1 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
4 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5 (parm_default_def_partition_arg): Ditto.
6 (set_parm_default_def_partition): Ditto.
7 (get_parm_default_def_partitions): Ditto and make it static.
8 (get_undefined_value_partitions): Ditto and make it static.
9 (remove_ssa_form): Refactor call to init_var_map here.
10 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
11 computation for loop region.
12 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
13 (register_default_def): Delete.
14 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
15 (parm_default_def_partition_arg): Ditto.
16 (set_parm_default_def_partition): Ditto.
17 (get_parm_default_def_partitions): Ditto and make it static.
18 (get_undefined_value_partitions): Ditto and make it static.
19 (coalesce_with_default, coalesce_with_default): Update comment.
20 (create_coalesce_list_for_region): New func factored out from
21 create_outofssa_var_map.
22 (populate_coalesce_list_for_outofssa): New func factored out from
23 create_outofssa_var_map and coalesce_ssa_name.
24 (create_outofssa_var_map): Delete.
25 (coalesce_ssa_name): Refactor to support live range computation.
26 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
27 (get_parm_default_def_partitions): Delete.
28 (get_undefined_value_partitions): Ditto.
29 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
30 computation for loop region.
31 (new_tree_live_info, loe_visit_block): Ditto.
32 (live_worklist, set_var_live_on_entry): Ditto.
33 (calculate_live_on_exit, verify_live_on_entry): Ditto.
34 * tree-ssa-live.h (struct _var_map): New fields.
35 (init_var_map): Change decl.
36 (region_contains_p): New.
38 2018-05-25 Bin Cheng <bin.cheng@arm.com>
40 * tree-ssa-live.h (live_merge_and_clear): Delete.
42 2018-05-25 Richard Biener <rguenther@suse.de>
45 * tree-dump.c (dequeue_and_dump): Remove access to removed
46 operand 2 of a SWITCH_EXPR.
48 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
50 * doc/sourcebuild.texi (vect_double_cond_arith): Include
51 multiplication and division.
52 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
53 (cond_udiv@var{m}, cond_umod@var{m}): Document.
54 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
55 (cond_udiv_optab, cond_umod_optab): New optabs.
56 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
57 (IFN_COND_RDIV): New internal functions.
58 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
59 TRUNC_MOD_EXPR and RDIV_EXPR.
60 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
61 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
63 (SVE_INT_BINARY): Include mult.
64 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
65 (optab, sve_int_op): Handle mult.
66 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
68 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
69 for SVE_INT_BINARY_SD.
71 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
73 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
74 (optab, sve_int_op): Handle div and udiv.
75 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
76 for SVE_INT_BINARY_SD.
77 (*<optab><mode>3): New insn for the same.
79 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
81 * tree-vect-patterns.c: Include predict.h.
82 (vect_recog_divmod_pattern): Restrict check for division support
83 to when optimizing for size.
85 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
87 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
88 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
89 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
90 (gimple_match_op::set_op): Likewise.
91 (gimple_resimplify4): Declare.
92 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
93 (expr::gen_transform): Likewise.
94 (decision_tree::gen): Generate a simplification routine for 4 operands.
95 * gimple-match-head.c (gimple_simplify): Add an overload for
96 4 operands. In the top-level function, handle up to 4 call
97 arguments and call gimple_resimplify4.
98 (gimple_resimplify4): New function.
99 (build_call_internal): Pass a fourth operand.
100 (maybe_push_to_seq): Likewise.
101 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
102 Fold VEC_COND_EXPRs of an operation and a default value into
103 an IFN_COND_* function if possible.
104 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
106 (SVE_COND_FP_BINARY): Include them.
107 (optab, sve_fp_op): Handle them.
108 (SVE_INT_BINARY_REV): New code iterator.
109 (SVE_COND_FP_BINARY_REV): New int iterator.
110 (commutative): New int attribute.
111 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
113 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
115 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
116 (*cond_<optab><mode>): New patterns for reversed operands.
118 2018-05-25 Richard Biener <rguenther@suse.de>
120 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
121 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
122 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
123 (STMT_VINFO_GROUPED_ACCESS): Adjust.
124 * tree-vect-data-refs.c (everywhere): Adjust users.
125 * tree-vect-loop.c (everywhere): Likewise.
126 * tree-vect-slp.c (everywhere): Likewise.
127 * tree-vect-stmts.c (everywhere): Likewise.
128 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
130 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
132 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
134 (gcc_cv_as_section_exclude): ... this.
135 Try Solaris as #exclude syntax.
136 * configure: Regenerate.
137 * config.in: Regenerate.
138 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
140 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
141 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
143 * varasm.c (default_elf_asm_named_section): Don't check if
144 HAVE_GAS_SECTION_EXCLUDE is defined.
146 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
148 * doc/md.texi: Update the documentation of the cond_* optabs
149 to mention the new final operand. Fix GET_MODE_NUNITS call.
150 Describe the scalar case too.
151 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
152 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
154 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
155 (get_conditional_internal_fn): Update comment.
156 * tree-vect-loop.c (vectorizable_reduction): Pass the original
157 accumulator value as a final argument to conditional functions.
158 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
159 a define_expand and add an "else" operand. Assert for now that
160 the else operand is equal to operand 2. Use SVE_INT_BINARY and
161 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
162 (*cond_<optab><mode>): New patterns.
163 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
164 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
165 (UNSPEC_COND_EOR): Delete.
166 (optab): Remove associated mappings.
167 (SVE_INT_BINARY): New code iterator.
168 (sve_int_op): Remove int attribute and add "minus" to the code
170 (SVE_COND_INT_OP): Delete.
171 (SVE_COND_FP_OP): Rename to...
172 (SVE_COND_FP_BINARY): ...this.
174 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
176 * optabs.c (can_reuse_operands_p): New function.
177 (maybe_legitimize_operands): Try to reuse the results for
180 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
182 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
183 Add {q} suffix to insn mnemonic.
185 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
187 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
188 (msp430_warn_func_return): New.
190 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
192 * fold-const.c (tree_nonzero_bits): New function.
193 * fold-const.h (tree_nonzero_bits): Likewise.
194 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
195 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
197 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
201 * varasm.c (assemble_alias): Check ifunc_resolver only on
204 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
207 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
208 when memory input operand is handled.
210 2018-05-24 Luis Machado <luis.machado@linaro.org>
212 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
214 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
216 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
218 * match.pd: Delay FMA folds until after vectorization.
220 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
223 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
224 address check not strict.
226 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
228 * gimple-match.h (gimple_match_op): New class.
229 (mprts_hook): Replace parameters with a gimple_match_op *.
230 (maybe_build_generic_op): Likewise.
231 (gimple_simplified_result_is_gimple_val): Replace parameters with
232 a const gimple_match_op *.
233 (gimple_simplify): Replace code_helper * and tree * parameters with
234 a gimple_match_op * parameter.
235 (gimple_resimplify1): Replace code_helper *, tree and tree *
236 parameters with a gimple_match_op * parameter.
237 (gimple_resimplify2): Likewise.
238 (gimple_resimplify3): Likewise.
239 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
240 parameters with a gimple_match_op * parameter.
241 * gimple-match-head.c (gimple_simplify): Change prototypes of
242 auto-generated functions to take a gimple_match_op * instead of
243 separate code_helper * and tree * parameters. Make the same
244 change in the top-level overload and update calls to the
245 gimple_resimplify routines. Update calls to the auto-generated
246 functions and to maybe_push_res_to_seq in the publicly-facing
247 operation-specific gimple_simplify overloads.
248 (gimple_match_op::MAX_NUM_OPS): Define.
249 (gimple_resimplify1): Replace rcode and ops with a single res_op
250 parameter. Update call to gimple_simplify.
251 (gimple_resimplify2): Likewise.
252 (gimple_resimplify3): Likewise.
253 (mprts_hook): Replace parameters with a gimple_match_op *.
254 (maybe_build_generic_op): Likewise.
255 (build_call_internal): Replace type, nargs and ops with
257 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
258 with a single gimple_match_op *. Update calls to mprts_hook,
259 build_call_internal and gimple_simplified_result_is_gimple_val.
260 Factor out code that is common to the tree_code and combined_fn cases.
261 * genmatch.c (expr::gen_transform): Replace tem_code and
262 tem_ops with a gimple_match_op called tem_op. Update calls
263 to the gimple_resimplify functions and maybe_push_res_to_seq.
264 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
265 res_ops. Update call to the gimple_resimplify functions.
266 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
267 (decision_tree::gen): Make the functions take a gimple_match_op *
268 called res_op instead of separate res_code and res_ops parameters.
269 Update call accordingly.
270 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
271 and ops with a single res_op parameter. Update calls to
272 maybe_build_generic_op and maybe_push_res_to_seq.
273 (fold_stmt_1): Update calls to gimple_simplify and
274 replace_stmt_with_simplification.
275 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
276 and gimple_simplified_result_is_gimple_val.
277 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
279 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
280 with a gimple_match_op *.
281 (vn_nary_build_or_lookup): Likewise. Update call to
282 vn_nary_build_or_lookup_1.
283 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
284 gimple_match_op *. Update calls to the gimple_resimplify routines
285 and to gimple_simplified_result_is_gimple_val.
286 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
287 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
288 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
289 (visit_nary_op): Likewise.
290 (visit_reference_op_load): Likewise.
292 2018-05-23 Luis Machado <luis.machado@linaro.org>
294 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
295 modifier for printing the step amount.
297 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
300 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
303 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
305 * doc/sourcebuild.texi (Endianness): New subsubsection.
307 2018-05-23 Luis Machado <luis.machado@linaro.org>
309 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
310 <prefetch_dynamic_strides>: New const bool field.
311 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
312 prefetch_dynamic_strides.
313 (exynosm1_prefetch_tune): Likewise.
314 (thunderxt88_prefetch_tune): Likewise.
315 (thunderx_prefetch_tune): Likewise.
316 (thunderx2t99_prefetch_tune): Likewise.
317 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
319 (aarch64_override_options_internal): Update to set
320 PARAM_PREFETCH_DYNAMIC_STRIDES.
321 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
322 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
323 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
324 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
325 prefetch-dynamic-strides setting.
327 2018-05-23 Luis Machado <luis.machado@linaro.org>
329 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
330 <minimum_stride>: New const int field.
331 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
332 minimum_stride field defaulting to -1.
333 (exynosm1_prefetch_tune): Likewise.
334 (thunderxt88_prefetch_tune): Likewise.
335 (thunderx_prefetch_tune): Likewise.
336 (thunderx2t99_prefetch_tune): Likewise.
337 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
338 <default_opt_level>: Set to 3.
339 (aarch64_override_options_internal): Update to set
340 PARAM_PREFETCH_MINIMUM_STRIDE.
341 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
342 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
343 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
344 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
345 stride is constant and is below the minimum stride threshold.
347 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
349 * config/arm/arm-cpus.in (mode26): Delete.
350 (armv4): Delete mode26 reference.
351 * config/arm/arm.c (arm_configure_build_target): Delete use of
354 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
356 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
358 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
359 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
361 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
362 (floatunsdidf2): Ditto.
364 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
366 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
367 (fixuns_trunc<mode>si2_avx512f): Ditto.
368 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
369 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
370 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
372 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
374 PR rtl-optimization/79985
375 * df-scan.c (df_insn_refs_collect): Remove special case for
376 global registers and asm statements.
378 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
380 * extend.texi (Global Register Variables): Rewrite the bullet list.
