporary registers to the
10913 functions above. Handle the case in which we need to emit new
10914 DW_CFA_expressions for registers that were originally saved
10915 relative to the stack pointer, but now have to be expressed
10916 relative to the frame pointer.
10917 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10919 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10920 IP0 and IP1 values for SVE frames.
10921 (aarch64_expand_vec_series): New function.
10922 (aarch64_expand_sve_widened_duplicate): Likewise.
10923 (aarch64_expand_sve_const_vector): Likewise.
10924 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10925 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10926 into the register, rather than emitting a SET directly.
10927 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10928 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10929 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10930 (offset_9bit_signed_scaled_p): New functions.
10931 (aarch64_replicate_bitmask_imm): New function.
10932 (aarch64_bitmask_imm): Use it.
10933 (aarch64_cannot_force_const_mem): Reject expressions involving
10934 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10935 (aarch64_classify_index): Handle SVE indices, by requiring
10936 a plain register index with a scale that matches the element size.
10937 (aarch64_classify_address): Handle SVE addresses. Assert that
10938 the mode of the address is VOIDmode or an integer mode.
10939 Update call to aarch64_classify_symbol.
10940 (aarch64_classify_symbolic_expression): Update call to
10941 aarch64_classify_symbol.
10942 (aarch64_const_vec_all_in_range_p): New function.
10943 (aarch64_print_vector_float_operand): Likewise.
10944 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10945 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10946 and the FP immediates 1.0 and 0.5.
10947 (aarch64_print_address_internal): Handle SVE addresses.
10948 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10949 (aarch64_regno_regclass): Handle predicate registers.
10950 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10952 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10953 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10954 (aarch64_convert_sve_vector_bits): New function.
10955 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10956 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10957 rather than an rtx.
10958 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10959 Handle SVE vector and predicate modes. Accept VL-based constants
10960 that need only one temporary register, and VL offsets that require
10961 no temporary registers.
10962 (aarch64_conditional_register_usage): Mark the predicate registers
10963 as fixed if SVE isn't available.
10964 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10965 Return true for SVE vector and predicate modes.
10966 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10967 rather than an unsigned int. Handle SVE modes.
10968 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10970 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10972 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10973 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10974 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10975 (aarch64_sve_float_mul_immediate_p): New functions.
10976 (aarch64_sve_valid_immediate): New function.
10977 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10978 Explicitly reject structure modes. Check for INDEX constants.
10979 Handle PTRUE and PFALSE constants.
10980 (aarch64_check_zero_based_sve_index_immediate): New function.
10981 (aarch64_simd_imm_zero_p): Delete.
10982 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10983 vector modes. Accept constants in the range of CNT[BHWD].
10984 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10985 ask for an Advanced SIMD mode.
10986 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10987 (aarch64_simd_vector_alignment): Handle SVE predicates.
10988 (aarch64_vectorize_preferred_vector_alignment): New function.
10989 (aarch64_simd_vector_alignment_reachable): Use it instead of
10991 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10992 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10994 (MAX_VECT_LEN): Delete.
10995 (expand_vec_perm_d): Add a vec_flags field.
10996 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10997 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10998 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11000 (aarch64_evpc_rev): Rename to...
11001 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11002 (aarch64_evpc_rev_global): New function.
11003 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11004 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11006 (aarch64_evpc_sve_tbl): New function.
11007 (aarch64_expand_vec_perm_const_1): Update after rename of
11008 aarch64_evpc_rev. Handle SVE permutes too, trying
11009 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11010 than aarch64_evpc_tbl.
11011 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11012 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11013 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11014 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11015 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11016 (aarch64_expand_sve_vcond): New functions.
11017 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11018 of aarch64_vector_mode_p.
11019 (aarch64_dwarf_poly_indeterminate_value): New function.
11020 (aarch64_compute_pressure_classes): Likewise.
11021 (aarch64_can_change_mode_class): Likewise.
11022 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11023 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11024 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11025 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11026 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11027 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11028 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11029 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11031 (Dn, Dl, Dr): Accept const as well as const_vector.
11032 (Dz): Likewise. Compare against CONST0_RTX.
11033 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11034 of "vector" where appropriate.
11035 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11036 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11037 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11038 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11039 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11040 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11041 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11042 (v_int_equiv): Extend to SVE modes.
11043 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11045 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11046 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11047 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11048 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11049 (SVE_COND_FP_CMP): New int iterators.
11050 (perm_hilo): Handle the new unpack unspecs.
11051 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11053 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11054 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11055 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11056 (aarch64_equality_operator, aarch64_constant_vector_operand)
11057 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11058 (aarch64_sve_nonimmediate_operand): Likewise.
11059 (aarch64_sve_general_operand): Likewise.
11060 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11061 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11062 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11063 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11064 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11065 (aarch64_sve_float_arith_immediate): Likewise.
11066 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11067 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11068 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11069 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11070 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11071 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11072 (aarch64_sve_float_arith_operand): Likewise.
11073 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11074 (aarch64_sve_float_mul_operand): Likewise.
11075 (aarch64_sve_vec_perm_operand): Likewise.
11076 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11077 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11078 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11079 as well as const_vector.
11080 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11081 in file. Use CONST0_RTX and CONSTM1_RTX.
11082 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11083 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11084 Use aarch64_simd_imm_zero.
11085 * config/aarch64/aarch64-sve.md: New file.
11086 * config/aarch64/aarch64.md: Include it.
11087 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11088 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11089 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11090 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11091 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11092 (sve): New attribute.
11093 (enabled): Disable instructions with the sve attribute unless
11095 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11096 aarch64_expand_mov_immediate.
11097 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11098 CNT[BHSD] immediates.
11099 (movti): Split CONST_POLY_INT moves into two halves.
11100 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11101 Split additions that need a temporary here if the destination
11102 is the stack pointer.
11103 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11104 (*add<mode>3_poly_1): New instruction.
11105 (set_clobber_cc): New expander.
11107 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11109 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11110 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11111 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11112 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11113 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11114 Change innermode from fixed_mode_size to machine_mode.
11115 (simplify_subreg): Update call accordingly. Handle a constant-sized
11116 subreg of a variable-length CONST_VECTOR.
11118 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11119 Alan Hayward <alan.hayward@arm.com>
11120 David Sherwood <david.sherwood@arm.com>
11122 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11123 (add_offset_to_base): New function, split out from...
11124 (create_mem_ref): ...here. When handling a scale other than 1,
11125 check first whether the address is valid without the offset.
11126 Add it into the base if so, leaving the index and scale as-is.
11128 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11131 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11132 fold_for_warn before checking if arg2 is INTEGER_CST.
11134 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11136 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11137 (store_multiple_operation): Delete.
11138 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11139 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11140 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11141 guarded by TARGET_STRING.
11142 (rs6000_output_load_multiple): Delete.
11143 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11144 OPTION_MASK_STRING / TARGET_STRING handling.
11145 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11146 (const rs6000_opt_masks) <"string">: Change mask to 0.
11147 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11148 (MASK_STRING): Delete.
11149 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11151 (load_multiple): Delete.
11158 (store_multiple): Delete.
11165 (movmemsi_8reg): Delete.
11166 (corresponding unnamed define_insn): Delete.
11167 (movmemsi_6reg): Delete.
11168 (corresponding unnamed define_insn): Delete.
11169 (movmemsi_4reg): Delete.
11170 (corresponding unnamed define_insn): Delete.
11171 (movmemsi_2reg): Delete.
11172 (corresponding unnamed define_insn): Delete.
11173 (movmemsi_1reg): Delete.
11174 (corresponding unnamed define_insn): Delete.
11175 * config/rs6000/rs6000.opt (mno-string): New.
11176 (mstring): Replace by deprecation warning stub.
11177 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11179 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11181 * regrename.c (regrename_do_replace): If replacing the same
11182 reg multiple times, try to reuse last created gen_raw_REG.
11185 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11186 main to workaround a bug in GDB.
11188 2018-01-12 Tom de Vries <tom@codesourcery.com>
11191 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11193 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11195 PR rtl-optimization/80481
11196 * ira-color.c (get_cap_member): New function.
11197 (allocnos_conflict_by_live_ranges_p): Use it.
11198 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11199 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11201 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11204 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11205 (*saddl_se_1): Ditto.
11206 (*ssubsi_1): Ditto.
11207 (*ssubl_se_1): Ditto.
11209 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11211 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11212 rather than wi::to_widest for DR_INITs.
11213 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11214 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11215 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11217 (vect_analyze_group_access_1): Note that here.
11219 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11221 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11222 polynomial type sizes.
11224 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11226 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11227 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11228 (gimple_add_tmp_var): Likewise.
11230 2018-01-12 Martin Liska <mliska@suse.cz>
11232 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11233 (gimple_alloc_sizes): Likewise.
11234 (dump_gimple_statistics): Use PRIu64 in printf format.
11235 * gimple.h: Change uint64_t to int.
11237 2018-01-12 Martin Liska <mliska@suse.cz>
11239 * tree-core.h: Use uint64_t instead of int.
11240 * tree.c (tree_node_counts): Likewise.
11241 (tree_node_sizes): Likewise.
11242 (dump_tree_statistics): Use PRIu64 in printf format.
11244 2018-01-12 Martin Liska <mliska@suse.cz>
11246 * Makefile.in: As qsort_chk is implemented in vec.c, add
11247 vec.o to linkage of gencfn-macros.
11248 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11249 passing the info to record_node_allocation_statistics.
11250 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11252 * ggc-common.c (struct ggc_usage): Add operator== and use
11253 it in operator< and compare function.
11254 * mem-stats.h (struct mem_usage): Likewise.
11255 * vec.c (struct vec_usage): Remove operator< and compare
11256 function. Can be simply inherited.
11258 2018-01-12 Martin Jambor <mjambor@suse.cz>
11261 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11262 * tree-ssa-math-opts.c: Include domwalk.h.
11263 (convert_mult_to_fma_1): New function.
11264 (fma_transformation_info): New type.
11265 (fma_deferring_state): Likewise.
11266 (cancel_fma_deferring): New function.
11267 (result_of_phi): Likewise.
11268 (last_fma_candidate_feeds_initial_phi): Likewise.
11269 (convert_mult_to_fma): Added deferring logic, split actual
11270 transformation to convert_mult_to_fma_1.
11271 (math_opts_dom_walker): New type.
11272 (math_opts_dom_walker::after_dom_children): New method, body moved
11273 here from pass_optimize_widening_mul::execute, added deferring logic
11275 (pass_optimize_widening_mul::execute): Moved most of code to
11276 math_opts_dom_walker::after_dom_children.
11277 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11278 * config/i386/i386.c (ix86_option_override_internal): Added
11279 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11281 2018-01-12 Richard Biener <rguenther@suse.de>
11284 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11285 inline instance vars.
11287 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11290 * config/rx/rx.c (rx_is_restricted_memory_address):
11291 Handle SUBREG case.
11293 2018-01-12 Richard Biener <rguenther@suse.de>
11295 PR tree-optimization/80846
11296 * target.def (split_reduction): New target hook.
11297 * targhooks.c (default_split_reduction): New function.
11298 * targhooks.h (default_split_reduction): Declare.
11299 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11300 target requests first reduce vectors by combining low and high
11302 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11303 (get_vectype_for_scalar_type_and_size): Export.
11304 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11305 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11306 * doc/tm.texi: Regenerate.
11307 * config/i386/i386.c (ix86_split_reduction): Implement
11308 TARGET_VECTORIZE_SPLIT_REDUCTION.
11310 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11313 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11314 in PIC mode except for TARGET_VXWORKS_RTP.
11315 * config/sparc/sparc.c: Include cfgrtl.h.
11316 (TARGET_INIT_PIC_REG): Define.
11317 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11318 (sparc_pic_register_p): New predicate.
11319 (sparc_legitimate_address_p): Use it.
11320 (sparc_legitimize_pic_address): Likewise.
11321 (sparc_delegitimize_address): Likewise.
11322 (sparc_mode_dependent_address_p): Likewise.
11323 (gen_load_pcrel_sym): Remove 4th parameter.
11324 (load_got_register): Adjust call to above. Remove obsolete stuff.