381 Note that the register is available for allocation. Note that access
382 via inline asm must use constraints. Add note about async-signal
383 handlers. Remove paragraph about automagic register selection.
385 2018-05-23 Richard Biener <rguenther@suse.de>
387 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
388 of fixed offset from memset VN.
390 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
392 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
394 (alloc_cand_and_find_basis): Initialize first_interp field.
395 (slsr_process_mul): Modify first_interp field.
396 (slsr_process_add): Likewise.
397 (slsr_process_cast): Modify first_interp field for each new
399 (slsr_process_copy): Likewise.
400 (dump_candidate): Dump first_interp field.
401 (replace_mult_candidate): Process all interpretations, not just
403 (replace_rhs_if_not_dup): Likewise.
404 (replace_one_candidate): Likewise.
406 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
408 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
410 (aarch64_needs_frame_chain): New function.
411 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
413 2018-05-23 Sudakshina Das <sudi.das@arm.com>
416 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
417 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
418 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
419 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
420 as true for strict-align.
421 (aarch64_can_inline_p): Perform checks even when callee has no
422 attributes to check for strict alignment.
423 * doc/extend.texi (AArch64 Function Attributes): Document
425 * doc/invoke.texi: (AArch64 Options): Likewise.
427 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
429 PR tree-optimization/85853
430 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
431 the handling of the root of the node to...
432 (vect_slp_analyze_node_operations_1): ...this new function,
433 and run the whole thing with the child nodes' def types
434 set according to their SLP node's def type.
436 2018-05-23 Richard Biener <rguenther@suse.de>
439 * tree-data-ref.c (create_runtime_alias_checks): Defer
440 and ignore overflow warnings.
442 2018-05-23 Yury Gribov <tetra2005@gmail.com>
444 PR tree-optimization/85822
445 * tree-vrp.c (is_masked_range_test): Fix handling of negative
448 2018-05-23 Richard Biener <rguenther@suse.de>
450 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
451 memset constants via native_interpret_expr.
453 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
456 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
458 (cgraph_node::create_alias): Likewise.
459 (cgraph_node::get_availability): Check ifunc_resolver instead
460 of looking up ifunc attribute.
461 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
462 * varasm.c (do_assemble_alias): Likewise.
463 (assemble_alias): Likewise.
464 (default_binds_local_p_3): Likewise.
465 * cgraph.h (cgraph_node): Add ifunc_resolver.
466 (cgraph_node::only_called_directly_or_aliased_p): Return false
468 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
470 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
471 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
472 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
473 instead of looking up ifunc attribute.
475 2018-05-22 Luis Machado <luis.machado@linaro.org>
477 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
479 2018-05-22 Martin Sebor <msebor@redhat.com>
482 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
483 only when expasion succeeds.
484 (expand_builtin_strcmp): Same.
485 (expand_builtin_strncmp): Same.
487 2018-05-22 Martin Sebor <msebor@redhat.com>
489 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
491 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
492 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
494 * config/aarch64/aarch64-ldpstp.md: Replace uses of
495 aarch64_mem_pair_operand with memory_operand and delete operand swapping
497 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
498 Add check for legitimate_address.
499 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
500 (aarch64_swap_ldrstr_operands): New.
501 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
504 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
505 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
507 * config/aarch64/aarch64.md: New patterns to generate stp
509 (store_pair_sw, store_pair_dw): New patterns to generate stp for
510 single words and double words.
511 (load_pair_sw, load_pair_dw): Likewise.
512 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
514 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
516 * config/aarch64/aarch64-ldpstp.md: Modify peephole
517 for different mode ldpstp and add peephole for merged zero stores.
519 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
521 (aarch64_gen_store_pair): Rename calls to match new patterns.
522 (aarch64_gen_load_pair): Rename calls to match new patterns.
523 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
524 (load_pair<DREG:mode><DREG2:mode>): ... This.
525 (store_pair<mode>): Rename to...
526 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
527 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
529 (V_INT_EQUIV): Handle SImode.
530 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
533 2018-05-22 Martin Sebor <msebor@redhat.com>
536 * calls.c (maybe_warn_nonstring_arg): Use string length to set
537 or ajust the presumed bound on an operation to avoid unnecessary
540 2018-05-22 Martin Sebor <msebor@redhat.com>
542 PR tree-optimization/85826
543 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
544 assuming that a DECL necesarily has a constant size.
546 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
549 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
551 2018-05-22 Richard Biener <rguenther@suse.de>
553 PR tree-optimization/85834
554 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
555 non-constant and non-zero memset arguments.
557 2018-05-22 Martin Liska <mliska@suse.cz>
560 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
562 2018-05-22 Richard Biener <rguenther@suse.de>
564 PR tree-optimization/85863
565 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
566 comparisons when vectype is specified.
567 (vectorizable_condition): Do not specify vectype for
568 vect_is_simple_cond when SLP vectorizing.
570 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
573 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
574 define __ibm128 as long double.
575 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
576 as a distinct type when IEEE 128-bit support is enabled.
577 (init_float128_ieee): Fix up conversions between IFmode and IEEE
578 128-bit types to use the correct functions.
579 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
580 convert between 128-bit floating point types that have different
581 modes but the same representation, instead of using gen_lowpart to
583 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
585 (IFKF_reg): New attributes to give the register constraints for
587 (extend<mode>tf2_internal): New insns to mark an explicit
588 conversion between 128-bit floating point types that have a
589 different mode but share the same representation.
591 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
593 PR tree-optimization/85814
594 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
595 a null return from get_strinfo when unsharing the next
596 strinfo in the chain.
598 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
601 * varasm.c (weak_finish): Clean up weak_decls.
603 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
605 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
606 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
607 UNSPEC_UADALP values.
608 * config/aarch64/iterators.md (ABAL): New int iterator.
611 (sur): Add mappings for the above.
612 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
614 (aarch64_<sur>abal<mode>_4): Likewise.
615 (aarch64_<sur>adalp<mode>_3): Likewise.
616 (<sur>sadv16qi): New define_expand.
618 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
620 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
621 (*movdf_internal): Ditto.
622 (*rcpsf2_sse): Ditto.
623 (*rsqrtsf2_sse): Ditto.
624 (*sqrt<mode>2_sse): Ditto.
626 2018-05-21 Tamar Christina <tamar.christina@arm.com>
628 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
630 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
631 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
632 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
633 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
635 * config/aarch64/arm_neon.h: Likewise.
636 * config/aarch64/iterators.md (VQ_I): New.
638 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
640 * config.gcc: Add arc/t-multilib-linux to tmake_file for
642 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
645 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
647 * config/nds32/constraints.md (S): New constraint.
648 * config/nds32/nds32.md (call_internal): Use constraint S.
649 (call_value_internal): Likewise.
650 (sibcall_internal): Likewise.
651 (sibcall_value_internal): Likewise.
653 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
654 Chung-Ju Wu <jasonwucj@gmail.com>
656 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
659 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
660 Chung-Ju Wu <jasonwucj@gmail.com>
662 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
663 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
664 (nds32_rtx_costs_impl): Simplify.
665 (nds32_address_cost_impl): Simplify.
666 (nds32_init_rtx_costs): New function.
667 (nds32_rtx_costs_speed_prefer): Likewise.
668 (nds32_rtx_costs_size_prefer): Likewise.
669 (nds32_address_cost_speed_prefer): Likewise.
670 (nds32_address_cost_speed_fwprop): Likewise.
671 (nds32_address_cost_size_prefer): Likewise.
672 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
673 * config/nds32/nds32.c (nds32_option_override): Use
674 nds32_init_rtx_costs function.
676 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
678 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
679 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
680 (TARGET_PIPELINE_N8): Likewise.
681 (TARGET_PIPELINE_N10): Likewise.
682 (TARGET_PIPELINE_N13): Likewise.
683 (TARGET_PIPELINE_GRAYWOLF): Likewise.
685 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
687 * config/nds32/nds32-fpu.md: Update copyright year.
689 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
691 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
693 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
695 * config/nds32/nds32.c
696 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
697 * config/nds32/nds32.opt (minline-asm-r15): New option.
699 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
701 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
703 * config/nds32/nds32.md (abssi2): New pattern.
705 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
707 * config/i386/i386.md (rex64namesuffix): New mode attribute.
708 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
709 Merge insn pattern from sse_cvtsi2ss<round_name> and
710 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
711 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
712 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
713 using SWI48 mode iterator.
714 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
715 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
716 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
717 pattern from sse_cvttss2si<round_saeonly_name>
718 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
719 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
720 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
721 using SWI48 mode iterator.
722 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
723 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
724 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
725 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
726 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
727 using SWI48 mode iterator.
728 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
729 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
730 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
731 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
732 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
734 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
735 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
736 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
737 pattern from sse_cvttsd2si<round_saeonly_name>
738 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
740 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
742 * config/nds32/nds32-md-auxiliary.c
743 (nds32_valid_smw_lwm_base_p): Refine.
744 (nds32_output_smw_single_word): Refine.
745 (nds32_output_smw_double_word): New.
746 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
748 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
750 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
751 (nds32_output_stack_pop): Refine.
752 (nds32_expand_unaligned_load): Refine.
753 (nds32_expand_unaligned_store): Refine.
755 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
756 Chung-Ju Wu <jasonwucj@gmail.com>
758 * config/nds32/constants.md: Add TP_REGNUM constant.
759 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
760 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
762 * config/nds32/nds32-doubleword.md: Consider flag_pic.
763 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
764 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
765 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
766 and PIC code generation.
767 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
769 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
771 * config/nds32/nds32.md: Support TLS and PIC.
772 * config/nds32/nds32.c: Support TLS and PIC.
773 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
774 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
777 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
779 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
782 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
783 Chung-Ju Wu <jasonwucj@gmail.com>
785 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
786 * config/nds32/nds32-md-auxiliary.c
787 (symbolic_reference_mentioned_p): New.
788 (nds32_legitimize_ict_address): New.
789 (nds32_expand_ict_move): New.
790 (nds32_indirect_call_referenced_p): New.
791 (nds32_symbol_binds_local_p): Delete.
792 (nds32_long_call_p): Modify.
793 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
794 * config/nds32/nds32-protos.h
795 (symbolic_reference_mentioned_p): Declare.
796 (nds32_legitimize_ict_address): Declare.
797 (nds32_expand_ict_move): Declare.
798 (nds32_indirect_call_referenced_p): Declare.
799 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
800 (nds32_relax_group): Use nds32_ict_const_p as condition.
801 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
802 (nds32_asm_file_start): Output ict_model directive in asm code.
803 (nds32_legitimate_address_p): Consider indirect call.
804 (nds32_print_operand): Consider indirect call.
805 (nds32_print_operand_address): Consider indirect call.
806 (nds32_insert_attributes): Handle "indirect_call" attribute.
807 (TARGET_LEGITIMATE_ADDRESS_P): Define.
808 (TARGET_LEGITIMATE_CONSTANT_P): Define.
809 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
810 (TARGET_DELEGITIMIZE_ADDRESS): Define.
811 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
812 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
813 (TARGET_ICT_MODEL_SMALL): Define.
814 (TARGET_ICT_MODEL_LARGE): Define.
815 * config/nds32/nds32.md (movsi): Consider ict model.
816 (call, call_value): Consider ict model.
817 (sibcall, sibcall_value): Consider ict model.
818 * config/nds32/nds32.opt (mict-model): New option.
819 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
822 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
823 Monk Chiang <sh.chiang04@gmail.com>
824 Jim Wilson <jimw@sifive.com>
826 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
827 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
828 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
829 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
830 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
831 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
832 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
833 compute save_libcall_adjustment properly.
834 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
835 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
836 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
837 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
838 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
839 (ABI_SPEC): Handle mabi=ilp32e.
840 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
842 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
843 <-march>: Add rv32e as an example.
845 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
848 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
849 (intra_create_variable_infos): Handle C++ constructors.
851 2018-05-18 Martin Liska <mliska@suse.cz>
853 * passes.def: Remove a redundant pass.
855 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
858 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
860 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
862 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
864 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
866 (armv2, armv2a, armv3, armv3m): Delete architectures.
867 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
868 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
869 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
871 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
872 (*mulsidi3adddi): Likewise.
873 (mulsidi3): Likewise.
874 (*mulsidi3_nov6): Likewise.
875 (umulsidi3): Likewise.
876 (umulsidi3_nov6): Likewise.
877 (umaddsidi4): Likewise.
878 (*umulsidi3adddi): Likewise.
879 (smulsi3_highpart): Likewise.
880 (*smulsi3_highpart_nov6): Likewise.
881 (umulsi3_highpart): Likewise.
882 (*umulsi3_highpart_nov6): Likewise.
883 * config/arm/arm.h (arm_arch3m): Delete.
884 * config/arm/arm.c (arm_arch3m): Delete.
885 (arm_option_override_internal): Update armv3-related comment.
886 (arm_configure_build_target): Delete use of isa_bit_mode32.
887 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
888 (arm_rtx_costs_internal): Delete check of arm_arch3m.
889 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
892 * config/arm/arm-protos.h (arm_arch3m): Delete.
893 * config/arm/arm-tables.opt: Regenerate.
894 * config/arm/arm-tune.md: Likewise.
895 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
896 deleted architectures.
898 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
900 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
901 (armv5t, armv5te): New features.
902 (ARMv5, ARMv5e): Delete fgroups.
903 (ARMv5t, ARMv5te): Adjust for above changes.
905 (armv5, armv5e): Delete arches.
906 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
908 (*call_reg_arm): Likewise.
909 (*call_value_reg_armv5): Likewise.
910 (*call_value_reg_arm): Likewise.
911 (*call_symbol): Likewise.
912 (*call_value_symbol): Likewise.
913 (*sibcall_insn): Likewise.
914 (*sibcall_value_insn): Likewise.
916 (prefetch): Likewise.
917 (define_split and define_peephole2 dependent on arm_arch5):
919 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
921 (TARGET_ARM_QBIT): Likewise.
922 (TARGET_DSP_MULTIPLY): Likewise.
923 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
924 (arm_arch5, arm_arch5e): Delete.
925 (arm_arch5t, arm_arch5te): Declare.
926 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
927 (arm_arch5t): Declare.
928 (arm_option_reconfigure_globals): Update for the above.
929 (arm_options_perform_arch_sanity_checks): Update comment, replace
930 use of arm_arch5 with arm_arch5t.
931 (use_return_insn): Likewise.
932 (arm_emit_call_insn): Likewise.
933 (output_return_instruction): Likewise.
934 (arm_final_prescan_insn): Likewise.
935 (arm_coproc_builtin_available): Likewise.
936 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
937 arm_arch5e with arm_arch5t and arm_arch5te.
938 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
939 (arm_arch5t, arm_arch5te): Declare.
940 * config/arm/arm-tables.opt: Regenerate.
941 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
942 * config/arm/t-multilib: Likewise.
943 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
944 instead of arm_arch5.
945 (*call_reg_thumb1): Likewise.
946 (*call_value_reg_thumb1_v5): Likewise.
947 (*call_value_reg_thumb1): Likewise.
948 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
950 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
952 2018-05-18 Martin Liska <mliska@suse.cz>
954 PR gcov-profile/84846
955 * doc/gcov.texi: Document -t option of gcov tool.
957 2018-05-18 Martin Liska <mliska@suse.cz>
959 PR gcov-profile/84846
960 * gcov.c (print_usage): Add new -t option.
961 (process_args): Handle the option.
962 (generate_results): Use stdout as output when requested by
965 2018-05-18 Martin Liska <mliska@suse.cz>
967 PR gcov-profile/84846
968 * coverage.c (coverage_init): Write PWD to .gcno file.
969 * doc/gcov.texi: Document how working directory is printed.
970 * gcov-dump.c (dump_gcov_file): Print PWD.
971 * gcov.c (output_intermediate_file): Likewise.
972 (read_graph_file): Read PWD string.
973 (output_lines): Print PWD.
975 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
978 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
979 for retval and return false if all args to phi are zero.
981 2018-05-18 Richard Biener <rguenther@suse.de>
983 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
985 (evrp_dom_walker::before_dom_children): Call it.
987 2018-05-18 Richard Biener <rguenther@suse.de>
989 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
990 results when processing array refs with variable index.
992 2018-05-18 Toon Moene <toon@moene.org>
994 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
995 directly after that of -floop-interchange. Indicate that both
996 options are enabled by default when specifying -O3.
998 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1000 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1001 iterator. Delete separate integer-mode vec_set<mode> expander.
1002 (aarch64_simd_vec_setv2di): Delete.
1003 (vec_setv2di): Delete.
1004 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1005 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1006 the "w, r" alternative.
1008 2018-05-18 Martin Liska <mliska@suse.cz>
1010 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1011 * tree-pass.h (make_pass_lower_switch_O0): New function.
1012 * tree-switch-conversion.c (node_has_low_bound): Remove.
1013 (node_has_high_bound): Likewise.
1014 (node_is_bounded): Likewise.
1015 (class pass_lower_switch): Make it a template type and create
1017 (pass_lower_switch::execute): Add template argument.
1018 (make_pass_lower_switch): New function.
1019 (make_pass_lower_switch_O0): New function.
1020 (do_jump_if_equal): Remove.
1021 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1022 all the hard work to tree optimization passes.
1024 2018-05-18 Martin Liska <mliska@suse.cz>
1026 * dbgcnt.c (limit_low): Renamed from limit.
1027 (limit_high): New variable.
1028 (dbg_cnt_is_enabled): Check for upper limit.
1029 (dbg_cnt): Adjust dumping.
1030 (dbg_cnt_set_limit_by_index): Add new argument for high
1032 (dbg_cnt_set_limit_by_name): Likewise.
1033 (dbg_cnt_process_single_pair): Parse new format.
1034 (dbg_cnt_process_opt): Use strtok.
1035 (dbg_cnt_list_all_counters): Remove 'value' and add
1037 * doc/invoke.texi: Document changes.
1039 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1041 * doc/sourcebuild.texi (scalar_all_fma): Document.
1042 * tree.def (FMA_EXPR): Delete.
1043 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1044 * internal-fn.c (ternary_direct): New macro.
1045 (expand_ternary_optab_fn): Likewise.
1046 (direct_ternary_optab_supported_p): Likewise.
1047 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1048 * builtins.c (fold_builtin_fma): Delete.
1049 (fold_builtin_3): Don't call it.
1050 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1051 * expr.c (expand_expr_real_2): Likewise.
1052 * fold-const.c (operand_equal_p): Likewise.
1053 (fold_ternary_loc): Likewise.
1054 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1055 * gimple.c (DEFTREECODE): Likewise.
1056 * gimplify.c (gimplify_expr): Likewise.
1057 * optabs-tree.c (optab_for_tree_code): Likewise.
1058 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1059 * tree-eh.c (operation_could_trap_p): Likewise.
1060 (stmt_could_throw_1_p): Likewise.
1061 * tree-inline.c (estimate_operator_cost): Likewise.
1062 * tree-pretty-print.c (dump_generic_node): Likewise.
1063 (op_code_prio): Likewise.
1064 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1065 * tree-ssa-operands.c (get_expr_operands): Likewise.
1066 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1067 * fold-const-call.h (fold_fma): Delete.
1068 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1069 CFN_FNMA and CFN_FNMS.
1071 * genmatch.c (combined_fn): New enum.
1072 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1073 (commutative_op): New function.
1074 (commutate): Use it. Handle more than 2 operands.
1075 (dt_operand::gen_gimple_expr): Use commutative_op.
1076 (parser::parse_expr): Allow :c to be used with non-binary
1077 operators if the commutative operand is known.
1078 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1079 CFN_FMS, CFN_FNMA and CFN_FNMS.
1080 (backprop::process_assign_use): Remove FMA_EXPR handling.
1081 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1082 (gen_hsa_fma): New function.
1083 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1084 IFN_FNMA and IFN_FNMS.
1085 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1086 * gimple-fold.h (follow_all_ssa_edges): Declare.
1087 * gimple-fold.c (follow_all_ssa_edges): New function.
1088 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1089 gimple_build interface and use follow_all_ssa_edges to fold the result.
1090 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1091 instead of checking for optabs directly.
1092 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1093 rather than FMA_EXPRs.
1094 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1095 call to IFN_FMA instead of an FMA_EXPR.
1097 2018-05-17 Jim Wilson <jimw@sifive.com>
1099 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1100 sign extended promoted subreg, and the range does not have the sign bit
1101 set, then do a sign extend.
1103 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1104 test, check for sign extended subreg and/or constant operands, and
1105 do a sign extend in that case.
1107 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1109 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1110 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1112 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1113 Change logics_shift_reg to logics_shift_imm.
1114 (thunderx2t99_fp_loadpair_basic): Delete.
1115 (thunderx2t99_fp_storepair_basic): Delete.
1116 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1117 (thunderx2t99_asimd_polynomial): Delete.
1118 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1119 and neon_fp_mul_d_scalar_q.
1120 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1121 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1122 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1123 (thunderx2t99_asimd_lut): Add missing tbl types.
1124 (thunderx2t99_asimd_ext): Delete.
1125 (thunderx2t99_asimd_load1_1_mult): Delete.
1126 (thunderx2t99_asimd_load1_2_mult): Delete.
1127 (thunderx2t99_asimd_load1_ldp): New.
1128 (thunderx2t99_asimd_load1): New.
1129 (thunderx2t99_asimd_load2): Add missing *load2* types.
1130 (thunderx2t99_asimd_load3): New.
1131 (thunderx2t99_asimd_load4): New.
1132 (thunderx2t99_asimd_store1_1_mult): Delete.
1133 (thunderx2t99_asimd_store1_2_mult): Delete.
1134 (thunderx2t99_asimd_store2_mult): Delete.
1135 (thunderx2t99_asimd_store2_onelane): Delete.
1136 (thunderx2t99_asimd_store_stp): New.
1137 (thunderx2t99_asimd_store1): New.
1138 (thunderx2t99_asimd_store2): New.
1139 (thunderx2t99_asimd_store3): New.
1140 (thunderx2t99_asimd_store4): New.
1142 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1144 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1145 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1147 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1148 Segher Boessenkool <segher@kernel.crashing.org>
1151 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1154 2018-05-17 Richard Biener <rguenther@suse.de>
1156 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1157 for pruning loop and prune defs feeding only already visited PHIs.
1159 2018-05-17 Richard Biener <rguenther@suse.de>
1161 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1163 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1164 Richard Biener <rguenther@suse.de>
1166 PR tree-optimization/85793
1167 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1168 for VMAT_ELEMENTWISE.
1170 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1172 * internal-fn.h (lookup_internal_fn): Declare
1173 * internal-fn.c (lookup_internal_fn): New function.