11325 (sparc_expand_prologue): Do not call load_got_register here.
11326 (sparc_flat_expand_prologue): Likewise.
11327 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11328 (sparc_use_pseudo_pic_reg): New function.
11329 (sparc_init_pic_reg): Likewise.
11330 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11331 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11333 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11335 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11336 Add item for branch_cost.
11338 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11340 PR rtl-optimization/83565
11341 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11342 not extend the result to a larger mode for rotate operations.
11343 (num_sign_bit_copies1): Likewise.
11345 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11348 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11350 Use values-Xc.o for -pedantic.
11351 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11353 2018-01-12 Martin Liska <mliska@suse.cz>
11356 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11358 (possible_polymorphic_call_targets): Use it.
11359 (ipa_devirt): Likewise.
11361 2018-01-12 Martin Liska <mliska@suse.cz>
11363 * profile-count.h (enum profile_quality): Use 0 as invalid
11364 enum value of profile_quality.
11366 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11368 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11369 -mext-string options.
11371 2018-01-12 Richard Biener <rguenther@suse.de>
11373 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11374 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11375 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11377 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11379 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11381 * configure.ac (--with-long-double-format): Add support for the
11382 configuration option to change the default long double format on
11384 * config.gcc (powerpc*-linux*-*): Likewise.
11385 * configure: Regenerate.
11386 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11387 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11388 used without modification.
11390 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11392 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11393 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11394 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11395 MISC_BUILTIN_SPEC_BARRIER.
11396 (rs6000_init_builtins): Likewise.
11397 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11399 (speculation_barrier): New define_insn.
11400 * doc/extend.texi: Document __builtin_speculation_barrier.
11402 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11405 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11406 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11407 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11409 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11410 integral modes instead of "ss" and "sd".
11411 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11412 vectors with 32-bit and 64-bit elements.
11413 (vecdupssescalarmodesuffix): New mode attribute.
11414 (vec_dup<mode>): Use it.
11416 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11419 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11420 frame if argument is passed on stack.
11422 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11425 * ree.c (combine_reaching_defs): Optimize also
11426 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11427 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11429 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11431 PR middle-end/83189
11432 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11434 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11436 PR middle-end/83718
11437 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11438 after they are computed.
11440 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11442 PR tree-optimization/83695
11443 * gimple-loop-linterchange.cc
11444 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11445 reset cached scev information after interchange.
11446 (pass_linterchange::execute): Remove call to scev_reset_htab.
11448 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11450 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11451 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11452 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11453 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11454 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11455 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11456 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11457 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11458 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11459 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11460 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11461 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11462 (V_lane_reg): Likewise.
11463 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11465 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11466 (vfmal_lane_low<mode>_intrinsic,
11467 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11468 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11469 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11470 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11471 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11472 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11474 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11476 * config/arm/arm-cpus.in (fp16fml): New feature.
11477 (ALL_SIMD): Add fp16fml.
11478 (armv8.2-a): Add fp16fml as an option.
11479 (armv8.3-a): Likewise.
11480 (armv8.4-a): Add fp16fml as part of fp16.
11481 * config/arm/arm.h (TARGET_FP16FML): Define.
11482 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11484 * config/arm/arm-modes.def (V2HF): Define.
11485 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11486 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11487 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11488 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11489 vfmsl_low, vfmsl_high): New set of builtins.
11490 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11491 (vfml_op): New code attribute.
11492 (VFMLHALVES): New int iterator.
11493 (VFML, VFMLSEL): New mode attributes.
11494 (V_reg): Define mapping for V2HF.
11495 (V_hi, V_lo): New mode attributes.
11496 (VF_constraint): Likewise.
11497 (vfml_half, vfml_half_selector): New int attributes.
11498 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11500 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11501 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11503 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11504 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11505 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11506 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11508 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11509 Document new effective target and option set.
11511 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11513 * config/arm/arm-cpus.in (armv8_4): New feature.
11514 (ARMv8_4a): New fgroup.
11515 (armv8.4-a): New arch.
11516 * config/arm/arm-tables.opt: Regenerate.
11517 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11518 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11519 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11520 Add matching rules for -march=armv8.4-a and extensions.
11521 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11523 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11526 * config/rx/rx.md (BW): New mode attribute.
11527 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11529 2018-01-11 Richard Biener <rguenther@suse.de>
11531 PR tree-optimization/83435
11532 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11533 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11534 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11536 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11537 Alan Hayward <alan.hayward@arm.com>
11538 David Sherwood <david.sherwood@arm.com>
11540 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11542 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11543 (aarch64_print_address_internal): Use it to check for a zero offset.
11545 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11546 Alan Hayward <alan.hayward@arm.com>
11547 David Sherwood <david.sherwood@arm.com>
11549 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11550 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11551 Return a poly_int64 rather than a HOST_WIDE_INT.
11552 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11553 rather than a HOST_WIDE_INT.
11554 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11555 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11556 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11557 final_offset from HOST_WIDE_INT to poly_int64.
11558 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11559 to_constant when getting the number of units in an Advanced SIMD
11561 (aarch64_builtin_vectorized_function): Check for a constant number
11563 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11565 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11566 attribute instead of GET_MODE_NUNITS.
11567 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11568 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11569 GET_MODE_SIZE for fixed-size registers.
11570 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11571 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11572 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11573 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11574 (aarch64_print_operand, aarch64_print_address_internal)
11575 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11576 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11577 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11578 Handle polynomial GET_MODE_SIZE.
11579 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11580 wider than SImode without modification.
11581 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11582 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11583 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11584 passing and returning SVE modes.
11585 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11586 rather than GEN_INT.
11587 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11588 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11589 (aarch64_allocate_and_probe_stack_space): Likewise.
11590 (aarch64_layout_frame): Cope with polynomial offsets.
11591 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11592 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11593 polynomial offsets.
11594 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11595 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11596 poly_int64 rather than a HOST_WIDE_INT.
11597 (aarch64_get_separate_components, aarch64_process_components)
11598 (aarch64_expand_prologue, aarch64_expand_epilogue)
11599 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11600 (aarch64_anchor_offset): New function, split out from...
11601 (aarch64_legitimize_address): ...here.
11602 (aarch64_builtin_vectorization_cost): Handle polynomial
11603 TYPE_VECTOR_SUBPARTS.
11604 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11606 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11607 number of elements from the PARALLEL rather than the mode.
11608 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11609 rather than GET_MODE_BITSIZE.
11610 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11611 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11612 (aarch64_expand_vec_perm_const_1): Handle polynomial
11613 d->perm.length () and d->perm elements.
11614 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11615 Apply to_constant to d->perm elements.
11616 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11617 polynomial CONST_VECTOR_NUNITS.
11618 (aarch64_move_pointer): Take amount as a poly_int64 rather
11620 (aarch64_progress_pointer): Avoid temporary variable.
11621 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11622 the mode attribute instead of GET_MODE.
11624 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11625 Alan Hayward <alan.hayward@arm.com>
11626 David Sherwood <david.sherwood@arm.com>
11628 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11629 x exists before using it.
11630 (aarch64_add_constant_internal): Rename to...
11631 (aarch64_add_offset_1): ...this. Replace regnum with separate
11632 src and dest rtxes. Handle the case in which they're different,
11633 including when the offset is zero. Replace scratchreg with an rtx.
11634 Use 2 additions if there is no spare register into which we can
11635 move a 16-bit constant.
11636 (aarch64_add_constant): Delete.
11637 (aarch64_add_offset): Replace reg with separate src and dest
11638 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11639 Use aarch64_add_offset_1.
11640 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11641 an rtx rather than an int. Take the delta as a poly_int64
11642 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11643 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11644 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11645 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11646 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11647 and aarch64_add_sp.
11648 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11649 aarch64_add_constant.
11651 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11653 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11654 Use scalar_float_mode.
11656 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11658 * config/aarch64/aarch64-simd.md
11659 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11660 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11661 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11662 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11663 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11664 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11665 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11666 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11667 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11668 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11670 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11673 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11674 targ_options->x_arm_arch_string is non NULL.
11676 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11678 * config/aarch64/aarch64.h
11679 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11681 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11684 * expmed.c (emit_store_flag_force): Swap if const op0
11685 and change VOIDmode to mode of op0.
11687 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11689 PR rtl-optimization/83761
11690 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11691 than bytes to mode_for_size.
11693 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11695 PR middle-end/83189
11696 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11697 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11700 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11702 PR middle-end/83575
11703 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11704 when in layout mode.
11705 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11706 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11709 2018-01-10 Michael Collison <michael.collison@arm.com>
11711 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11712 * config/aarch64/aarch64-option-extension.def: Add
11713 AARCH64_OPT_EXTENSION of 'fp16fml'.
11714 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11715 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11716 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11717 * config/aarch64/constraints.md (Ui7): New constraint.
11718 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11719 (VFMLA_SEL_W): Ditto.
11722 (VFMLA16_LOW): New int iterator.
11723 (VFMLA16_HIGH): Ditto.
11724 (UNSPEC_FMLAL): New unspec.
11725 (UNSPEC_FMLSL): Ditto.
11726 (UNSPEC_FMLAL2): Ditto.
11727 (UNSPEC_FMLSL2): Ditto.
11728 (f16mac): New code attribute.
11729 * config/aarch64/aarch64-simd-builtins.def
11730 (aarch64_fmlal_lowv2sf): Ditto.
11731 (aarch64_fmlsl_lowv2sf): Ditto.
11732 (aarch64_fmlalq_lowv4sf): Ditto.
11733 (aarch64_fmlslq_lowv4sf): Ditto.
11734 (aarch64_fmlal_highv2sf): Ditto.
11735 (aarch64_fmlsl_highv2sf): Ditto.
11736 (aarch64_fmlalq_highv4sf): Ditto.
11737 (aarch64_fmlslq_highv4sf): Ditto.
11738 (aarch64_fmlal_lane_lowv2sf): Ditto.
11739 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11740 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11741 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11742 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11743 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11744 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11745 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11746 (aarch64_fmlal_lane_highv2sf): Ditto.
11747 (aarch64_fmlsl_lane_highv2sf): Ditto.
11748 (aarch64_fmlal_laneq_highv2sf): Ditto.
11749 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11750 (aarch64_fmlalq_lane_highv4sf): Ditto.
11751 (aarch64_fmlsl_lane_highv4sf): Ditto.
11752 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11753 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11754 * config/aarch64/aarch64-simd.md:
11755 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11756 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11757 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11758 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11759 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11760 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11761 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11762 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11763 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11764 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11765 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11766 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11767 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11768 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11769 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11770 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11771 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11772 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11773 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11774 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11775 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11776 (vfmlsl_low_u32): Ditto.
11777 (vfmlalq_low_u32): Ditto.
11778 (vfmlslq_low_u32): Ditto.
11779 (vfmlal_high_u32): Ditto.
11780 (vfmlsl_high_u32): Ditto.
11781 (vfmlalq_high_u32): Ditto.
11782 (vfmlslq_high_u32): Ditto.
11783 (vfmlal_lane_low_u32): Ditto.
11784 (vfmlsl_lane_low_u32): Ditto.
11785 (vfmlal_laneq_low_u32): Ditto.
11786 (vfmlsl_laneq_low_u32): Ditto.
11787 (vfmlalq_lane_low_u32): Ditto.
11788 (vfmlslq_lane_low_u32): Ditto.
11789 (vfmlalq_laneq_low_u32): Ditto.
11790 (vfmlslq_laneq_low_u32): Ditto.
11791 (vfmlal_lane_high_u32): Ditto.
11792 (vfmlsl_lane_high_u32): Ditto.
11793 (vfmlal_laneq_high_u32): Ditto.
11794 (vfmlsl_laneq_high_u32): Ditto.
11795 (vfmlalq_lane_high_u32): Ditto.
11796 (vfmlslq_lane_high_u32): Ditto.
11797 (vfmlalq_laneq_high_u32): Ditto.
11798 (vfmlslq_laneq_high_u32): Ditto.
11799 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11800 (AARCH64_FL_FOR_ARCH8_4): New.
11801 (AARCH64_ISA_F16FML): New ISA flag.
11802 (TARGET_F16FML): New feature flag for fp16fml.