1174 * gimple.c (gimple_build_call_from_tree): Handle calls to
1176 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1177 internal function names.
1178 * tree-pretty-print.c (dump_generic_node): Likewise.
1179 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1181 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1183 * gimple-fold.h (gimple_build): Make the function forms take
1184 combined_fn rather than built_in_function.
1185 (gimple_simplify): Likewise.
1186 * gimple-match-head.c (gimple_simplify): Likewise.
1187 * gimple-fold.c (gimple_build): Likewise.
1188 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1189 rather than gimple_build_call_internal.
1190 (get_initial_defs_for_reduction): Likewise.
1191 (vect_create_epilog_for_reduction): Likewise.
1192 (vectorizable_live_operation): Likewise.
1194 2018-05-17 Martin Liska <mliska@suse.cz>
1196 * gimple-ssa-sprintf.c (format_directive): Do not use
1197 space in between 'G_' and '('.
1199 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1202 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1203 even if the mask is not all ones.
1206 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1208 (ix86_gimple_fold_builtin): Likewise.
1211 * config/i386/i386.c: Include tree-vector-builder.h.
1212 (ix86_vector_shift_count): New function.
1213 (ix86_fold_builtin): Fold shift builtins by scalar count.
1214 (ix86_gimple_fold_builtin): Likewise.
1216 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1217 _mm512_setzero): New intrinsics.
1219 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1220 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1222 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1223 code generation for cases where splatting a value is not useful.
1224 * simplify-rtx.c (simplify_ternary_operation): Simplify
1225 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1226 mode to a vec_concat.
1228 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1230 * config.gcc: Support "goldmont-plus".
1231 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1233 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1234 PROCESSOR_GOLDMONT_PLUS.
1235 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1236 (processor_target_table): Add "goldmont-plus".
1237 (PTA_GOLDMONT_PLUS): Define.
1238 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1239 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1240 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1241 (fold_builtin_cpu): Add "goldmont-plus".
1242 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1243 (ix86_option_override_internal): Add "goldmont-plus".
1244 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1245 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1246 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1247 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1249 2018-05-17 Richard Biener <rguenther@suse.de>
1251 PR tree-optimization/85757
1252 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1253 remove defs that only feed that PHI from further processing.
1255 2018-05-16 Jim Wilson <jimw@sifive.com>
1257 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1259 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1261 2018-05-16 Mark Wielaard <mark@klomp.org>
1263 * dwarf2out.c (count_index_strings): New function.
1264 (output_indirect_strings): Call count_index_strings and generate
1265 header for dwarf_version >= 5.
1267 2018-05-16 Mark Wielaard <mark@klomp.org>
1269 * dwarf2out.c (dwarf_FORM): New function.
1270 (set_indirect_string): Use dwarf_FORM.
1271 (reset_indirect_string): Likewise.
1272 (size_of_die): Likewise.
1273 (value_format): Likewise.
1274 (output_die): Likewise.
1275 (add_skeleton_AT_string): Likewise.
1276 (output_macinfo_op): Likewise.
1277 (index_string): Likewise.
1278 (output_index_string_offset): Likewise.
1279 (output_index_string): Likewise.
1280 (count_index_strings): Likewise.
1282 2018-05-16 Carl Love <cel@us.ibm.com>
1284 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1285 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1287 2018-05-16 Martin Jambor <mjambor@suse.cz>
1289 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1290 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1292 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1294 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1295 (fnma<mode>4): Likewise.
1296 (fms<mode>4): Likewise.
1297 (fnms<mode>4): Likewise.
1298 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1299 (aarch64_fnma<mode>4): Likewise.
1300 (aarch64_fms<mode>4): Likewise.
1301 (aarch64_fnms<mode>4): Likewise.
1302 (aarch64_fnmadd<mode>4): Likewise.
1304 2018-05-16 Jason Merrill <jason@redhat.com>
1306 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1308 2018-05-16 Richard Biener <rguenther@suse.de>
1310 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1311 (dump_stmt_cost): Declare.
1312 (add_stmt_cost): Dump cost we add.
1313 (add_stmt_costs): New function.
1314 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1316 (vect_analyze_stmt): Adjust prototype.
1317 (vectorizable_condition): Likewise.
1318 (vectorizable_live_operation): Likewise.
1319 (vectorizable_reduction): Likewise.
1320 (vectorizable_induction): Likewise.
1321 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1322 cost vector to pass to vectorizable_ and record afterwards.
1323 (vect_model_reduction_cost): Take cost vector argument and adjust.
1324 (vect_model_induction_cost): Likewise.
1325 (vectorizable_reduction): Likewise.
1326 (vectorizable_induction): Likewise.
1327 (vectorizable_live_operation): Likewise.
1328 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1329 SLP_TREE_NUMBER_OF_VEC_STMTS.
1330 (vect_analyze_slp_cost_1): Remove.
1331 (vect_analyze_slp_cost): Likewise.
1332 (vect_slp_analyze_node_operations): Take visited args and
1333 a target cost vector. Avoid processing already visited stmt sets.
1334 (vect_slp_analyze_operations): Use a local cost vector to gather
1335 costs and register those of non-discarded instances.
1336 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1337 (vect_schedule_slp_instance): Remove copying of
1338 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1340 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1341 adding cost. Record cost entry location.
1342 (vect_prologue_cost_for_slp_op): Function to compute cost of
1343 a constant or invariant generated for SLP vect in the prologue,
1344 split out from vect_analyze_slp_cost_1.
1345 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1346 (vect_model_promotion_demotion_cost): Likewise.
1347 (vect_model_store_cost): Likewise, make static.
1348 (vect_model_load_cost): Likewise.
1349 (vectorizable_bswap): Add cost vector arg and adjust.
1350 (vectorizable_call): Likewise.
1351 (vectorizable_simd_clone_call): Likewise.
1352 (vectorizable_conversion): Likewise.
1353 (vectorizable_assignment): Likewise.
1354 (vectorizable_shift): Likewise.
1355 (vectorizable_operation): Likewise.
1356 (vectorizable_store): Likewise.
1357 (vectorizable_load): Likewise.
1358 (vectorizable_condition): Likewise.
1359 (vectorizable_comparison): Likewise.
1360 (can_vectorize_live_stmts): Likewise.
1361 (vect_analyze_stmt): Likewise.
1362 (vect_transform_stmt): Adjust calls to vectorizable_*.
1363 * tree-vectorizer.c: Include gimple-pretty-print.h.
1364 (dump_stmt_cost): New function.
1366 2018-05-16 Richard Biener <rguenther@suse.de>
1368 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1369 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1370 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1371 (check_name): New callback.
1372 (dse_classify_store): Track cycles via a visited bitmap of PHI
1373 defs and simplify handling of in-loop and across loop dead stores
1374 and properly fail for loop-variant refs. Handle byte-tracking with
1375 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1378 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1380 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1381 (vect_get_mask_type_for_stmt): Likewise.
1382 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1384 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1385 to determine the statement's vector type and the vector type that
1386 should be used for calculating nunits. Deal with cases in which
1387 the type has to be deferred.
1388 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1389 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1390 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1391 (vect_determine_vf_for_stmt): New functions, split out from...
1392 (vect_determine_vectorization_factor): ...here.
1393 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1394 (vect_get_mask_type_for_stmt): New functions, split out from
1395 vect_determine_vectorization_factor.
1397 2018-05-16 Richard Biener <rguenther@suse.de>
1399 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1400 verify the [VEC_]COND_EXPR embedded comparison.
1402 2018-05-15 Martin Sebor <msebor@redhat.com>
1404 PR tree-optimization/85753
1405 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1406 RECORD_TYPE in addition to ARRAY_TYPE.
1408 2018-05-15 Martin Sebor <msebor@redhat.com>
1411 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1413 2018-05-15 Richard Biener <rguenther@suse.de>
1415 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1416 add by_clobber_p one. Change algorithm to collect all defs
1417 representing uses we need to walk and try reducing them to
1418 a single one before failing.
1419 (dse_dom_walker::dse_optimize_stmt): Adjust.
1421 2018-05-13 Mark Wielaard <mark@klomp.org>
1423 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1424 (size_of_loc_descr): Likewise.
1425 (output_loc_operands): Likewise.
1426 (output_loc_operands_raw): Likewise.
1427 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1428 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1429 (hash_loc_operands): Likewise.
1430 (compare_loc_operands): Likewise.
1432 2018-05-14 Mark Wielaard <mark@klomp.org>
1434 * dwarf2out.c (count_index_addrs): New function.
1435 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1437 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1439 PR tree-optimization/83648
1440 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1441 return value as malloc candidate.
1443 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1446 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1447 as true in call to suggest_attribute.
1449 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1451 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1452 -mreadonly-in-sdata.
1454 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1456 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1458 (aarch64_crypto_aesd_fused): Likewise.
1460 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1462 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1463 (movsi_aarch64): Likewise.
1464 (load_pairsi): Likewise.
1465 (load_pairdi): Likewise.
1466 (store_pairsi): Likewise.
1467 (store_pairdi): Likewise.
1468 (load_pairsf): Likewise.
1469 (load_pairdf): Likewise.
1470 (store_pairsf): Likewise.
1471 (store_pairdf): Likewise.
1472 (zero_extend): Likewise.
1473 (trunc): Swap alternatives.
1474 (fcvt_target): Add '?' to prefer w over r.
1476 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1479 * config/i386/i386.md: Disallow non-commutative arithmetics in
1480 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1481 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1482 in the peephole2 before it.
1484 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1486 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1487 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1488 (ix86_handle_option): Handle -mcldemote.
1489 * config.gcc: New header.
1490 * config/i386/cldemoteintrin.h: New file.
1491 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1492 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1494 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1495 OPTION_MASK_ISA_CLDEMOTE.
1496 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1497 (ix86_valid_target_attribute_inner_p): Ditto.
1498 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1499 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1500 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1501 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1502 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1504 * config/i386/i386.opt: Add -mcldemote.
1505 * config/i386/x86intrin.h: New header.
1506 * doc/invoke.texi: Add -mcldemote.
1508 2018-05-14 Richard Biener <rguenther@suse.de>
1510 * doc/match-and-simplify.texi: Adjust :s documentation.
1512 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1514 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1515 intended memcpy size.
1516 (REORDER_45): Likewise.
1518 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1520 * sort.cc: New file.
1521 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1522 * vec.c (qsort_chk): Use gcc_qsort.
1523 * Makefile.in (OBJS-libcommon): Add sort.o.
1524 (build/sort.o): New target. Use it...
1525 (BUILD_RTL): ... here, and...
1526 (build/gencfn-macros): ... here, and...
1527 (build/genmatch): ... here.
1529 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1530 Chung-Ju Wu <jasonwucj@gmail.com>
1532 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1533 * config/nds32/nds32-graywolf.md: New file.
1534 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1535 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1537 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1538 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1539 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1540 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1541 * config/nds32/pipelines.md: Include n15 settings.
1543 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1544 Chung-Ju Wu <jasonwucj@gmail.com>
1546 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1547 * config/nds32/nds32-n13.md: New file.
1548 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1549 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1551 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1552 * config/nds32/nds32.md (pipeline_model): Add n13.
1553 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1554 * config/nds32/pipelines.md: Include n13 settings.
1556 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1557 Chung-Ju Wu <jasonwucj@gmail.com>
1559 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1560 * config/nds32/nds32-n10.md: New file.
1561 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1562 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1564 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1565 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1566 * config/nds32/nds32.md (pipeline_model): Add n10.