11803 (doc/invoke.texi): Document new fp16fml option.
11805 2018-01-10 Michael Collison <michael.collison@arm.com>
11807 * config/aarch64/aarch64-builtins.c:
11808 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11809 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11810 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11811 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11812 (AARCH64_ISA_SHA3): New ISA flag.
11813 (TARGET_SHA3): New feature flag for sha3.
11814 * config/aarch64/iterators.md (sha512_op): New int attribute.
11815 (CRYPTO_SHA512): New int iterator.
11816 (UNSPEC_SHA512H): New unspec.
11817 (UNSPEC_SHA512H2): Ditto.
11818 (UNSPEC_SHA512SU0): Ditto.
11819 (UNSPEC_SHA512SU1): Ditto.
11820 * config/aarch64/aarch64-simd-builtins.def
11821 (aarch64_crypto_sha512hqv2di): New builtin.
11822 (aarch64_crypto_sha512h2qv2di): Ditto.
11823 (aarch64_crypto_sha512su0qv2di): Ditto.
11824 (aarch64_crypto_sha512su1qv2di): Ditto.
11825 (aarch64_eor3qv8hi): Ditto.
11826 (aarch64_rax1qv2di): Ditto.
11827 (aarch64_xarqv2di): Ditto.
11828 (aarch64_bcaxqv8hi): Ditto.
11829 * config/aarch64/aarch64-simd.md:
11830 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11831 (aarch64_crypto_sha512su0qv2di): Ditto.
11832 (aarch64_crypto_sha512su1qv2di): Ditto.
11833 (aarch64_eor3qv8hi): Ditto.
11834 (aarch64_rax1qv2di): Ditto.
11835 (aarch64_xarqv2di): Ditto.
11836 (aarch64_bcaxqv8hi): Ditto.
11837 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11838 (vsha512h2q_u64): Ditto.
11839 (vsha512su0q_u64): Ditto.
11840 (vsha512su1q_u64): Ditto.
11841 (veor3q_u16): Ditto.
11842 (vrax1q_u64): Ditto.
11843 (vxarq_u64): Ditto.
11844 (vbcaxq_u16): Ditto.
11845 * config/arm/types.md (crypto_sha512): New type attribute.
11846 (crypto_sha3): Ditto.
11847 (doc/invoke.texi): Document new sha3 option.
11849 2018-01-10 Michael Collison <michael.collison@arm.com>
11851 * config/aarch64/aarch64-builtins.c:
11852 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11853 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11854 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11855 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11856 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11857 (AARCH64_ISA_SM4): New ISA flag.
11858 (TARGET_SM4): New feature flag for sm4.
11859 * config/aarch64/aarch64-simd-builtins.def
11860 (aarch64_sm3ss1qv4si): Ditto.
11861 (aarch64_sm3tt1aq4si): Ditto.
11862 (aarch64_sm3tt1bq4si): Ditto.
11863 (aarch64_sm3tt2aq4si): Ditto.
11864 (aarch64_sm3tt2bq4si): Ditto.
11865 (aarch64_sm3partw1qv4si): Ditto.
11866 (aarch64_sm3partw2qv4si): Ditto.
11867 (aarch64_sm4eqv4si): Ditto.
11868 (aarch64_sm4ekeyqv4si): Ditto.
11869 * config/aarch64/aarch64-simd.md:
11870 (aarch64_sm3ss1qv4si): Ditto.
11871 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11872 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11873 (aarch64_sm4eqv4si): Ditto.
11874 (aarch64_sm4ekeyqv4si): Ditto.
11875 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11876 (sm3part_op): Ditto.
11877 (CRYPTO_SM3TT): Ditto.
11878 (CRYPTO_SM3PART): Ditto.
11879 (UNSPEC_SM3SS1): New unspec.
11880 (UNSPEC_SM3TT1A): Ditto.
11881 (UNSPEC_SM3TT1B): Ditto.
11882 (UNSPEC_SM3TT2A): Ditto.
11883 (UNSPEC_SM3TT2B): Ditto.
11884 (UNSPEC_SM3PARTW1): Ditto.
11885 (UNSPEC_SM3PARTW2): Ditto.
11886 (UNSPEC_SM4E): Ditto.
11887 (UNSPEC_SM4EKEY): Ditto.
11888 * config/aarch64/constraints.md (Ui2): New constraint.
11889 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11890 * config/arm/types.md (crypto_sm3): New type attribute.
11891 (crypto_sm4): Ditto.
11892 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11893 (vsm3tt1aq_u32): Ditto.
11894 (vsm3tt1bq_u32): Ditto.
11895 (vsm3tt2aq_u32): Ditto.
11896 (vsm3tt2bq_u32): Ditto.
11897 (vsm3partw1q_u32): Ditto.
11898 (vsm3partw2q_u32): Ditto.
11899 (vsm4eq_u32): Ditto.
11900 (vsm4ekeyq_u32): Ditto.
11901 (doc/invoke.texi): Document new sm4 option.
11903 2018-01-10 Michael Collison <michael.collison@arm.com>
11905 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11906 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11907 (AARCH64_FL_FOR_ARCH8_4): New.
11908 (AARCH64_FL_V8_4): New flag.
11909 (doc/invoke.texi): Document new armv8.4-a option.
11911 2018-01-10 Michael Collison <michael.collison@arm.com>
11913 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11914 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11915 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11916 * config/aarch64/aarch64-option-extension.def: Add
11917 AARCH64_OPT_EXTENSION of 'sha2'.
11918 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11919 (crypto): Disable sha2 and aes if crypto disabled.
11920 (crypto): Enable aes and sha2 if enabled.
11921 (simd): Disable sha2 and aes if simd disabled.
11922 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11924 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11925 (TARGET_SHA2): New feature flag for sha2.
11926 (TARGET_AES): New feature flag for aes.
11927 * config/aarch64/aarch64-simd.md:
11928 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11929 conditional on TARGET_AES.
11930 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11931 (aarch64_crypto_sha1hsi): Make pattern conditional
11933 (aarch64_crypto_sha1hv4si): Ditto.
11934 (aarch64_be_crypto_sha1hv4si): Ditto.
11935 (aarch64_crypto_sha1su1v4si): Ditto.
11936 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11937 (aarch64_crypto_sha1su0v4si): Ditto.
11938 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11939 (aarch64_crypto_sha256su0v4si): Ditto.
11940 (aarch64_crypto_sha256su1v4si): Ditto.
11941 (doc/invoke.texi): Document new aes and sha2 options.
11943 2018-01-10 Martin Sebor <msebor@redhat.com>
11945 PR tree-optimization/83781
11946 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11949 2018-01-11 Martin Sebor <msebor@gmail.com>
11950 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11952 PR tree-optimization/83501
11953 PR tree-optimization/81703
11955 * tree-ssa-strlen.c (get_string_cst): Rename...
11956 (get_string_len): ...to this. Handle global constants.
11957 (handle_char_store): Adjust.
11959 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11960 Jim Wilson <jimw@sifive.com>
11962 * config/riscv/riscv-protos.h (riscv_output_return): New.
11963 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11964 (riscv_attribute_table, riscv_output_return),
11965 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11966 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11967 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11969 (riscv_expand_prologue): Add early return for naked function.
11970 (riscv_expand_epilogue): Likewise.
11971 (riscv_function_ok_for_sibcall): Return false for naked function.
11972 (riscv_set_current_function): New.
11973 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11974 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11975 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11976 * doc/extend.texi (RISC-V Function Attributes): New.
11978 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11980 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11981 check for 128-bit long double before checking TCmode.
11982 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11983 128-bit long doubles before checking TFmode or TCmode.
11984 (FLOAT128_IBM_P): Likewise.
11986 2018-01-10 Martin Sebor <msebor@redhat.com>
11988 PR tree-optimization/83671
11989 * builtins.c (c_strlen): Unconditionally return zero for the empty
11991 Use -Warray-bounds for warnings.
11992 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11993 for non-constant array indices with COMPONENT_REF, arrays of
11994 arrays, and pointers to arrays.
11995 (gimple_fold_builtin_strlen): Determine and set length range for
11996 non-constant character arrays.
11998 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12000 PR middle-end/81897
12001 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12004 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12006 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12008 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12011 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12012 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12013 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12014 indexed_or_indirect_operand predicate.
12015 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12016 (*vsx_le_perm_load_v8hi): Likewise.
12017 (*vsx_le_perm_load_v16qi): Likewise.
12018 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12019 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12020 (*vsx_le_perm_store_v8hi): Likewise.
12021 (*vsx_le_perm_store_v16qi): Likewise.
12022 (eight unnamed splitters): Likewise.
12024 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12026 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12027 * config/rs6000/emmintrin.h: Likewise.
12028 * config/rs6000/mmintrin.h: Likewise.
12029 * config/rs6000/xmmintrin.h: Likewise.
12031 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12034 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12036 * tree.c (tree_nop_conversion): Return true for location wrapper
12038 (maybe_wrap_with_location): New function.
12039 (selftest::check_strip_nops): New function.
12040 (selftest::test_location_wrappers): New function.
12041 (selftest::tree_c_tests): Call it.
12042 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12043 (maybe_wrap_with_location): New decl.
12044 (EXPR_LOCATION_WRAPPER_P): New macro.
12045 (location_wrapper_p): New inline function.
12046 (tree_strip_any_location_wrapper): New inline function.
12048 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12051 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12052 stack_realign_offset for the largest alignment of stack slot
12054 (ix86_find_max_used_stack_alignment): New function.
12055 (ix86_finalize_stack_frame_flags): Use it. Set
12056 max_used_stack_alignment if we don't realign stack.
12057 * config/i386/i386.h (machine_function): Add
12058 max_used_stack_alignment.
12060 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12062 * config/arm/arm.opt (-mbranch-cost): New option.
12063 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12066 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12069 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12070 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12072 2018-01-10 Richard Biener <rguenther@suse.de>
12075 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12076 early out so it also covers the case where we have a non-NULL
12079 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12081 PR tree-optimization/83753
12082 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12083 for non-strided grouped accesses if the number of elements is 1.
12085 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12088 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12089 * i386.h (TARGET_USE_GATHER): Define.
12090 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12092 2018-01-10 Martin Liska <mliska@suse.cz>
12095 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12096 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12098 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12099 CLEANUP_NO_PARTITIONING is not set.
12101 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12103 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12104 for vectors, as a partial revert of r254296.
12105 * rtl.h (const_vec_p): Delete.
12106 (const_vec_duplicate_p): Don't test for vector CONSTs.
12107 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12108 * expmed.c (make_tree): Likewise.
12111 * common.md (E, F): Use CONSTANT_P instead of checking for
12113 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12114 checking for CONST_VECTOR.
12116 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12118 PR middle-end/83575
12119 * predict.c (force_edge_cold): Handle in more sane way edges
12120 with no prediction.
12122 2018-01-09 Carl Love <cel@us.ibm.com>
12124 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12126 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12127 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12128 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12129 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12130 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12131 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12132 * config/rs6000/rs6000-protos.h: Add extern defition for
12133 rs6000_generate_float2_double_code.
12134 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12136 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12137 (float2_v2df): Add define_expand.
12139 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12142 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12143 op_mode in the force_to_mode call.
12145 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12147 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12148 instead of checking each element individually.
12149 (aarch64_evpc_uzp): Likewise.
12150 (aarch64_evpc_zip): Likewise.
12151 (aarch64_evpc_ext): Likewise.
12152 (aarch64_evpc_rev): Likewise.
12153 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12154 instead of checking each element individually. Return true without
12156 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12157 whether all selected elements come from the same input, instead of
12158 checking each element individually. Remove calls to gen_rtx_REG,
12159 start_sequence and end_sequence and instead assert that no rtl is
12162 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12164 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12165 order of HIGH and CONST checks.
12167 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12169 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12170 if the destination isn't an SSA_NAME.
12172 2018-01-09 Richard Biener <rguenther@suse.de>
12174 PR tree-optimization/83668
12175 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12177 (canonicalize_loop_form): ... here, renamed from ...
12178 (canonicalize_loop_closed_ssa_form): ... this and amended to
12179 swap successor edges for loop exit blocks to make us use
12180 the RPO order we need for initial schedule generation.
12182 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12184 PR tree-optimization/64811
12185 * match.pd: When optimizing comparisons with Inf, avoid
12186 introducing or losing exceptions from comparisons with NaN.
12188 2018-01-09 Martin Liska <mliska@suse.cz>
12191 * asan.c (shadow_mem_size): Add gcc_assert.
12193 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12195 Don't save registers in main().
12198 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12199 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12200 * config/avr/avr.c (avr_set_current_function): Don't error if
12201 naked, OS_task or OS_main are specified at the same time.
12202 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12204 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12206 * common/config/avr/avr-common.c (avr_option_optimization_table):
12207 Switch on -mmain-is-OS_task for optimizing compilations.
12209 2018-01-09 Richard Biener <rguenther@suse.de>
12211 PR tree-optimization/83572
12212 * graphite.c: Include cfganal.h.
12213 (graphite_transform_loops): Connect infinite loops to exit
12214 and remove fake edges at the end.
12216 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12218 * ipa-inline.c (edge_badness): Revert accidental checkin.
12220 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12223 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12224 symbols; not inline clones.
12226 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12229 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12230 hard registers. Formatting fixes.
12232 PR preprocessor/83722
12233 * gcc.c (try_generate_repro): Pass
12234 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12235 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12238 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12239 Kito Cheng <kito.cheng@gmail.com>
12241 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12242 (riscv_leaf_function_p): Delete.
12243 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12245 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12247 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12249 (do_ifelse): New function.
12250 (do_isel): New function.
12251 (do_sub3): New function.
12252 (do_add3): New function.
12253 (do_load_mask_compare): New function.
12254 (do_overlap_load_compare): New function.
12255 (expand_compare_loop): New function.
12256 (expand_block_compare): Call expand_compare_loop() when appropriate.
12257 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12258 option description.
12259 (-mblock-compare-inline-loop-limit): New option.
12261 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12264 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12265 Reverse order of second and third operands in first alternative.
12266 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12267 of first and second elements in UNSPEC_VPERMR vector.
12268 (altivec_expand_vec_perm_le): Likewise.
12270 2018-01-08 Jeff Law <law@redhat.com>
12272 PR rtl-optimizatin/81308
12273 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12274 (process_switch): If group_case_labels makes a change, then set
12276 (pass_convert_switch::execute): If a switch is converted, then
12277 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12279 PR rtl-optimization/81308
12280 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12283 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12285 PR target/83663 - Revert r255946
12286 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12287 generation for cases where splatting a value is not useful.
12288 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12289 across a vec_duplicate and a paradoxical subreg forming a vector
12290 mode to a vec_concat.
12292 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12294 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12295 -march=armv8.3-a variants.
12296 * config/arm/t-multilib: Likewise.
12297 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12299 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12301 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12303 (cceq_ior_compare_complement): Give it a name so I can use it, and
12304 change boolean_or_operator predicate to boolean_operator so it can
12305 be used to generate a crand.
12306 (eqne): New code iterator.
12307 (bd/bd_neg): New code_attrs.
12308 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12309 a single define_insn.
12310 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12311 decrement (bdnzt/bdnzf/bdzt/bdzf).
12312 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12313 with the new names of the branch decrement patterns, and added the
12314 names of the branch decrement conditional patterns.
12316 2018-01-08 Richard Biener <rguenther@suse.de>
12318 PR tree-optimization/83563
12319 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12322 2018-01-08 Richard Biener <rguenther@suse.de>
12324 PR middle-end/83713
12325 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12327 2018-01-08 Richard Biener <rguenther@suse.de>
12329 PR tree-optimization/83685
12330 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12331 references to abnormals.
12333 2018-01-08 Richard Biener <rguenther@suse.de>
12336 * dwarf2out.c (output_indirect_strings): Handle empty
12337 skeleton_debug_str_hash.
12338 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12340 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12342 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12343 (emit_store_direct): Likewise.
12344 (arc_trampoline_adjust_address): Likewise.
12345 (arc_asm_trampoline_template): New function.
12346 (arc_initialize_trampoline): Use asm_trampoline_template.
12347 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12348 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12349 * config/arc/arc.md (flush_icache): Delete pattern.
12351 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12353 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12354 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12357 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12360 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12361 by not USED_FOR_TARGET.
12362 (make_pass_resolve_sw_modes): Likewise.
12364 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12366 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12369 2018-01-08 Richard Biener <rguenther@suse.de>
12371 PR middle-end/83580
12372 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12374 2018-01-08 Richard Biener <rguenther@suse.de>
12376 PR middle-end/83517
12377 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12379 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12381 PR middle-end/81897
12382 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12383 basic blocks with a small number of successors.
12384 (convert_control_dep_chain_into_preds): Improve handling of
12386 (dump_predicates): Split apart into...
12387 (dump_pred_chain): ...here...
12388 (dump_pred_info): ...and here.
12389 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12390 (can_chain_union_be_invalidated_p): Improve check for invalidation
12392 (uninit_uses_cannot_happen): Avoid unnecessary if
12393 convert_control_dep_chain_into_preds yielded nothing.
12395 2018-01-06 Martin Sebor <msebor@redhat.com>
12397 PR tree-optimization/83640
12398 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12399 subtracting negative offset from size.
12400 (builtin_access::overlap): Adjust offset bounds of the access to fall
12401 within the size of the object if possible.
12403 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12405 PR rtl-optimization/83699
12406 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12407 extract_bit_field_as_subreg to cases in which the extracted
12408 value is also a vector.
12410 * lra-constraints.c (process_alt_operands): Test for the equivalence
12411 substitutions when detecting a possible reload cycle.
12413 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12416 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12417 by default if flag_selective_schedling{,2}. Formatting fixes.
12419 PR rtl-optimization/83682
12420 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12421 if it has non-VECTOR_MODE element mode.
12422 (vec_duplicate_p): Likewise.
12424 PR middle-end/83694
12425 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12426 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12428 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12431 * config/i386/i386-builtin.def
12432 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12433 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12434 Require also OPTION_MASK_ISA_AVX512F in addition to
12435 OPTION_MASK_ISA_GFNI.
12436 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12437 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12438 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12439 to OPTION_MASK_ISA_GFNI.
12440 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12441 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12442 OPTION_MASK_ISA_AVX512BW.
12443 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12444 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12445 addition to OPTION_MASK_ISA_GFNI.
12446 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12447 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12448 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12449 to OPTION_MASK_ISA_GFNI.
12450 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12451 a requirement for all ISAs rather than any of them with a few
12453 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12455 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12456 bitmasks to be enabled with 3 exceptions, instead of requiring any
12457 enabled ISA with lots of exceptions.
12458 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12459 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12460 Change avx512bw in isa attribute to avx512f.
12461 * config/i386/sgxintrin.h: Add license boilerplate.
12462 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12463 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12464 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12465 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12467 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12468 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12469 temporarily sse2 rather than sse if not enabled already.
12472 * config/i386/sse.md (VI248_VLBW): Rename to ...
12473 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12474 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12475 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12476 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12477 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12478 mode iterator instead of VI248_VLBW.
12480 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12482 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12483 (record_modified): Skip clobbers; add debug output.
12484 (param_change_prob): Use sreal frequencies.
12486 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12488 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12489 punt for user-aligned variables.
12491 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12493 * tree-chrec.c (chrec_contains_symbols): Return true for
12496 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12499 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12500 of (x|y) == x for BICS pattern.
12502 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12504 PR tree-optimization/83605
12505 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12506 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12509 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12511 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12512 * config/epiphany/rtems.h: New file.
12514 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12515 Uros Bizjak <ubizjak@gmail.com>
12518 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12519 QIreg_operand instead of register_operand predicate.
12520 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12521 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12522 comments instead of -fmitigate[-_]rop.
12524 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12527 * cgraphunit.c (symbol_table::compile): Switch to text_section
12528 before calling assembly_start debug hook.
12529 * run-rtl-passes.c (run_rtl_passes): Likewise.
12532 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12534 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12535 range_int_cst_p rather than !symbolic_range_p before calling
12536 extract_range_from_multiplicative_op_1.
12538 2018-01-04 Jeff Law <law@redhat.com>
12540 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12541 redundant test in assertion.
12543 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12545 * doc/rtl.texi: Document machine_mode wrapper classes.
12547 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12549 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12550 using tree_to_uhwi.
12552 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12554 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12555 the VEC_PERM_EXPR fold to fail.
12557 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12560 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12561 to switched_sections.
12563 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12566 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12567 test for d.testing.
12569 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12572 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12573 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12575 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12578 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12579 is BLKmode and bitpos not zero or mode change is needed.
12581 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12584 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12587 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12590 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12591 instead of MULT rtx. Update all corresponding splitters.
12592 (*saddl_se): Ditto.
12593 (*ssub<modesuffix>): Ditto.
12594 (*ssubl_se): Ditto.
12595 (*cmp_sadd_di): Update split patterns.
12596 (*cmp_sadd_si): Ditto.
12597 (*cmp_sadd_sidi): Ditto.
12598 (*cmp_ssub_di): Ditto.
12599 (*cmp_ssub_si): Ditto.
12600 (*cmp_ssub_sidi): Ditto.
12601 * config/alpha/predicates.md (const23_operand): New predicate.
12602 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12603 Look for ASHIFT, not MULT inner operand.
12604 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12606 2018-01-04 Martin Liska <mliska@suse.cz>
12608 PR gcov-profile/83669
12609 * gcov.c (output_intermediate_file): Add version to intermediate
12611 * doc/gcov.texi: Document new field 'version' in intermediate
12612 file format. Fix location of '-k' option of gcov command.
12614 2018-01-04 Martin Liska <mliska@suse.cz>
12617 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12619 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12621 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12623 2018-01-03 Martin Sebor <msebor@redhat.com>
12625 PR tree-optimization/83655
12626 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12627 checking calls with invalid arguments.
12629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12631 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12632 (vectorizable_mask_load_store): Delete.
12633 (vectorizable_call): Return false for masked loads and stores.
12634 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12635 instead of gimple_assign_rhs1.
12636 (vectorizable_load): Handle IFN_MASK_LOAD.
12637 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12641 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12643 (vectorizable_mask_load_store): ...here.
12644 (vectorizable_load): ...and here.
12646 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12648 * tree-vect-stmts.c (vect_build_all_ones_mask)
12649 (vect_build_zero_merge_argument): New functions, split out from...
12650 (vectorizable_load): ...here.
12652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12654 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12656 (vectorizable_mask_load_store): ...here.
12657 (vectorizable_store): ...and here.
12659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12661 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12663 (vectorizable_mask_load_store): ...here.
12665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12667 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12668 (vect_model_store_cost): Take a vec_load_store_type instead of a
12670 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12671 (vect_model_store_cost): Take a vec_load_store_type instead of a
12673 (vectorizable_mask_load_store): Update accordingly.
12674 (vectorizable_store): Likewise.
12675 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12679 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12680 IFN_MASK_LOAD calls here rather than...
12681 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12684 Alan Hayward <alan.hayward@arm.com>
12685 David Sherwood <david.sherwood@arm.com>
12687 * expmed.c (extract_bit_field_1): For vector extracts,
12688 fall back to extract_bit_field_as_subreg if vec_extract
12691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12692 Alan Hayward <alan.hayward@arm.com>
12693 David Sherwood <david.sherwood@arm.com>
12695 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12696 they are variable or constant sized.
12697 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12698 slots for constant-sized data.
12700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12701 Alan Hayward <alan.hayward@arm.com>
12702 David Sherwood <david.sherwood@arm.com>
12704 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12705 handling COND_EXPRs with boolean comparisons, try to find a better
12706 basis for the mask type than the boolean itself.
12708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12710 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12711 is calculated and how it can be overridden.
12712 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12713 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12715 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12719 Alan Hayward <alan.hayward@arm.com>
12720 David Sherwood <david.sherwood@arm.com>
12722 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12723 Remove the mode argument.
12724 (aarch64_simd_valid_immediate): Remove the mode and inverse
12726 * config/aarch64/iterators.md (bitsize): New iterator.
12727 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12728 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12729 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12730 aarch64_simd_valid_immediate.