1567 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1568 * config/nds32/pipelines.md: Include n10 settings.
1570 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1571 Kito Cheng <kito.cheng@gmail.com>
1572 Chung-Ju Wu <jasonwucj@gmail.com>
1574 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1575 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1576 Add enum values for DSP extension instructions.
1577 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1579 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1580 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1582 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1583 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1584 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1585 * config/nds32/nds32-intrinsic.md: Likewise.
1586 * config/nds32/nds32_intrinsic.h: Likewise.
1587 * config/nds32/nds32-md-auxiliary.c: Likewise.
1588 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1589 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1590 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1591 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1592 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1593 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1594 TYPE_DMAC in switch statement.
1595 * config/nds32/nds32.c: New checking and implementation for DSP
1596 extension instructions.
1597 * config/nds32/nds32.h: Likewise.
1598 * config/nds32/nds32.md: Likewise.
1599 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1600 * config/nds32/predicates.md: Implement new predicates for DSP
1603 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1605 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1606 Reformat alternatives and attributes so it is easier to identify
1607 which constraints/attributes go with which instruction.
1608 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1609 (mov<mode>_softfloat32, FMOVE64): Likewise.
1610 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1611 (mov<mode>_softfloat64, FMOVE64): Likewise.
1613 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1615 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1617 (Basic PowerPC Built-in Functions): The new name of the
1618 subsection previously known as "PowerPC Built-in Functions".
1619 (Basic PowerPC Built-in Functions Available on all Configurations):
1621 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1622 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1623 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1624 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1626 2018-05-11 Martin Jambor <mjambor@suse.cz>
1629 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1632 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1635 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1637 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1639 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1640 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1641 (ix86_handle_option): Handle -mwaitpkg.
1642 * config.gcc: New header.
1643 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1644 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1645 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1647 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1648 OPTION_MASK_ISA_WAITPKG.
1649 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1650 (ix86_option_override_internal): Add PTA_WAITPKG.
1651 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1652 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1653 IX86_BUILTIN_TPAUSE.
1654 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1655 __builtin_ia32_umwait and __builtin_ia32_tpause.
1656 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1657 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1658 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1659 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1660 UNSPECV_TPAUSE): New.
1661 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1662 * config/i386/i386.opt: Add -mwaitpkg.
1663 * config/i386/waitpkgintrin.h: New file.
1664 * config/i386/x86intrin.h: New header.
1665 * doc/invoke.texi: Add -mwaitpkg.
1667 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1670 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1672 (cortex-m0): Use armv6s-m isa.
1673 (cortex-m0plus): Likewise.
1674 (cortex-m1): Likewise.
1675 (cortex-m0.small-multiply): Likewise.
1676 (cortex-m0plus.small-multiply): Likewise.
1677 (cortex-m1.small-multiply): Likewise.
1679 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1680 Jakub Jelinek <jakub@redhat.com>
1682 PR tree-optimization/85692
1683 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1684 source permute as well.
1686 2018-05-11 Martin Liska <mliska@suse.cz>
1689 * doc/extend.texi: Document LLVM style format for no_sanitize
1692 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1694 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1695 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1696 (mode_supports_vsx_dform_quad): Likewise.
1697 (mode_supports_vmx_dform): Move these functions to be next to the
1698 other mode_supports functions.
1699 (mode_supports_dq_form): Likewise.
1700 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1701 mode_supports_dq_form.
1702 (reg_offset_addressing_ok_p): Likewise.
1703 (offsettable_ok_by_alignment): Likewise.
1704 (rs6000_legitimate_offset_address_p): Likewise.
1705 (legitimate_lo_sum_address_p): Likewise.
1706 (rs6000_legitimize_address): Likewise.
1707 (rs6000_legitimize_reload_address): Likewise.
1708 (rs6000_secondary_reload_inner): Likewise.
1709 (rs6000_preferred_reload_class): Likewise.
1710 (rs6000_output_move_128bit): Likewise.
1712 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1714 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1715 Generate SImode target register for null target.
1716 <case IX86_BUILTIN_XGETBV>: Ditto.
1717 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1718 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1720 2018-05-10 Carl Love <cel@us.ibm.com>
1722 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1723 dcbtt and dcbtstt if operands[2] is 0.
1725 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1728 * config/i386/sse.md (usadv64qi): New expander.
1730 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1732 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1733 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1734 -maltivec=be support.
1735 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1736 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1737 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1738 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1739 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1740 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1741 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1742 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1743 altivec_vsumsws): Adjust.
1744 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1745 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1746 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1748 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1749 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1750 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1751 (altivec_lve<VI_char>x): Delete expand.
1752 (*altivec_lve<VI_char>x_internal): Rename to...
1753 (altivec_lve<VI_char>x): ... this.
1754 (altivec_lvxl_<mode>): Delete expand.
1755 (*altivec_lvxl_<mode>_internal): Rename to ...
1756 (altivec_lvxl_<mode>): ... this.
1757 (altivec_stvxl_<mode>): Delete expand.
1758 (*altivec_stvxl_<mode>_internal): Rename to ...
1759 (altivec_stvxl_<mode>): ... this.
1760 (altivec_stve<VI_char>x): Delete expand.
1761 (*altivec_stve<VI_char>x_internal): Rename to ...
1762 (altivec_stve<VI_char>x): ... this.
1763 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1764 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1765 reduc_plus_scal_<mode>): Adjust.
1766 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1768 (rs6000_cpu_cpp_builtins): Adjust.
1769 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1770 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1771 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1772 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1773 -maltivec=be support.
1774 (rs6000_split_vec_extract_var): Adjust.
1775 (rs6000_split_v4si_init): Adjust.
1776 (swap_selector_for_mode): Delete.
1777 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1778 altivec_expand_stvex_be): Delete.
1779 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1780 -maltivec=be support.
1781 (rs6000_gimple_fold_builtin): Ditto.
1782 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1784 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1785 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1786 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1787 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1788 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1789 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1790 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1791 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1792 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1793 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1794 anonymous split): Adjust.
1795 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1796 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1798 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1800 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1801 when --with-gxx-include-dir is also specified.
1802 * configure: Regenerate.
1804 2018-05-09 Jim Wilson <jimw@sifive.com>
1807 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1808 * config/riscv/t-withmultilib: New.
1809 * config/riscv/withmultilib.h: New.
1810 * doc/install.texi: Document RISC-V --with-multilib-list support.
1812 2018-05-09 Richard Biener <rguenther@suse.de>
1814 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1816 (vect_bb_vectorization_profitable_p): Adjust. Compute
1817 actual scalar cost using the cost vector and the add_stmt_cost
1820 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1822 PR rtl-optimization/85645
1823 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1824 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1826 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1828 PR rtl-optimization/85645
1829 * shrink-wrap.c (spread_components): Return a boolean saying if
1830 anything was changed.
1831 (try_shrink_wrapping_separate): Iterate spread_components until
1832 nothing changes anymore.
1834 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1836 PR rtl-optimization/85645
1837 * regrename.c (build_def_use): Also kill the chains that include the
1838 destination of a REG_CFA_REGISTER note.
1840 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1842 PR rtl-optimization/85645
1843 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1844 insn that has a REG_CFA_REGISTER note.
1846 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1848 * cfgexpand.c (expand_clobber): New function.
1849 (expand_gimple_stmt_1): Use it.
1850 * tree-vect-stmts.c (vect_clobber_variable): New function,
1852 (vectorizable_simd_clone_call): ...here.
1853 (vectorizable_store): Emit a clobber either side of an
1854 IFN_STORE_LANES sequence.
1855 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1857 2018-05-09 Tom de Vries <tom@codesourcery.com>
1860 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1861 (define_insn "trap_if_false"): Add exit after trap.
1863 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1865 PR rtl-optimization/85638
1866 * bb-reorder.c: Include common/common-target.h.
1867 (create_forwarder_block): New function extracted from...
1868 (fix_up_crossing_landing_pad): ...here. Rename into...
1869 (dw2_fix_up_crossing_landing_pad): ...this.
1870 (sjlj_fix_up_crossing_landing_pad): New function.
1871 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1872 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1873 from both partitions and exit the loop after one iteration.
1875 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1878 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1880 (Basic PowerPC Built-in Functions): The new name of the
1881 subsection previously known as "PowerPC Built-in Functions".
1882 (Basic PowerPC Built-in Functions Available on all Configurations):
1884 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1886 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1888 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1890 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1893 2018-05-08 Jim Wilson <jimw@sifive.com>
1895 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1896 (LD_EMUL_SUFFIX): New.
1897 (LINK_SPEC): Use it.
1899 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1901 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1903 (Basic PowerPC Built-in Functions): The new name of the
1904 subsection previously known as "PowerPC Built-in Functions".
1905 (Basic PowerPC Built-in Functions Available on all Configurations):
1907 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1909 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1911 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1913 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1916 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1919 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1920 after cmpelim optimization.
1922 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1924 * config.gcc: Support "goldmont".
1925 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1926 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1928 * config/i386/i386.c (m_GOLDMONT): Define.
1929 (processor_target_table): Add "goldmont".
1930 (PTA_GOLDMONT): Define.
1931 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1932 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1933 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1934 (fold_builtin_cpu): Add "goldmont".
1935 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1936 (ix86_option_override_internal): Add "goldmont".
1937 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1938 (processor_type): Add PROCESSOR_GOLDMONT.
1939 * config/i386/i386.md: Add CPU "glm".
1940 * config/i386/glm.md: New file.
1941 * config/i386/x86-tune.def: Add m_GOLDMONT.
1942 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1944 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1947 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1949 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1950 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1951 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1954 * config/i386/i386.c (ix86_fold_builtin): Handle
1955 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1958 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1959 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1961 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1964 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1965 (check_arch): Likewise.
1966 (check_fpu): Return the result rather than printing it.
1967 (end arch): Fix operator precedence.
1968 (end cpu): Likewise.
1969 (END): Print the result from check_fpu.
1971 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1972 Alan Hayward <alan.hayward@arm.com>
1973 David Sherwood <david.sherwood@arm.com>
1975 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1976 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1977 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1978 (*fcmuo<mode>_and): New patterns.
1980 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1982 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1983 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1984 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1985 (cmp_op, sve_imm_con): New code attributes.
1986 (SVE_COND_INT_CMP, imm_con): Delete.
1987 (cmp_op): Remove above unspecs from int attribute.
1988 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1990 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1991 comparison-specific unspecs.
1992 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1993 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1994 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1995 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1996 (*vec_fcm<cmp_op><mode>): Rename to...
1997 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1998 (*vec_fcmuo<mode>): Rename to...
1999 (*fcmuo<mode>): ...this and adjust likewise.
2000 (*pred_fcm<cmp_op><mode>): New pattern.
2001 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2002 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2004 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2006 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2007 (aarch64_emit_sve_predicated_cond): New function.
2008 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2009 (aarch64_emit_unspec_cond_or): Replace with...
2010 (aarch64_emit_sve_or_conds): ...this new function. Use
2011 aarch64_emit_sve_ptrue_op for the individual comparisons and
2012 aarch64_emit_binop to OR them together.
2013 (aarch64_emit_inverted_unspec_cond): Replace with...
2014 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2015 aarch64_emit_sve_ptrue_op for the comparison and
2016 aarch64_emit_unop to invert the result.
2017 (aarch64_expand_sve_vec_cmp_float): Update after the above
2018 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2020 2018-05-07 Nathan Sidwell <nathan@acm.org>
2022 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2023 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2024 (Backwards Compatibility): Likewise.