12731 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12732 (aarch64_reg_or_bic_imm): Likewise.
12733 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12734 with an insn_type enum and msl with a modifier_type enum.
12735 Replace element_width with a scalar_mode. Change the shift
12736 to unsigned int. Add constructors for scalar_float_mode and
12737 scalar_int_mode elements.
12738 (aarch64_vect_float_const_representable_p): Delete.
12739 (aarch64_can_const_movi_rtx_p)
12740 (aarch64_simd_scalar_immediate_valid_for_move)
12741 (aarch64_simd_make_constant): Update call to
12742 aarch64_simd_valid_immediate.
12743 (aarch64_advsimd_valid_immediate_hs): New function.
12744 (aarch64_advsimd_valid_immediate): Likewise.
12745 (aarch64_simd_valid_immediate): Remove mode and inverse
12746 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12747 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12748 and aarch64_float_const_representable_p on the result.
12749 (aarch64_output_simd_mov_immediate): Remove mode argument.
12750 Update call to aarch64_simd_valid_immediate and use of
12751 simd_immediate_info.
12752 (aarch64_output_scalar_simd_mov_immediate): Update call
12755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12756 Alan Hayward <alan.hayward@arm.com>
12757 David Sherwood <david.sherwood@arm.com>
12759 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12760 (mode_nunits): Likewise CONST_MODE_NUNITS.
12761 * machmode.def (ADJUST_NUNITS): Document.
12762 * genmodes.c (mode_data::need_nunits_adj): New field.
12763 (blank_mode): Update accordingly.
12764 (adj_nunits): New variable.
12765 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12767 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12768 listed in adj_nunits.
12769 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12770 listed in adj_nunits. Don't emit case statements for such modes.
12771 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12772 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12773 nothing if adj_nunits is nonnull.
12774 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12775 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12776 (emit_mode_fbit): Update use of print_maybe_const_decl.
12777 (emit_move_size): Likewise. Treat the array as non-const
12779 (emit_mode_adjustments): Handle adj_nunits.
12781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12783 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12784 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12785 (VECTOR_MODES): Use it.
12786 (make_vector_modes): Take the prefix as an argument.
12788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12789 Alan Hayward <alan.hayward@arm.com>
12790 David Sherwood <david.sherwood@arm.com>
12792 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12793 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12794 for MODE_VECTOR_BOOL.
12795 * machmode.def (VECTOR_BOOL_MODE): Document.
12796 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12797 (make_vector_bool_mode): New function.
12798 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12800 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12801 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12803 * stor-layout.c (int_mode_for_mode): Likewise.
12804 * tree.c (build_vector_type_for_mode): Likewise.
12805 * varasm.c (output_constant_pool_2): Likewise.
12806 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12807 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12808 for MODE_VECTOR_BOOL.
12809 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12810 of mode class checks.
12811 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12812 instead of a list of mode class checks.
12813 (expand_vector_scalar_condition): Likewise.
12814 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12817 Alan Hayward <alan.hayward@arm.com>
12818 David Sherwood <david.sherwood@arm.com>
12820 * machmode.h (mode_size): Change from unsigned short to
12822 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12823 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12824 or if measurement_type is not polynomial.
12825 (fixed_size_mode::includes_p): Check for constant-sized modes.
12826 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12827 return a poly_uint16 rather than an unsigned short.
12828 (emit_mode_size): Change the type of mode_size from unsigned short
12829 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12830 (emit_mode_adjustments): Cope with polynomial vector sizes.
12831 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12833 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12835 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12836 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12837 * caller-save.c (setup_save_areas): Likewise.
12838 (replace_reg_with_saved_mem): Likewise.
12839 * calls.c (emit_library_call_value_1): Likewise.
12840 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12841 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12842 (gen_lowpart_for_combine): Likewise.
12843 * convert.c (convert_to_integer_1): Likewise.
12844 * cse.c (equiv_constant, cse_insn): Likewise.
12845 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12846 (cselib_subst_to_values): Likewise.
12847 * dce.c (word_dce_process_block): Likewise.
12848 * df-problems.c (df_word_lr_mark_ref): Likewise.
12849 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12850 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12851 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12852 (rtl_for_decl_location): Likewise.
12853 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12854 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12855 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12856 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12857 (expand_expr_real_1): Likewise.
12858 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12859 (pad_below): Likewise.
12860 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12861 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12862 * ira.c (get_subreg_tracking_sizes): Likewise.
12863 * ira-build.c (ira_create_allocno_objects): Likewise.
12864 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12865 (ira_sort_regnos_for_alter_reg): Likewise.
12866 * ira-costs.c (record_operand_costs): Likewise.
12867 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12868 (resolve_simple_move): Likewise.
12869 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12870 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12871 (lra_constraints): Likewise.
12872 (CONST_POOL_OK_P): Reject variable-sized modes.
12873 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12874 (add_pseudo_to_slot, lra_spill): Likewise.
12875 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12876 * optabs-query.c (get_best_extraction_insn): Likewise.
12877 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12878 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12879 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12880 * recog.c (offsettable_address_addr_space_p): Likewise.
12881 * regcprop.c (maybe_mode_change): Likewise.
12882 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12883 * regrename.c (build_def_use): Likewise.
12884 * regstat.c (dump_reg_info): Likewise.
12885 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12886 (find_reloads, find_reloads_subreg_address): Likewise.
12887 * reload1.c (eliminate_regs_1): Likewise.
12888 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12889 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12890 (simplify_binary_operation_1, simplify_subreg): Likewise.
12891 * targhooks.c (default_function_arg_padding): Likewise.
12892 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12893 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12894 (verify_gimple_assign_ternary): Likewise.
12895 * tree-inline.c (estimate_move_cost): Likewise.
12896 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12897 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12898 (get_address_cost_ainc): Likewise.
12899 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12900 (vect_supportable_dr_alignment): Likewise.
12901 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12902 (vectorizable_reduction): Likewise.
12903 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12904 (vectorizable_operation, vectorizable_load): Likewise.
12905 * tree.c (build_same_sized_truth_vector_type): Likewise.
12906 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12907 * var-tracking.c (emit_note_insn_var_location): Likewise.
12908 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12909 (ADDR_VEC_ALIGN): Likewise.
12911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12912 Alan Hayward <alan.hayward@arm.com>
12913 David Sherwood <david.sherwood@arm.com>
12915 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12917 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12918 or if measurement_type is polynomial.
12919 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12920 * combine.c (make_extraction): Likewise.
12921 * dse.c (find_shift_sequence): Likewise.
12922 * dwarf2out.c (mem_loc_descriptor): Likewise.
12923 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12924 (extract_bit_field, extract_low_bits): Likewise.
12925 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12926 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12927 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12928 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12929 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12930 * reload.c (find_reloads): Likewise.
12931 * reload1.c (alter_reg): Likewise.
12932 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12933 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12934 * tree-if-conv.c (predicate_mem_writes): Likewise.
12935 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12936 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12937 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12938 * valtrack.c (dead_debug_insert_temp): Likewise.
12939 * varasm.c (mergeable_constant_section): Likewise.
12940 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12943 Alan Hayward <alan.hayward@arm.com>
12944 David Sherwood <david.sherwood@arm.com>
12946 * expr.c (expand_assignment): Cope with polynomial mode sizes
12947 when assigning to a CONCAT.
12949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12950 Alan Hayward <alan.hayward@arm.com>
12951 David Sherwood <david.sherwood@arm.com>
12953 * machmode.h (mode_precision): Change from unsigned short to
12955 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12957 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12958 or if measurement_type is not polynomial.
12959 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12960 in which the mode is already known to be a scalar_int_mode.
12961 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12962 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12964 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12965 for GET_MODE_PRECISION.
12966 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12967 for GET_MODE_PRECISION.
12968 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12970 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12971 (expand_field_assignment, make_extraction): Likewise.
12972 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12973 (get_last_value): Likewise.
12974 * convert.c (convert_to_integer_1): Likewise.
12975 * cse.c (cse_insn): Likewise.
12976 * expr.c (expand_expr_real_1): Likewise.
12977 * lra-constraints.c (simplify_operand_subreg): Likewise.
12978 * optabs-query.c (can_atomic_load_p): Likewise.
12979 * optabs.c (expand_atomic_load): Likewise.
12980 (expand_atomic_store): Likewise.
12981 * ree.c (combine_reaching_defs): Likewise.
12982 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12983 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12984 * tree.h (type_has_mode_precision_p): Likewise.
12985 * ubsan.c (instrument_si_overflow): Likewise.
12987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12988 Alan Hayward <alan.hayward@arm.com>
12989 David Sherwood <david.sherwood@arm.com>
12991 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12992 polynomial numbers of units.
12993 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12994 (valid_vector_subparts_p): New function.
12995 (build_vector_type): Remove temporary shim and take the number
12996 of units as a poly_uint64 rather than an int.
12997 (build_opaque_vector_type): Take the number of units as a
12998 poly_uint64 rather than an int.
12999 * tree.c (build_vector_from_ctor): Handle polynomial
13000 TYPE_VECTOR_SUBPARTS.
13001 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13002 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13003 (build_vector_from_val): If the number of units is variable,
13004 use build_vec_duplicate_cst for constant operands and
13005 VEC_DUPLICATE_EXPR otherwise.
13006 (make_vector_type): Remove temporary is_constant ().
13007 (build_vector_type, build_opaque_vector_type): Take the number of
13008 units as a poly_uint64 rather than an int.
13009 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13011 * cfgexpand.c (expand_debug_expr): Likewise.
13012 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13013 (store_constructor, expand_expr_real_1): Likewise.
13014 (const_scalar_mask_from_tree): Likewise.
13015 * fold-const-call.c (fold_const_reduction): Likewise.
13016 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13017 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13018 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13019 (fold_relational_const): Likewise.
13020 (native_interpret_vector): Likewise. Change the size from an
13021 int to an unsigned int.
13022 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13023 TYPE_VECTOR_SUBPARTS.
13024 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13025 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13026 duplicating a non-constant operand into a variable-length vector.
13027 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13028 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13029 * ipa-icf.c (sem_variable::equals): Likewise.
13030 * match.pd: Likewise.
13031 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13032 * print-tree.c (print_node): Likewise.
13033 * stor-layout.c (layout_type): Likewise.
13034 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13035 * tree-cfg.c (verify_gimple_comparison): Likewise.
13036 (verify_gimple_assign_binary): Likewise.
13037 (verify_gimple_assign_ternary): Likewise.
13038 (verify_gimple_assign_single): Likewise.
13039 * tree-pretty-print.c (dump_generic_node): Likewise.
13040 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13041 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13042 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13043 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13044 (vect_shift_permute_load_chain): Likewise.
13045 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13046 (expand_vector_condition, optimize_vector_constructor): Likewise.
13047 (lower_vec_perm, get_compute_type): Likewise.
13048 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13049 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13050 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13051 (vect_recog_mask_conversion_pattern): Likewise.
13052 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13053 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13054 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13055 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13056 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13057 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13058 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13059 (supportable_widening_operation): Likewise.
13060 (supportable_narrowing_operation): Likewise.
13061 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13063 * varasm.c (output_constant): Likewise.
13065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13066 Alan Hayward <alan.hayward@arm.com>
13067 David Sherwood <david.sherwood@arm.com>
13069 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13070 so that both the length == 3 and length != 3 cases set up their
13071 own permute vectors. Add comments explaining why we know the
13072 number of elements is constant.
13073 (vect_permute_load_chain): Likewise.
13075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13076 Alan Hayward <alan.hayward@arm.com>
13077 David Sherwood <david.sherwood@arm.com>
13079 * machmode.h (mode_nunits): Change from unsigned char to
13081 (ONLY_FIXED_SIZE_MODES): New macro.
13082 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13083 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13084 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13086 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13087 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13088 or if measurement_type is not polynomial.
13089 * genmodes.c (ZERO_COEFFS): New macro.
13090 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13092 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13093 Use ZERO_COEFFS when emitting initializers.
13094 * data-streamer.h (bp_pack_poly_value): New function.
13095 (bp_unpack_poly_value): Likewise.
13096 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13097 for GET_MODE_NUNITS.
13098 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13099 for GET_MODE_NUNITS.
13100 * tree.c (make_vector_type): Remove temporary shim and make
13101 the real function take the number of units as a poly_uint64
13102 rather than an int.
13103 (build_vector_type_for_mode): Handle polynomial nunits.
13104 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13105 * emit-rtl.c (const_vec_series_p_1): Likewise.
13106 (gen_rtx_CONST_VECTOR): Likewise.
13107 * fold-const.c (test_vec_duplicate_folding): Likewise.
13108 * genrecog.c (validate_pattern): Likewise.
13109 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13110 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13111 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13112 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13113 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13114 * rtlanal.c (subreg_get_info): Likewise.
13115 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13116 (vect_grouped_load_supported): Likewise.
13117 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13118 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13119 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13120 (simplify_const_unary_operation, simplify_binary_operation_1)
13121 (simplify_const_binary_operation, simplify_ternary_operation)
13122 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13123 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13124 instead of CONST_VECTOR_NUNITS.
13125 * varasm.c (output_constant_pool_2): Likewise.
13126 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13127 explicit-encoded elements in the XVEC for variable-length vectors.
13129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13131 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13134 Alan Hayward <alan.hayward@arm.com>
13135 David Sherwood <david.sherwood@arm.com>
13137 * coretypes.h (fixed_size_mode): Declare.
13138 (fixed_size_mode_pod): New typedef.
13139 * builtins.h (target_builtins::x_apply_args_mode)
13140 (target_builtins::x_apply_result_mode): Change type to
13141 fixed_size_mode_pod.
13142 * builtins.c (apply_args_size, apply_result_size, result_vector)
13143 (expand_builtin_apply_args_1, expand_builtin_apply)
13144 (expand_builtin_return): Update accordingly.
13146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13148 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13149 * cselib.c (cselib_hash_rtx): Likewise.
13150 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13151 CONST_VECTOR encoding.
13153 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13154 Jeff Law <law@redhat.com>
13157 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13158 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13159 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13160 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13163 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13164 explicitly probe *sp in a noreturn function if there were any callee
13165 register saves or frame pointer is needed.
13167 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13170 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13171 BLKmode for ternary, binary or unary expressions.
13174 * var-tracking.c (delete_vta_debug_insn): New inline function.
13175 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13176 insns from get_insns () to NULL instead of each bb separately.
13177 Use delete_vta_debug_insn. No longer static.
13178 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13179 delete_vta_debug_insns callers.
13180 * rtl.h (delete_vta_debug_insns): Declare.
13181 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13182 instead of variable_tracking_main.
13184 2018-01-03 Martin Sebor <msebor@redhat.com>
13186 PR tree-optimization/83603
13187 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13188 arguments past the endof the argument list in functions declared
13189 without a prototype.
13190 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13191 Avoid checking when arguments are null.
13193 2018-01-03 Martin Sebor <msebor@redhat.com>
13196 * doc/extend.texi (attribute const): Fix a typo.
13197 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13198 issuing -Wsuggest-attribute for void functions.
13200 2018-01-03 Martin Sebor <msebor@redhat.com>
13202 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13203 offset_int::from instead of wide_int::to_shwi.
13204 (maybe_diag_overlap): Remove assertion.
13205 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13206 * gimple-ssa-sprintf.c (format_directive): Same.
13207 (parse_directive): Same.
13208 (sprintf_dom_walker::compute_format_length): Same.
13209 (try_substitute_return_value): Same.
13211 2018-01-03 Jeff Law <law@redhat.com>
13213 PR middle-end/83654
13214 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13215 non-constant residual for zero at runtime and avoid probing in
13216 that case. Reorganize code for trailing problem to mirror handling
13219 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13221 PR tree-optimization/83501
13222 * tree-ssa-strlen.c (get_string_cst): New.
13223 (handle_char_store): Call get_string_cst.
13225 2018-01-03 Martin Liska <mliska@suse.cz>
13227 PR tree-optimization/83593
13228 * tree-ssa-strlen.c: Include tree-cfg.h.
13229 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13230 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13231 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13233 (strlen_dom_walker::before_dom_children): Call
13234 gimple_purge_dead_eh_edges. Dump tranformation with details
13236 (strlen_dom_walker::before_dom_children): Update call by adding
13237 new argument cleanup_eh.
13238 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13240 2018-01-03 Martin Liska <mliska@suse.cz>
13243 * cif-code.def (VARIADIC_THUNK): New enum value.
13244 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13247 2018-01-03 Jan Beulich <jbeulich@suse.com>
13249 * sse.md (mov<mode>_internal): Tighten condition for when to use
13250 vmovdqu<ssescalarsize> for TI and OI modes.
13252 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13254 Update copyright years.
13256 2018-01-03 Martin Liska <mliska@suse.cz>
13259 * ipa-visibility.c (function_and_variable_visibility): Skip
13260 functions with noipa attribure.
13262 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13264 * gcc.c (process_command): Update copyright notice dates.
13265 * gcov-dump.c (print_version): Ditto.
13266 * gcov.c (print_version): Ditto.
13267 * gcov-tool.c (print_version): Ditto.
13268 * gengtype.c (create_file): Ditto.
13269 * doc/cpp.texi: Bump @copying's copyright year.
13270 * doc/cppinternals.texi: Ditto.
13271 * doc/gcc.texi: Ditto.
13272 * doc/gccint.texi: Ditto.
13273 * doc/gcov.texi: Ditto.
13274 * doc/install.texi: Ditto.
13275 * doc/invoke.texi: Ditto.
13277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13279 * vector-builder.h (vector_builder::m_full_nelts): Change from
13280 unsigned int to poly_uint64.
13281 (vector_builder::full_nelts): Update prototype accordingly.
13282 (vector_builder::new_vector): Likewise.
13283 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13284 (vector_builder::operator ==): Likewise.
13285 (vector_builder::finalize): Likewise.
13286 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13287 Take the number of elements as a poly_uint64 rather than an
13289 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13290 from unsigned int to poly_uint64.
13291 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13292 (vec_perm_indices::new_vector): Likewise.
13293 (vec_perm_indices::length): Likewise.
13294 (vec_perm_indices::nelts_per_input): Likewise.
13295 (vec_perm_indices::input_nelts): Likewise.
13296 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13297 number of elements per input as a poly_uint64 rather than an
13298 unsigned int. Use the original encoding for variable-length
13299 vectors, rather than clamping each individual element.
13300 For the second and subsequent elements in each pattern,
13301 clamp the step and base before clamping their sum.
13302 (vec_perm_indices::series_p): Handle polynomial element counts.
13303 (vec_perm_indices::all_in_range_p): Likewise.
13304 (vec_perm_indices_to_tree): Likewise.
13305 (vec_perm_indices_to_rtx): Likewise.
13306 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13307 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13308 (tree_vector_builder::new_binary_operation): Handle polynomial
13309 element counts. Return false if we need to know the number
13310 of elements at compile time.
13311 * fold-const.c (fold_vec_perm): Punt if the number of elements
13312 isn't known at compile time.
13314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13316 * vec-perm-indices.h (vec_perm_builder): Change element type
13317 from HOST_WIDE_INT to poly_int64.
13318 (vec_perm_indices::element_type): Update accordingly.
13319 (vec_perm_indices::clamp): Handle polynomial element_types.
13320 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13321 (vec_perm_indices::all_in_range_p): Likewise.
13322 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13324 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13325 polynomial vec_perm_indices element types.
13326 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13327 * fold-const.c (fold_vec_perm): Likewise.
13328 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13329 * tree-vect-generic.c (lower_vec_perm): Likewise.
13330 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13331 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13332 element type to HOST_WIDE_INT.
13334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13335 Alan Hayward <alan.hayward@arm.com>
13336 David Sherwood <david.sherwood@arm.com>
13338 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13339 rather than an int. Use plus_constant.
13340 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13341 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13344 Alan Hayward <alan.hayward@arm.com>
13345 David Sherwood <david.sherwood@arm.com>
13347 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13348 a HOST_WIDE_INT to a poly_int64.
13350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13351 Alan Hayward <alan.hayward@arm.com>
13352 David Sherwood <david.sherwood@arm.com>
13354 * calls.c (load_register_parameters): Cope with polynomial
13355 mode sizes. Require a constant size for BLKmode parameters
13356 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13357 forces a parameter to be padded at the lsb end in order to
13358 fill a complete number of words, require the parameter size
13359 to be ordered wrt UNITS_PER_WORD.
13361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13362 Alan Hayward <alan.hayward@arm.com>
13363 David Sherwood <david.sherwood@arm.com>
13365 * reload1.c (spill_stack_slot_width): Change element type
13366 from unsigned int to poly_uint64_pod.
13367 (alter_reg): Treat mode sizes as polynomial.
13369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13370 Alan Hayward <alan.hayward@arm.com>
13371 David Sherwood <david.sherwood@arm.com>
13373 * reload.c (complex_word_subreg_p): New function.
13374 (reload_inner_reg_of_subreg, push_reload): Use it.
13376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13377 Alan Hayward <alan.hayward@arm.com>
13378 David Sherwood <david.sherwood@arm.com>
13380 * lra-constraints.c (process_alt_operands): Reject matched
13381 operands whose sizes aren't ordered.
13382 (match_reload): Refer to this check here.
13384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13385 Alan Hayward <alan.hayward@arm.com>
13386 David Sherwood <david.sherwood@arm.com>
13388 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13389 that the mode size is in the set {1, 2, 4, 8, 16}.
13391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13392 Alan Hayward <alan.hayward@arm.com>
13393 David Sherwood <david.sherwood@arm.com>
13395 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13396 Use plus_constant instead of gen_rtx_PLUS.
13398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13399 Alan Hayward <alan.hayward@arm.com>
13400 David Sherwood <david.sherwood@arm.com>
13402 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13403 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13404 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13405 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13406 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13407 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13408 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13409 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13410 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13411 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13413 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13414 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13415 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13416 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13417 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13418 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13419 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13420 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13421 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13422 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13424 * expr.c (emit_move_resolve_push): Treat the input and result
13425 of PUSH_ROUNDING as a poly_int64.
13426 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13427 (emit_push_insn): Likewise.
13428 * lra-eliminations.c (mark_not_eliminable): Likewise.
13429 * recog.c (push_operand): Likewise.
13430 * reload1.c (elimination_effects): Likewise.
13431 * rtlanal.c (nonzero_bits1): Likewise.
13432 * calls.c (store_one_arg): Likewise. Require the padding to be
13433 known at compile time.
13435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13436 Alan Hayward <alan.hayward@arm.com>
13437 David Sherwood <david.sherwood@arm.com>
13439 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13440 Use plus_constant instead of gen_rtx_PLUS.
13442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13443 Alan Hayward <alan.hayward@arm.com>
13444 David Sherwood <david.sherwood@arm.com>
13446 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13447 rather than an int.
13449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13450 Alan Hayward <alan.hayward@arm.com>
13451 David Sherwood <david.sherwood@arm.com>
13453 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13454 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13455 via stack temporaries. Treat the mode size as polynomial too.
13457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13458 Alan Hayward <alan.hayward@arm.com>
13459 David Sherwood <david.sherwood@arm.com>
13461 * expr.c (expand_expr_real_2): When handling conversions involving
13462 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13463 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13464 as a poly_uint64 too.
13466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13467 Alan Hayward <alan.hayward@arm.com>
13468 David Sherwood <david.sherwood@arm.com>
13470 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13473 Alan Hayward <alan.hayward@arm.com>
13474 David Sherwood <david.sherwood@arm.com>
13476 * combine.c (can_change_dest_mode): Handle polynomial
13477 REGMODE_NATURAL_SIZE.
13478 * expmed.c (store_bit_field_1): Likewise.
13479 * expr.c (store_constructor): Likewise.
13480 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13481 and polynomial REGMODE_NATURAL_SIZE.
13482 (gen_lowpart_common): Likewise.
13483 * reginfo.c (record_subregs_of_mode): Likewise.
13484 * rtlanal.c (read_modify_subreg_p): Likewise.
13486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13487 Alan Hayward <alan.hayward@arm.com>
13488 David Sherwood <david.sherwood@arm.com>
13490 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13491 numbers of elements.