2026 2018-05-07 Luis Machado <luis.machado@linaro.org>
2030 2018-05-07 Luis Machado <luis.machado@linaro.org>
2032 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2033 <prefetch_dynamic_strides>: New const bool field.
2034 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2035 prefetch_dynamic_strides.
2036 (exynosm1_prefetch_tune): Likewise.
2037 (thunderxt88_prefetch_tune): Likewise.
2038 (thunderx_prefetch_tune): Likewise.
2039 (thunderx2t99_prefetch_tune): Likewise.
2040 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2041 (aarch64_override_options_internal): Update to set
2042 PARAM_PREFETCH_DYNAMIC_STRIDES.
2043 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2044 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2045 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2046 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2047 prefetch-dynamic-strides setting.
2049 2018-05-07 Luis Machado <luis.machado@linaro.org>
2051 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2052 <minimum_stride>: New const int field.
2053 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2054 minimum_stride field.
2055 (exynosm1_prefetch_tune): Likewise.
2056 (thunderxt88_prefetch_tune): Likewise.
2057 (thunderx_prefetch_tune): Likewise.
2058 (thunderx2t99_prefetch_tune): Likewise.
2059 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2060 (aarch64_override_options_internal): Update to set
2061 PARAM_PREFETCH_MINIMUM_STRIDE.
2062 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2063 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2064 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2065 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2066 stride is constant and is below the minimum stride threshold.
2068 2018-05-07 Luis Machado <luis.machado@linaro.org>
2070 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2073 2018-05-07 Luis Machado <luis.machado@linaro.org>
2075 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2076 <prefetch_dynamic_strides>: New const bool field.
2077 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2078 prefetch_dynamic_strides.
2079 (exynosm1_prefetch_tune): Likewise.
2080 (thunderxt88_prefetch_tune): Likewise.
2081 (thunderx_prefetch_tune): Likewise.
2082 (thunderx2t99_prefetch_tune): Likewise.
2083 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2084 (aarch64_override_options_internal): Update to set
2085 PARAM_PREFETCH_DYNAMIC_STRIDES.
2086 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2087 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2088 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2089 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2090 prefetch-dynamic-strides setting.
2092 2018-05-07 Luis Machado <luis.machado@linaro.org>
2094 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2095 <minimum_stride>: New const int field.
2096 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2097 minimum_stride field.
2098 (exynosm1_prefetch_tune): Likewise.
2099 (thunderxt88_prefetch_tune): Likewise.
2100 (thunderx_prefetch_tune): Likewise.
2101 (thunderx2t99_prefetch_tune): Likewise.
2102 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2103 (aarch64_override_options_internal): Update to set
2104 PARAM_PREFETCH_MINIMUM_STRIDE.
2105 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2106 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2107 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2108 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2109 stride is constant and is below the minimum stride threshold.
2111 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2114 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2115 the type is addressable. Don't force op into register if it has
2118 2018-05-05 Roland McGrath <mcgrathr@google.com>
2121 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2122 any section for which we don't know a specific type it should have,
2123 regardless of name. Previously this was done only for the exact
2124 names ".init_array", ".fini_array", and ".preinit_array".
2125 (default_elf_asm_named_section): Add comment about
2126 relationship with default_section_type_flags and SECTION_NOTYPE.
2127 (get_section): Don't consider it a type conflict if one side has
2128 SECTION_NOTYPE and the other doesn't, as long as neither has the
2129 SECTION_BSS et al used in the default_section_type_flags logic.
2131 2018-05-05 Tom de Vries <tom@codesourcery.com>
2134 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2135 (workaround_barsyncs): New function.
2136 (nvptx_reorg): Use workaround_barsyncs.
2137 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2138 (define_expand "nvptx_membar_cta"): New define_expand.
2139 (define_insn "*nvptx_membar_cta"): New insn.
2141 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2143 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2144 To improve optimization opportunities.
2145 * builtin-types.def: The new needed builtin types for the above.
2147 2018-05-04 Richard Biener <rguenther@suse.de>
2149 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2150 * gimple-ssa-store-merging.c
2151 (imm_store_chain_info::output_merged_store): Remove redundant create,
2152 release split_store vector contents on failure.
2153 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2154 scalar stmt vector on cache hit.
2156 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2158 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2160 * config.gcc (powerpc-xilinx-eabi*): Remove.
2161 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2163 (fusion_addis_mem_combo_load): Ditto.
2164 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2166 (rs6000_cpu_cpp_builtins): Ditto.
2167 * config/rs6000/rs6000-linux.c
2168 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2169 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2170 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2172 (rs6000_setup_reg_addr_masks): Ditto.
2173 (rs6000_init_hard_regno_mode_ok): Ditto.
2174 (rs6000_option_override_internal): Ditto.
2175 (legitimate_lo_sum_address_p): Ditto.
2176 (rs6000_legitimize_address): Ditto.
2177 (rs6000_legitimize_reload_address): Ditto.
2178 (rs6000_legitimate_address_p): Ditto.
2179 (abi_v4_pass_in_fpr): Ditto.
2180 (setup_incoming_varargs): Ditto.
2181 (rs6000_gimplify_va_arg): Ditto.
2182 (rs6000_split_multireg_move): Ditto.
2183 (rs6000_savres_strategy): Ditto.
2184 (rs6000_emit_prologue_components): Ditto.
2185 (rs6000_emit_epilogue_components): Ditto.
2186 (rs6000_emit_prologue): Ditto.
2187 (rs6000_emit_epilogue): Ditto.
2188 (rs6000_elf_file_end): Ditto.
2189 (rs6000_function_value): Ditto.
2190 (rs6000_libcall_value): Ditto.
2191 * config/rs6000/rs6000.h: Ditto.
2192 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2193 (TARGET_MINMAX): ... this. New.
2194 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2195 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2196 (*movsi_internal1_single): Delete.
2197 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2198 mfpu=, mxilinx-fpu): Delete.
2199 * config/rs6000/singlefp.h: Delete.
2200 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2201 * config/rs6000/t-rs6000: Ditto.
2202 * config/rs6000/t-xilinx: Delete.
2203 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2204 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2205 (VStype_simple): Delete.
2206 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2207 * config/rs6000/xfpu.h: Delete.
2208 * config/rs6000/xfpu.md: Delete.
2209 * config/rs6000/xilinx.h: Delete.
2210 * config/rs6000/xilinx.opt: Delete.
2211 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2212 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2214 2018-05-04 Tom de Vries <tom@codesourcery.com>
2217 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2220 2018-05-04 Richard Biener <rguenther@suse.de>
2223 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2224 (expand_complex_div_wide): Likewise.
2225 (expand_complex_operations_1): Likewise.
2226 (expand_complex_libcall): Preserve EH info of the original stmt.
2227 (tree_lower_complex): Handle removed blocks.
2228 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2229 on complex multiplication and division libcall builtins.
2231 2018-05-04 Richard Biener <rguenther@suse.de>
2234 * fold-const.c (negate_expr_p): Restrict negation of operand
2235 zero of a division to when we know that can happen without
2237 (fold_negate_expr_1): Likewise.
2239 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2242 * real.h (real_nextafter): Declare.
2243 * real.c (real_nextafter): New function.
2244 * fold-const-call.c (fold_const_nextafter): New function.
2245 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2246 CASE_CFN_NEXTTOWARD.
2247 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2248 even when arg1_mode is different from arg0_mode.
2250 2018-05-03 Nathan Sidwell <nathan@acm.org>
2252 * doc/extend.texi (Deprecated Features): Remove
2254 (Backwards Compatibility): Likewise.
2255 * doc/invoke.texi (C++ Language Options): Likewise.
2256 (C++ Dialect Options): Likewise.
2258 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2261 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2262 _mm512_mask_mullox_epi64): New intrinsics.
2264 2018-05-03 Tom de Vries <tom@codesourcery.com>
2267 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2268 dump files): Add offload-tree.
2270 2018-05-03 Richard Biener <rguenther@suse.de>
2272 PR tree-optimization/85615
2273 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2274 to loops not nested in BBs loop father to avoid creating multi-entry
2277 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2279 PR tree-optimization/70291
2280 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2281 arguments. Change return type to tree. Emit libcall as a new
2282 statement rather than replacing existing one when inplace_p is true.
2283 (expand_complex_multiplication_components): New function.
2284 (expand_complex_multiplication): Expand floating-point complex
2285 multiplication using the above.
2286 (expand_complex_division): Rename inner_type parameter to type.
2287 Update expand_complex_libcall call-site.
2288 (expand_complex_operations_1): Update expand_complex_multiplication
2289 and expand_complex_division call-sites.
2291 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2294 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2295 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2296 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2297 the highest significant bit of the shift count mask is clear. In
2298 check whether and[sq]i3 is needed verify that all significant bits
2299 of the shift count other than the highest are set.
2301 2018-05-02 Tom de Vries <tom@codesourcery.com>
2304 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2305 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2306 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2307 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2308 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2309 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2310 __builtin_goacc_parlevel_size.
2312 2018-05-02 Richard Biener <rguenther@suse.de>
2314 PR tree-optimization/85597
2315 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2316 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2319 2018-05-02 Tom de Vries <tom@codesourcery.com>
2322 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2323 dump files): Add ltrans-tree.
2325 2018-05-02 Tom de Vries <tom@codesourcery.com>
2328 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2329 dump files): Add wpa-ipa.
2331 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2333 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2334 powerpc*-*-linux*paired* target.
2335 * config/rs6000/750cl.h: Delete.
2336 * config/rs6000/paired.h: Delete.
2337 * config/rs6000/paired.md: Delete.
2338 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2340 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2341 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2342 comment. Remove paired float support.
2343 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2344 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2346 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2347 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2349 * config/rs6000/rs6000.c: Remove paired float support.
2350 (paired_expand_vector_init, paired_expand_vector_move,
2351 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2352 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2353 paired_expand_builtin, paired_expand_predicate_builtin,
2354 paired_init_builtins): Delete.
2355 * config/rs6000/rs6000.h: Remove paired float support.
2356 * config/rs6000/rs6000.md: Remove paired float support.
2357 (move_from_CR_ov_bit): Delete.
2358 * config/rs6000/rs6000.opt (mpaired): Delete.
2359 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2360 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2362 2018-05-02 Richard Biener <rguenther@suse.de>
2365 * gimplify.c (gimplify_save_expr): When in SSA form allow
2366 SAVE_EXPRs to compute to SSA vars.
2368 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2371 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2372 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2373 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2374 clobber operands[2], instead use a new pseudo. Formatting fixes.
2376 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2378 PR tree-optimization/85586
2379 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2380 exit early for statements in the same group if the accesses are
2383 2018-05-02 Tom de Vries <tom@codesourcery.com>
2386 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2389 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2391 PR tree-optimization/85143
2392 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2394 2018-05-01 Tom de Vries <tom@codesourcery.com>
2397 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2398 not found" error message.
2400 2018-05-01 Tom de Vries <tom@codesourcery.com>
2403 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2404 * vec.c (test_ordered_remove_if): New function.
2405 (vec_c_tests): Call test_ordered_remove_if.
2406 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2407 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2408 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2409 VEC_ORDERED_REMOVE_IF.
2411 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2413 PR tree-optimization/82665
2414 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2415 pointer subtraction where arguments come from a memchr call.
2417 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2419 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2420 --push-state --as-needed and --pop-state instead of --as-needed and
2421 --no-as-needed if ld supports it.