13493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13494 Alan Hayward <alan.hayward@arm.com>
13495 David Sherwood <david.sherwood@arm.com>
13497 * match.pd: Cope with polynomial numbers of vector elements.
13499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13500 Alan Hayward <alan.hayward@arm.com>
13501 David Sherwood <david.sherwood@arm.com>
13503 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13504 in a POINTER_PLUS_EXPR.
13506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13507 Alan Hayward <alan.hayward@arm.com>
13508 David Sherwood <david.sherwood@arm.com>
13510 * omp-simd-clone.c (simd_clone_subparts): New function.
13511 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13512 (ipa_simd_modify_function_body): Likewise.
13514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13515 Alan Hayward <alan.hayward@arm.com>
13516 David Sherwood <david.sherwood@arm.com>
13518 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13519 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13520 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13521 (expand_vector_condition, vector_element): Likewise.
13522 (subparts_gt): New function.
13523 (get_compute_type): Use subparts_gt.
13524 (count_type_subparts): Delete.
13525 (expand_vector_operations_1): Use subparts_gt instead of
13526 count_type_subparts.
13528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13529 Alan Hayward <alan.hayward@arm.com>
13530 David Sherwood <david.sherwood@arm.com>
13532 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13533 (vect_compile_time_alias): ...this new function. Do the calculation
13534 on poly_ints rather than trees.
13535 (vect_prune_runtime_alias_test_list): Update call accordingly.
13537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13538 Alan Hayward <alan.hayward@arm.com>
13539 David Sherwood <david.sherwood@arm.com>
13541 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13543 (vect_schedule_slp_instance): Likewise.
13545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13546 Alan Hayward <alan.hayward@arm.com>
13547 David Sherwood <david.sherwood@arm.com>
13549 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13550 constant and extern definitions for variable-length vectors.
13551 (vect_get_constant_vectors): Note that the number of units
13552 is known to be constant.
13554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13555 Alan Hayward <alan.hayward@arm.com>
13556 David Sherwood <david.sherwood@arm.com>
13558 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13559 of units as polynomial. Choose between WIDE and NARROW based
13562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13563 Alan Hayward <alan.hayward@arm.com>
13564 David Sherwood <david.sherwood@arm.com>
13566 * tree-vect-stmts.c (simd_clone_subparts): New function.
13567 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13570 Alan Hayward <alan.hayward@arm.com>
13571 David Sherwood <david.sherwood@arm.com>
13573 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13574 vectors as polynomial. Use build_index_vector for
13575 IFN_GOMP_SIMD_LANE.
13577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13578 Alan Hayward <alan.hayward@arm.com>
13579 David Sherwood <david.sherwood@arm.com>
13581 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13582 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13583 for variable-length vectors.
13584 (vectorizable_mask_load_store): Treat the number of units as
13585 polynomial, asserting that it is constant if the condition has
13586 already been enforced.
13587 (vectorizable_store, vectorizable_load): Likewise.
13589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13590 Alan Hayward <alan.hayward@arm.com>
13591 David Sherwood <david.sherwood@arm.com>
13593 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13594 of units as polynomial. Punt if we can't tell at compile time
13595 which vector contains the final result.
13597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13598 Alan Hayward <alan.hayward@arm.com>
13599 David Sherwood <david.sherwood@arm.com>
13601 * tree-vect-loop.c (vectorizable_induction): Treat the number
13602 of units as polynomial. Punt on SLP inductions. Use an integer
13603 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13604 cast of such a series for variable-length floating-point
13607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13608 Alan Hayward <alan.hayward@arm.com>
13609 David Sherwood <david.sherwood@arm.com>
13611 * tree.h (build_index_vector): Declare.
13612 * tree.c (build_index_vector): New function.
13613 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13614 of units as polynomial, forcibly converting it to a constant if
13615 vectorizable_reduction has already enforced the condition.
13616 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13617 to create a {1,2,3,...} vector.
13618 (vectorizable_reduction): Treat the number of units as polynomial.
13619 Choose vectype_in based on the largest scalar element size rather
13620 than the smallest number of units. Enforce the restrictions
13623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13624 Alan Hayward <alan.hayward@arm.com>
13625 David Sherwood <david.sherwood@arm.com>
13627 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13628 number of units as polynomial.
13630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13631 Alan Hayward <alan.hayward@arm.com>
13632 David Sherwood <david.sherwood@arm.com>
13634 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13635 * target.def (autovectorize_vector_sizes): Return the vector sizes
13636 by pointer, using vector_sizes rather than a bitmask.
13637 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13638 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13639 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13641 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13642 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13643 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13644 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13645 * omp-general.c (omp_max_vf): Likewise.
13646 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13647 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13648 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13649 * tree-vect-slp.c (vect_slp_bb): Likewise.
13650 * doc/tm.texi: Regenerate.
13651 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13653 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13654 the vector size as a poly_uint64 rather than an unsigned int.
13655 (current_vector_size): Change from an unsigned int to a poly_uint64.
13656 (get_vectype_for_scalar_type): Update accordingly.
13657 * tree.h (build_truth_vector_type): Take the size and number of
13658 units as a poly_uint64 rather than an unsigned int.
13659 (build_vector_type): Add a temporary overload that takes
13660 the number of units as a poly_uint64 rather than an unsigned int.
13661 * tree.c (make_vector_type): Likewise.
13662 (build_truth_vector_type): Take the number of units as a poly_uint64
13663 rather than an unsigned int.
13665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13666 Alan Hayward <alan.hayward@arm.com>
13667 David Sherwood <david.sherwood@arm.com>
13669 * target.def (get_mask_mode): Take the number of units and length
13670 as poly_uint64s rather than unsigned ints.
13671 * targhooks.h (default_get_mask_mode): Update accordingly.
13672 * targhooks.c (default_get_mask_mode): Likewise.
13673 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13674 * doc/tm.texi: Regenerate.
13676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13677 Alan Hayward <alan.hayward@arm.com>
13678 David Sherwood <david.sherwood@arm.com>
13680 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13681 * omp-general.c (omp_max_vf): Likewise.
13682 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13683 (expand_omp_simd): Handle polynomial safelen.
13684 * omp-low.c (omplow_simd_context): Add a default constructor.
13685 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13686 (lower_rec_simd_input_clauses): Update accordingly.
13687 (lower_rec_input_clauses): Likewise.
13689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13690 Alan Hayward <alan.hayward@arm.com>
13691 David Sherwood <david.sherwood@arm.com>
13693 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13694 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13695 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13696 (vect_analyze_slp_cost): Likewise.
13697 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13698 (vect_model_load_cost): Likewise.
13700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13701 Alan Hayward <alan.hayward@arm.com>
13702 David Sherwood <david.sherwood@arm.com>
13704 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13705 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13706 from an unsigned int * to a poly_uint64_pod *.
13707 (calculate_unrolling_factor): New function.
13708 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13711 Alan Hayward <alan.hayward@arm.com>
13712 David Sherwood <david.sherwood@arm.com>
13714 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13715 from an unsigned int to a poly_uint64.
13716 (_loop_vec_info::slp_unrolling_factor): Likewise.
13717 (_loop_vec_info::vectorization_factor): Change from an int
13719 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13720 (vect_get_num_vectors): New function.
13721 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13722 (vect_get_num_copies): Use vect_get_num_vectors.
13723 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13724 to an unsigned int *.
13725 (vect_analyze_data_refs): Change min_vf from an int * to a
13727 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13728 than an unsigned HOST_WIDE_INT.
13729 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13730 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13731 to an unsigned int *.
13732 (vect_analyze_data_ref_dependences): Likewise.
13733 (vect_compute_data_ref_alignment): Handle polynomial vf.
13734 (vect_enhance_data_refs_alignment): Likewise.
13735 (vect_prune_runtime_alias_test_list): Likewise.
13736 (vect_shift_permute_load_chain): Likewise.
13737 (vect_supportable_dr_alignment): Likewise.
13738 (dependence_distance_ge_vf): Take the vectorization factor as a
13739 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13740 (vect_analyze_data_refs): Change min_vf from an int * to a
13742 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13743 vfm1 as a poly_uint64 rather than an int. Make the same change
13744 for the returned bound_scalar.
13745 (vect_gen_vector_loop_niters): Handle polynomial vf.
13746 (vect_do_peeling): Likewise. Update call to
13747 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13748 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13750 * tree-vect-loop.c (vect_determine_vectorization_factor)
13751 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13752 (vect_get_known_peeling_cost): Likewise.
13753 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13754 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13755 (vect_transform_loop): Likewise. Use the lowest possible VF when
13756 updating the upper bounds of the loop.
13757 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13758 rather than an int.
13759 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13760 polynomial unroll factors.
13761 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13762 (vect_make_slp_decision): Likewise.
13763 (vect_supported_load_permutation_p): Likewise, and polynomial
13765 (vect_analyze_slp_cost): Handle polynomial vf.
13766 (vect_slp_analyze_node_operations): Likewise.
13767 (vect_slp_analyze_bb_1): Likewise.
13768 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13769 than an unsigned HOST_WIDE_INT.
13770 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13771 (vectorizable_load): Handle polynomial vf.
13772 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13774 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13777 Alan Hayward <alan.hayward@arm.com>
13778 David Sherwood <david.sherwood@arm.com>
13780 * match.pd: Handle bit operations involving three constants
13781 and try to fold one pair.
13783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13785 * tree-vect-loop-manip.c: Include gimple-fold.h.
13786 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13787 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13788 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13789 Add a path that uses a step of VF instead of 1, but disable it
13791 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13792 and niters_no_overflow parameters. Update calls to
13793 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13794 Create a new SSA name if the latter choses to use a ste other
13795 than zero, and return it via niters_vector_mult_vf_var.
13796 * tree-vect-loop.c (vect_transform_loop): Update calls to
13797 vect_do_peeling, vect_gen_vector_loop_niters and
13798 slpeel_make_loop_iterate_ntimes.
13799 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13800 (vect_gen_vector_loop_niters): Update declarations after above changes.
13802 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13804 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13805 128-bit round to integer instructions.
13806 (ceil<mode>2): Likewise.
13807 (btrunc<mode>2): Likewise.
13808 (round<mode>2): Likewise.
13810 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13812 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13813 unaligned VSX load/store on P8/P9.
13814 (expand_block_clear): Allow the use of unaligned VSX
13815 load/store on P8/P9.
13817 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13819 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13821 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13822 swap associated with both a load and a store.
13824 2018-01-02 Andrew Waterman <andrew@sifive.com>
13826 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13827 * config/riscv/riscv.md (clear_cache): Use it.
13829 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13831 * web.c: Remove out-of-date comment.
13833 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13835 * expr.c (fixup_args_size_notes): Check that any existing
13836 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13837 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13838 (emit_single_push_insn): ...here.
13840 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13842 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13843 (const_vector_encoded_nelts): New function.
13844 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13845 (const_vector_int_elt, const_vector_elt): Declare.
13846 * emit-rtl.c (const_vector_int_elt_1): New function.
13847 (const_vector_elt): Likewise.
13848 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13849 of CONST_VECTOR_ELT.
13851 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13853 * expr.c: Include rtx-vector-builder.h.
13854 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13855 directly on the tree encoding.
13856 (const_vector_from_tree): Likewise.
13857 * optabs.c: Include rtx-vector-builder.h.
13858 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13859 sequence of "u" values.
13860 * vec-perm-indices.c: Include rtx-vector-builder.h.
13861 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13862 directly on the vec_perm_indices encoding.
13864 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13866 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13867 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13868 * rtx-vector-builder.h: New file.
13869 * rtx-vector-builder.c: Likewise.
13870 * rtl.h (rtx_def::u2): Add a const_vector field.
13871 (CONST_VECTOR_NPATTERNS): New macro.
13872 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13873 (CONST_VECTOR_DUPLICATE_P): Likewise.
13874 (CONST_VECTOR_STEPPED_P): Likewise.
13875 (CONST_VECTOR_ENCODED_ELT): Likewise.
13876 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13877 (unwrap_const_vec_duplicate): Likewise.
13878 (const_vec_series_p): Check for a non-duplicated vector encoding.
13879 Say that the function only returns true for integer vectors.
13880 * emit-rtl.c: Include rtx-vector-builder.h.
13881 (gen_const_vec_duplicate_1): Delete.
13882 (gen_const_vector): Call gen_const_vec_duplicate instead of
13883 gen_const_vec_duplicate_1.
13884 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13885 (gen_const_vec_duplicate): Use rtx_vector_builder.
13886 (gen_const_vec_series): Likewise.
13887 (gen_rtx_CONST_VECTOR): Likewise.
13888 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13889 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13890 Build a new vector rather than modifying a CONST_VECTOR in-place.
13891 (handle_special_swappables): Update call accordingly.
13892 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13893 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13894 Build a new vector rather than modifying a CONST_VECTOR in-place.
13895 (handle_special_swappables): Update call accordingly.
13897 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13899 * simplify-rtx.c (simplify_const_binary_operation): Use
13900 CONST_VECTOR_ELT instead of XVECEXP.
13902 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13904 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13905 the selector elements to be different from the data elements
13906 if the selector is a VECTOR_CST.
13907 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13908 ssizetype for the selector.
13910 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13912 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13913 before testing each element individually.
13914 * tree-vect-generic.c (lower_vec_perm): Likewise.
13916 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13918 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13919 * selftest-run-tests.c (selftest::run_tests): Call it.
13920 * vector-builder.h (vector_builder::operator ==): New function.
13921 (vector_builder::operator !=): Likewise.
13922 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13923 (vec_perm_indices::all_from_input_p): New function.
13924 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13925 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13926 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13927 instead of reading the VECTOR_CST directly. Detect whether both
13928 vector inputs are the same before constructing the vec_perm_indices,
13929 and update the number of inputs argument accordingly. Use the
13930 utility functions added above. Only construct sel2 if we need to.
13932 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13934 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13935 the broadcast of the low byte.
13936 (expand_mult_highpart): Use an explicit encoding for the permutes.
13937 * optabs-query.c (can_mult_highpart_p): Likewise.
13938 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13939 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13940 (vectorizable_bswap): Likewise.
13941 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13942 explicit encoding for the power-of-2 permutes.
13943 (vect_permute_store_chain): Likewise.
13944 (vect_grouped_load_supported): Likewise.
13945 (vect_permute_load_chain): Likewise.
13947 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13949 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13950 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13951 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13952 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13953 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13954 (vect_gen_perm_mask_any): Likewise.
13956 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13958 * int-vector-builder.h: New file.
13959 * vec-perm-indices.h: Include int-vector-builder.h.
13960 (vec_perm_indices): Redefine as an int_vector_builder.
13961 (auto_vec_perm_indices): Delete.
13962 (vec_perm_builder): Redefine as a stand-alone class.
13963 (vec_perm_indices::vec_perm_indices): New function.
13964 (vec_perm_indices::clamp): Likewise.
13965 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13966 (vec_perm_indices::new_vector): New function.
13967 (vec_perm_indices::new_expanded_vector): Update for new
13968 vec_perm_indices class.
13969 (vec_perm_indices::rotate_inputs): New function.
13970 (vec_perm_indices::all_in_range_p): Operate directly on the
13971 encoded form, without computing elided elements.
13972 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13973 encoding. Update for new vec_perm_indices class.
13974 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13975 the given vec_perm_builder.
13976 (expand_vec_perm_var): Update vec_perm_builder constructor.
13977 (expand_mult_highpart): Use vec_perm_builder instead of
13978 auto_vec_perm_indices.
13979 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13980 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13981 or double series encoding as appropriate.
13982 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13983 vec_perm_indices instead of auto_vec_perm_indices.
13984 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13985 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13986 (vect_permute_store_chain): Likewise.
13987 (vect_grouped_load_supported): Likewise.
13988 (vect_permute_load_chain): Likewise.
13989 (vect_shift_permute_load_chain): Likewise.
13990 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13991 (vect_transform_slp_perm_load): Likewise.
13992 (vect_schedule_slp_instance): Likewise.
13993 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13994 (vectorizable_mask_load_store): Likewise.
13995 (vectorizable_bswap): Likewise.
13996 (vectorizable_store): Likewise.
13997 (vectorizable_load): Likewise.
13998 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13999 vec_perm_indices instead of auto_vec_perm_indices. Use
14000 tree_to_vec_perm_builder to read the vector from a tree.
14001 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14002 vec_perm_builder instead of a vec_perm_indices.
14003 (have_whole_vector_shift): Use vec_perm_builder and
14004 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14005 truncation to calc_vec_perm_mask_for_shift.
14006 (vect_create_epilog_for_reduction): Likewise.
14007 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14008 from auto_vec_perm_indices to vec_perm_indices.
14009 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14010 instead of changing individual elements.
14011 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14012 the vector in d.perm.
14013 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14014 from auto_vec_perm_indices to vec_perm_indices.
14015 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14016 instead of changing individual elements.
14017 (arm_vectorize_vec_perm_const): Use new_vector to install
14018 the vector in d.perm.
14019 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14020 Update vec_perm_builder constructor.
14021 (rs6000_expand_interleave): Likewise.
14022 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14023 (rs6000_expand_interleave): Likewise.
14025 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14027 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14028 to qimode could truncate the indices.
14029 * optabs.c (expand_vec_perm_var): Likewise.
14031 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14033 * Makefile.in (OBJS): Add vec-perm-indices.o.
14034 * vec-perm-indices.h: New file.
14035 * vec-perm-indices.c: Likewise.
14036 * target.h (vec_perm_indices): Replace with a forward class
14038 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14039 * optabs.h: Include vec-perm-indices.h.
14040 (expand_vec_perm): Delete.
14041 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14042 (expand_vec_perm_const): Declare.
14043 * target.def (vec_perm_const_ok): Replace with...
14044 (vec_perm_const): ...this new hook.
14045 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14046 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14047 * doc/tm.texi: Regenerate.
14048 * optabs.def (vec_perm_const): Delete.
14049 * doc/md.texi (vec_perm_const): Likewise.
14050 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14051 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14052 expand_vec_perm for constant permutation vectors. Assert that
14053 the mode of variable permutation vectors is the integer equivalent
14054 of the mode that is being permuted.
14055 * optabs-query.h (selector_fits_mode_p): Declare.
14056 * optabs-query.c: Include vec-perm-indices.h.
14057 (selector_fits_mode_p): New function.
14058 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14059 is defined, instead of checking whether the vec_perm_const_optab
14060 exists. Use targetm.vectorize.vec_perm_const instead of
14061 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14062 fit in the vector mode before using a variable permute.
14063 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14064 vec_perm_indices instead of an rtx.
14065 (expand_vec_perm): Replace with...
14066 (expand_vec_perm_const): ...this new function. Take the selector
14067 as a vec_perm_indices rather than an rtx. Also take the mode of
14068 the selector. Update call to shift_amt_for_vec_perm_mask.
14069 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14070 Use vec_perm_indices::new_expanded_vector to expand the original
14071 selector into bytes. Check whether the indices fit in the vector
14072 mode before using a variable permute.
14073 (expand_vec_perm_var): Make global.
14074 (expand_mult_highpart): Use expand_vec_perm_const.
14075 * fold-const.c: Includes vec-perm-indices.h.
14076 * tree-ssa-forwprop.c: Likewise.
14077 * tree-vect-data-refs.c: Likewise.
14078 * tree-vect-generic.c: Likewise.
14079 * tree-vect-loop.c: Likewise.
14080 * tree-vect-slp.c: Likewise.
14081 * tree-vect-stmts.c: Likewise.
14082 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14084 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14085 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14086 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14087 (aarch64_vectorize_vec_perm_const): ...this new function.
14088 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14089 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14090 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14091 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14092 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14093 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14094 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14096 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14097 check for NEON modes.
14098 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14099 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14100 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14101 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14103 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14104 the old VEC_PERM_CONST conditions.
14105 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14106 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14107 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14108 (ia64_vectorize_vec_perm_const_ok): Merge into...
14109 (ia64_vectorize_vec_perm_const): ...this new function.
14110 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14111 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14112 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14113 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14114 * config/mips/mips.c (mips_expand_vec_perm_const)
14115 (mips_vectorize_vec_perm_const_ok): Merge into...
14116 (mips_vectorize_vec_perm_const): ...this new function.
14117 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14118 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14119 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14120 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14121 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14122 (rs6000_expand_vec_perm_const): Delete.
14123 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14125 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14126 (altivec_expand_vec_perm_const_le): Take each operand individually.
14127 Operate on constant selectors rather than rtxes.
14128 (altivec_expand_vec_perm_const): Likewise. Update call to
14129 altivec_expand_vec_perm_const_le.
14130 (rs6000_expand_vec_perm_const): Delete.
14131 (rs6000_vectorize_vec_perm_const_ok): Delete.
14132 (rs6000_vectorize_vec_perm_const): New function.
14133 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14134 an element count and rtx array.
14135 (rs6000_expand_extract_even): Update call accordingly.
14136 (rs6000_expand_interleave): Likewise.
14137 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14138 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14139 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14140 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14141 (rs6000_expand_vec_perm_const): Delete.
14142 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14143 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14144 (altivec_expand_vec_perm_const_le): Take each operand individually.
14145 Operate on constant selectors rather than rtxes.
14146 (altivec_expand_vec_perm_const): Likewise. Update call to
14147 altivec_expand_vec_perm_const_le.
14148 (rs6000_expand_vec_perm_const): Delete.
14149 (rs6000_vectorize_vec_perm_const_ok): Delete.
14150 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14151 reference to the SPE evmerge intructions.
14152 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14153 an element count and rtx array.
14154 (rs6000_expand_extract_even): Update call accordingly.
14155 (rs6000_expand_interleave): Likewise.
14156 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14157 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14159 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14161 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14163 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14164 vector mode and that that mode matches the mode of the data
14166 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14167 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14168 directly using expand_vec_perm_1 when forcing selectors into
14170 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14172 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14174 * optabs-query.h (can_vec_perm_p): Delete.
14175 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14176 * optabs-query.c (can_vec_perm_p): Split into...
14177 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14178 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14179 particular selector is valid.
14180 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14181 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14182 (vect_grouped_load_supported): Likewise.
14183 (vect_shift_permute_load_chain): Likewise.
14184 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14185 (vect_transform_slp_perm_load): Likewise.
14186 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14187 (vectorizable_bswap): Likewise.
14188 (vect_gen_perm_mask_checked): Likewise.
14189 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14190 implementations of variable permutation vectors into account
14191 when deciding which selector to use.
14192 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14193 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14194 with a false third argument.
14195 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14196 to test whether the constant selector is valid and can_vec_perm_var_p
14197 to test whether a variable selector is valid.
14199 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14201 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14202 * optabs-query.c (can_vec_perm_p): Likewise.
14203 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14204 instead of vec_perm_indices.
14205 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14206 (vect_gen_perm_mask_checked): Likewise,
14207 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14208 (vect_gen_perm_mask_checked): Likewise,
14210 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14212 * optabs-query.h (qimode_for_vec_perm): Declare.
14213 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14214 (qimode_for_vec_perm): ...this new function.
14215 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14217 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14219 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14220 does not have a conditional at the top.
14222 2018-01-02 Richard Biener <rguenther@suse.de>
14224 * ipa-inline.c (big_speedup_p): Fix expression.
14226 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14229 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14232 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14236 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14237 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14238 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14239 cond_taken_branch_cost 3->4.
14241 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14243 PR tree-optimization/83581
14244 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14245 TODO_cleanup_cfg if any changes have been made.
14247 PR middle-end/83608
14248 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14249 convert_modes if target mode has the right side, but different mode
14252 PR middle-end/83609
14253 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14254 last argument when extracting from CONCAT. If either from_real or
14255 from_imag is NULL, use expansion through memory. If result is not
14256 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14257 the parts directly to inner mode, if even that fails, use expansion
14260 PR middle-end/83623
14261 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14262 check for bswap in mode rather than HImode and use that in expand_unop
14265 Copyright (C) 2018 Free Software Foundation, Inc.
14267 Copying and distribution of this file, with or without modification,
14268 are permitted in any medium without royalty provided the copyright
14269 notice and this notice are preserved.