2422 * configure: Regenerated.
2425 * doc/install.texi2html: Replace _002d with - and _002a with * in
2426 generated html files using sed.
2428 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2431 * gcc-rich-location.c (blank_line_before_p): New function.
2432 (use_new_line): New function.
2433 (gcc_rich_location::add_fixit_insert_formatted): New function.
2434 * gcc-rich-location.h
2435 (gcc_rich_location::add_fixit_insert_formatted): New function.
2437 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2439 * selftest.c (assert_streq): Rename "expected" and "actual" to
2440 "val1" and "val2". Extend NULL-handling to cover both inputs
2441 symmetrically, while still requiring both to be non-NULL for a pass.
2442 * selftest.h (assert_streq): Rename "expected" and "actual" to
2444 (ASSERT_EQ): Likewise.
2445 (ASSERT_EQ_AT): Likewise.
2446 (ASSERT_KNOWN_EQ): Likewise.
2447 (ASSERT_KNOWN_EQ_AT): Likewise.
2448 (ASSERT_NE): Likewise.
2449 (ASSERT_MAYBE_NE): Likewise.
2450 (ASSERT_MAYBE_NE_AT): Likewise.
2451 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2452 the assertion to pass.
2453 (ASSERT_STREQ_AT): Likewise.
2455 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2457 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2458 interaction with -pie.
2460 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2462 * selftest.h: Fix alphabetization of per-source-file selftest
2465 2018-04-30 Jason Merrill <jason@redhat.com>
2467 PR c++/61982 - dead stores to destroyed objects.
2468 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2471 2018-04-30 Jason Merrill <jason@redhat.com>
2473 * tree.c (build_clobber): New.
2474 * tree.h: Declare it.
2475 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2477 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2479 * diagnostic-show-locus.c (layout::layout): Update for
2480 location_get_source_line returning a char_span.
2481 (struct char_span): Move to input.h.
2482 (struct correction): Update for fields in char_span becoming
2484 (struct source_line): Update for location_get_source_line
2485 returning a char_span.
2486 (layout::print_line): Likewise.
2487 * edit-context.c (edited_file::print_content): Likewise.
2488 (edited_file::print_diff_hunk): Likewise.
2489 (edited_file::print_run_of_changed_lines): Likewise.
2490 (edited_file::get_num_lines): Likewise.
2491 (edited_line::edited_line): Likewise.
2492 * final.c (asm_show_source): Likewise.
2493 * input.c (location_get_source_line): Convert return type
2494 from const char * to char_span, losing the final "line_len"
2496 (dump_location_info): Update for the above.
2497 (get_substring_ranges_for_loc): Likewise. Use a char_span
2498 when handling the literal within the line.
2499 (test_reading_source_line): Update for location_get_source_line
2500 returning a char_span.
2501 * input.h (class char_span): Move here from
2502 diagnostic-show-locus.c, converting from a struct to a class.
2503 Make data members private.
2504 (char_span::operator bool): New.
2505 (char_span::length): New.
2506 (char_span::get_buffer): New.
2507 (char_span::operator[]): New.
2508 (char_span::subspan): Make const.
2509 (char_span::xstrdup): New.
2510 (location_get_source_line): Convert return type from const char *
2511 to char_span, losing the final "line_size" param.
2513 2018-04-30 Jan Hubicka <jh@suse.cz>
2515 * lto-wrapper.c (ltrans_priorities): New static var.
2516 (cmp_priority): New.
2517 (run_gcc): Read priorities and if doing parallel build order
2518 the Makefile by them.
2520 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2522 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2524 2018-04-30 Richard Biener <rguenther@suse.de>
2526 * tree-cfg.c (verify_address): Remove base argument, add
2527 flag whether to check TREE_ADDRESSABLE and do that.
2528 (verify_expr): Remove.
2529 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2530 (verify_gimple_assign_single): Likewise.
2531 (verify_gimple_switch): Likewise.
2532 (verify_expr_location_1): Dereference tp once. Add (disabled)
2533 piece from verify_expr.
2534 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2536 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2538 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2540 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2542 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2543 (small_data_pattern): Likewise.
2544 (arc_rewrite_small_data): Likewise.
2545 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2546 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2547 (get_symbol_alignment): New function.
2548 (legitimate_small_data_address_p): Likewise.
2549 (legitimate_scaled_address): Update, call
2550 legitimate_small_data_address_p.
2551 (output_sdata): New static variable.
2552 (arc_print_operand): Update how we handle small data operands.
2553 (arc_print_operand_address): Likewise.
2554 (arc_legitimate_address_p): Update, use
2555 legitimate_small_data_address_p.
2556 (arc_rewrite_small_data_p): Remove.
2557 (arc_rewrite_small_data_1): Likewise.
2558 (arc_rewrite_small_data): Likewise.
2559 (small_data_pattern): Likewise.
2560 (compact_sda_memory_operand): Update to use
2561 legitimate_small_data_address_p and get_symbol_alignment.
2562 (prepare_move_operands): Don't rewite sdata pattern.
2563 (prepare_extend_operands): Remove.
2564 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2566 (zero_extendqisi2): Likewise.
2567 (zero_extendhisi2): Likewise.
2568 (extendqihi2): Likewise.
2569 (extendqisi2): Likewise.
2570 (extendhisi2): Likewise.
2574 * config/arc/constraints.md (Usd): Change it to memory constraint.
2576 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2578 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2579 as source of std instructions.
2580 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2581 allow 6-bit constants as source for store instructions.
2582 (movdi_insn): Update instruction pattern to allow 6-bit constants
2583 as source for store instructions.
2585 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2587 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2589 2018-04-30 Nathan Sidwell <nathan@acm.org>
2590 Sandra Loosemore <sandra@codesourcery.com>
2592 * dumpfile.c (dump_open): Allow '-' for stdout.
2593 * doc/invoke.texi (Developer Options): Document dump filename
2594 determination early. Document stdin/stdout selection.
2596 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2598 Microblaze Target: PIC data text relative
2600 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2601 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2603 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2604 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2605 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2606 New addressing mode for data-text relative position indepenedent code.
2607 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2608 'ADDRESS_SYMBOLIC_TXT_REL'.
2609 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2610 (microblaze_legitimate_pic_operand): Exclude function calls from
2611 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2612 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2614 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2615 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2616 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2617 for 'address + offset'.
2618 (microblaze_expand_prologue): Add new function prologue call for
2620 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2621 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2622 table in case of TARGET_PIC_DATA_TEXT_REL.
2623 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2624 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2625 Add new macros 'UNSPEC_TEXT',
2626 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2627 + exclude function calls from 'UNSPEC_PLT' in case of data text
2629 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2630 new target hook for generating address diff vector tables in case of
2632 * doc/tm.texi : Regenerate.
2633 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2634 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2635 of addr diff vector generation.
2636 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2637 target hook definition.
2638 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2639 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2640 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2641 option for data text relative.
2643 2018-04-30 Richard Biener <rguenther@suse.de>
2645 * tree-chrec.h (evolution_function_is_constant_p): Remove
2647 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2649 2018-04-30 Richard Biener <rguenther@suse.de>
2652 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2654 2018-04-30 Richard Biener <rguenther@suse.de>
2656 PR tree-optimization/28364
2657 PR tree-optimization/85275
2658 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2659 copying first exit test.
2661 2018-04-28 Mark Wielaard <mark@klomp.org>
2663 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2665 (dwarf_AT): Handle DW_AT_addr_base.
2666 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2668 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2671 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2672 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2673 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2674 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2676 2018-04-28 Richard Biener <rguenther@suse.de>
2678 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2679 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2680 to reflect use. Only add interesting stmts.
2682 2018-04-27 Martin Jambor <mjambor@suse.cz>
2685 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2686 the jump function allows for passing through aggregate values.
2688 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2690 * input.h (in_system_header_at): Convert from macro to inline
2692 (from_macro_expansion_at): Likewise.
2693 (from_macro_definition_at): Likewise.
2695 2018-04-27 Jeff Law <law@redhat.com>
2697 * config.gcc: Mark tile* targets as deprecated/obsolete.
2699 2018-04-27 Richard Biener <rguenther@suse.de>
2701 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2704 2018-04-27 Richard Biener <rguenther@suse.de>
2706 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2708 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2710 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2711 with Yd constraint. Set "preferred_for_speed" attribute from
2712 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2714 (*movdi_internal): Ditto.
2715 (movti_interunit splitters): Remove
2716 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2717 (movdi_interunit splitters): Ditto.
2718 * config/i386/constraints.md (Ye): Remove.
2719 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2721 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2724 * config/aarch64/constraints.md (Usg): Limit to 31.
2727 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2729 PR tree-optimization/85529
2730 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2731 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2732 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2733 zero extension or masking of the MSB bit.
2734 (optimize_range_tests): Add FIRST_BB argument, pass it through
2735 to optimize_range_tests_var_bound.
2736 (maybe_optimize_range_tests, reassociate_bb): Adjust
2737 optimize_range_tests callers.
2739 2018-04-26 Richard Biener <rguenther@suse.de>
2740 Jakub Jelinek <jakub@redhat.com>
2742 * cgraph.h (symbol_table): Just declare debug method here.
2743 * symtab.c (symbol_table::debug): Define.
2745 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2747 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2749 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2751 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2752 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2753 (*movdi_internal): Substitute Yi and Yj constraint with x
2754 and Ym and Yn constraint with y constraint. Update "isa"
2755 attribute and set "preferred_for_speed" attribute from
2756 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2757 (*movsi_internal): Ditto.
2758 (*movdf_internal): Ditto.
2759 (*movsf_internal): Ditto.
2760 (*zero_extendsidi2): Ditto.
2761 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2762 (sse2_loadld): Ditto.
2763 (*vec_extract<ssevecmodelower>_0): Ditto.
2764 (*vec_extractv4si_0_zext_sse4): Ditto.
2765 (vec_concatv2di): Ditto.
2766 (*vec_dup<mode>): Ditto.
2767 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2768 * config/i386/constraints.md (Yi): Remove.
2773 2018-04-26 Nathan Sidwell <nathan@acm.org>
2775 * dumpfile.c (dump_open): New.
2776 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2777 (dump_finish): Detect stdio/stderr by value not name.
2779 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2781 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2783 2018-04-26 Tom de Vries <tom@codesourcery.com>
2786 * config/nvptx/nvptx.c (verify_neutering_jumps)
2787 (verify_neutering_labels): New function
2788 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2790 2018-04-26 Tom de Vries <tom@codesourcery.com>
2793 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2794 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2797 2018-04-26 Richard Biener <rguenther@suse.de>
2798 Tom de Vries <tom@codesourcery.com>
2801 * lto-streamer-out.c (output_function): Fixup loops if required to match
2802 discovery done in the reader.
2804 2018-04-26 Richard Biener <rguenther@suse.de>
2806 PR tree-optimization/85116
2807 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2808 have a loop exit from the single latch predecessor. Remove
2809 case of header with just condition.
2810 (ch_base::copy_headers): Exclude infinite loops from any
2812 (pass_ch::execute): Record exits.
2814 2018-04-26 Richard Biener <rguenther@suse.de>
2816 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2817 prologue cost vector and pass it to vect_get_load_cost.
2818 (vect_get_peeling_costs_all_drs): Likewise.
2819 (vect_peeling_hash_get_lowest_cost): Likewise.
2820 (vect_enhance_data_refs_alignment): Likewise.
2822 2018-04-26 Richard Biener <rguenther@suse.de>
2825 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2826 checking of integer<->pointer conversions.
2827 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2828 sign-/zero-extending pointer types.
2829 (expand_omp_for_static_chunk): Likewise.
2831 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2832 Jean Lee <xiaoyur347@gmail.com>
2834 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2835 (TARGET_ASAN_SHADOW_OFFSET): Define.
2836 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2837 true for -fsanitize=address.
2839 2018-04-25 Mark Wielaard <mark@klomp.org>
2841 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2844 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2846 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2847 than "alu", remove explicit "memory" and "imm_disp" attributes.
2848 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2851 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2852 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2853 gen_lowpart_no_emit.
2855 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2858 * config/i386/i386.c (ix86_expand_builtin): Change memory
2859 operand to XI, extend p0 to Pmode.
2860 * config/i386/i386.md: Change unspec volatile and operand
2861 1 mode to XI, change operand 0 mode to P.
2863 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2865 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2866 GET_MODE_MASK before any checking.
2867 (nds32_can_use_bset_p): Likewise.
2868 (nds32_can_use_btgl_p): Likewise.
2870 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2872 * config/nds32/nds32-doubleword.md: New define_split pattern for
2873 illegal register number.
2875 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2877 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2879 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2881 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2883 2018-04-25 Richard Biener <rguenther@suse.de>
2885 * lto-streamer.h (LTO_major_version): Bump to 8.
2887 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2889 * BASE-VER: Set to 9.0.0.
2891 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2893 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2894 in __abskf2 and __powikf2.
2896 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2899 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2900 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2901 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2902 Use the above on operand 2. Reindent.
2903 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2905 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2908 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2910 * config/i386/i386.opt (mcet): Removed.
2911 * doc/install.texi: Remove -mcet documentation.
2912 * doc/invoke.texi: Likewise.
2914 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2917 * doc/install.texi: Remove -mcet from bootstrap-cet.
2919 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2922 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2923 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2927 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2928 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2929 containing a CONST_VECTOR.
2931 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2933 * doc/install.texi: Update newlib dependency for nvptx.
2935 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2938 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2939 instead of INTVAL when shifting x left.
2941 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2943 PR tree-optimization/85478
2944 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2945 vect_grouped_store_supported for single element vectors.
2947 2018-04-24 Richard Biener <rguenther@suse.de>
2950 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2951 load cost increase to the case of non-constant step.
2953 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2956 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2957 destination if any_malformed_asm.
2959 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2962 * expr.c (store_field): In the bitfield case, if the value comes from
2963 a function call and is returned in registers by means of a PARALLEL,
2964 do not change the mode of the temporary unless BLKmode and VOIDmode.
2966 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2968 PR rtl-optimization/85423
2969 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2970 dependencies to debug insns when the previous insn is non-debug.
2972 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2974 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2975 enums into a single definition.
2976 (fls): Fix predicates and printing.
2979 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2981 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2982 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2983 and short u6 immediate.
2984 (check_if_valid_sleep_operand): Remove.
2985 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2987 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2989 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2990 flag_always_save_lp condition.
2991 * config/nds32/nds32.opt (malways-save-lp): New option.
2993 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2995 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2996 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2997 * config/nds32/nds32.h
2998 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2999 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3001 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3003 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3004 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3006 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3007 Chung-Ju Wu <jasonwucj@gmail.com>
3009 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3011 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3012 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3014 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3016 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3018 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3020 * config/nds32/nds32-protos.h (nds32_data_alignment,
3021 nds32_local_alignment): Declare.
3022 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3023 nds32_local_alignment): New functions.
3024 (TARGET_CONSTANT_ALIGNMENT): Define.
3025 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3027 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3029 * config/nds32/nds32.c
3030 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3031 (TARGET_MODES_TIEABLE_P): Likewise.
3033 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3035 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3038 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3039 Chung-Ju Wu <jasonwucj@gmail.com>
3041 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3042 for unaligned access.
3043 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3045 * config/nds32/nds32-intrinsic.md: Likewise.
3046 * config/nds32/nds32_intrinsic.h: Likewise.
3047 * config/nds32/nds32.h (nds32_builtins): Likewise.
3048 * config/nds32/nds32.opt (munaligned-access): New option.
3049 * config/nds32/nds32.c (nds32_asm_file_start): Display
3050 flag_unaligned_access status.
3052 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3054 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3055 -mno-relax is present.
3056 * config/riscv/linux.h (LINK_SPEC): Ditto.
3058 2018-04-20 Martin Sebor <msebor@redhat.com>
3061 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3063 (gimple_fold_builtin_stxcpy_chk): Same.
3064 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3066 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3069 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3070 __powikf2 when long double is IEEE 128-bit.
3072 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3074 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3075 step to make sure stack always aligned.
3077 2018-04-20 Carl Love <cel@us.ibm.com>
3080 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3081 size check for arg0.
3083 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3084 Tom de Vries <tom@codesourcery.com>
3087 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3088 Emit insns for calls too.
3089 (nvptx_find_par): Always look for worker-level predecessor insn.
3090 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3092 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3093 (nvptx_process_pars): Propagate frames for calls.
3095 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3098 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3100 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3101 (ix86_handle_option): Don't handle OPT_mibt.
3102 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3104 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3106 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3107 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3108 (ix86_target_macros): Define __CET__ with flag_cf_protection
3109 for -fcf-protection.
3110 * config/i386/i386.c (isa2_opts): Remove -mibt.
3111 * config/i386/i386.h (TARGET_IBT): Removed.
3112 (TARGET_IBT_P): Likewise.
3113 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3114 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3115 * config/i386/i386.opt (mcet): Update help message.
3118 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3119 -mcet as an alias for -mshstk.
3121 2018-04-20 Richard Biener <rguenther@suse.de>
3124 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3125 complexity by forcing a single use of the multiply operand.
3127 2018-04-20 Martin Jambor <mjambor@suse.cz>
3130 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3131 recursion dependency to only apply to non-clones.
3133 2018-04-20 Martin Jambor <mjambor@suse.cz>
3136 * ipa-cp.c (create_specialized_node): Check that clones of
3137 self-recursive edges exist during IPA-CP.
3139 2018-04-19 Toon Moene <toon@moene.org>
3141 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3144 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3146 PR tree-optimization/85467
3147 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3148 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3149 VECTOR_CST element to type.
3151 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3154 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3155 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3156 (builtin_longjmp): Likewise.
3157 (save_stack_nonlocal): New pattern.
3158 (restore_stack_nonlocal): Likewise.
3160 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3163 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3164 Replace ASM_OUTPUT_LABEL with fprintf.
3166 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3169 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3170 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3171 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3172 define __IBT__ and __SHSTK__ for -fcf-protection.
3173 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3175 (ix86_trampoline_init): Likewise.
3176 (x86_output_mi_thunk): Likewise.
3177 (ix86_notrack_prefixed_insn_p): Likewise.
3178 (ix86_option_override_internal): Don't disallow -fcf-protection.
3179 * config/i386/i386.md (rdssp<mode>): Also enable for
3181 (incssp<mode>): Likewise.
3182 (nop_endbr): Likewise.
3183 * config/i386/i386.opt (mcet): Change help message to built-in
3187 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3188 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3189 enable CET built-in functions.
3191 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3193 * common/config/i386/i386-common.c
3194 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3195 OPTION_MASK_ISA_MOVDIRI_UNSET,
3196 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3197 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3198 * config.gcc (movdirintrin.h): New header.
3199 * config/i386/cpuid.h (bit_MOVDIRI,
3200 bit_MOVDIR64B): New bits.
3201 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3203 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3204 (VOID, PVOID, PCVOID)): New function types.
3205 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3206 __builtin_ia32_directstoreu_u64,
3207 __builtin_ia32_movdir64b): New builtins.
3208 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3209 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3211 (ix86_valid_target_attribute_inner_p): Ditto.
3212 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3213 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3214 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3215 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3216 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3217 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3218 (movdiri<mode>, movdir64b_<mode>): New.
3219 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3220 * config/i386/immintrin.h: Include movdirintrin.h.
3221 * config/i386/movdirintrin.h: New file.
3222 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3224 2018-04-19 Richard Biener <rguenther@suse.de>
3227 * cfg.c (clear_bb_flags): When loop state says we have
3228 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3230 2018-04-19 Richard Biener <rguenther@suse.de>
3232 PR tree-optimization/84737
3233 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3234 copying restrict info.
3235 (vect_setup_realignment): Use it.
3236 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3237 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3238 the first DR to all generated stores.
3239 (vectorizable_load): Likewise for loads.
3241 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3243 PR tree-optimization/85446
3244 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3245 the integral and pointer types to have the same precision.
3247 * doc/install.texi: Document --disable-cet being the default and
3250 2018-04-18 Martin Liska <mliska@suse.cz>
3252 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3255 2018-04-18 Martin Liska <mliska@suse.cz>
3258 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3261 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3262 arguments if they are comparable.
3264 2018-04-18 Martin Liska <mliska@suse.cz>
3267 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3270 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3273 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3276 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3277 ENDBR after calling __morestack.
3279 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3282 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3283 by using gcc_base_ver to generate a gcc_driver_version, and use
3284 it when generating GCC_DRIVER_NAME.
3285 * configure: Regenerate.
3287 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3290 * config.gcc: Obsolete powerpc*-*-*spe*.
3292 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3295 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3296 (stabstr_D): Change type of unum from unsigned int to
3297 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3300 2018-04-17 Jim Wilson <jimw@sifive.com>
3303 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3304 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3305 Set arg_pointer_offset after using pretend_args_size.
3307 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3309 PR rtl-optimization/85431
3310 * dse.c (record_store): Ignore zero width stores.
3313 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3314 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3315 __builtin_stack_restore rather than after it.
3316 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3317 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3318 argument instead of virtual_dynamic_stack_rtx.
3320 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3322 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3324 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3325 Add note to error message to explain internal mapping of overloaded
3326 built-in function name to non-overloaded built-in function name.
3327 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3330 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3333 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3334 where the inputs overlap with the output.
3336 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3338 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3339 (=v, v) alternative and explicit "memory" attribute.
3340 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3341 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3343 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3344 "sselog1" type instead of "sselog".
3345 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3346 "sselog". Remove explicit "memory" attribute.
3347 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3348 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3350 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3351 "sselog1" type instead of "sselog". Remove explicit "memory"
3353 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3354 use "sselog1" type instead of "sselog". Remove explicit "memory"
3356 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3357 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3359 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3360 "sselog1" type instead of "sselog". Remove explicit "memory"
3362 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3363 use "sselog1" type instead of "sselog". Remove explicit "memory"
3367 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3370 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3373 2018-04-17 Martin Jambor <mjambor@suse.cz>
3376 * ipa-cp.c (create_specialized_node): Call
3377 expand_all_artificial_thunks if necessary.
3379 2018-04-17 Martin Liska <mliska@suse.cz>
3382 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3383 in message, remote space in between '_G' and '('.
3385 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3388 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3389 avx512f_vmcmp<mode>3<round_saeonly_name>,
3390 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3391 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3392 avx512f_rndscale<mode><round_saeonly_name>,
3393 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3394 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3395 Use %<iptr>2 instead of %2 for -masm=intel.
3396 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3397 avx512f_vcvttss2usi<round_saeonly_name>,
3398 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3400 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3401 avx512f_vcvttsd2usi<round_saeonly_name>,
3402 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3403 Use %q1 instead of %1 for -masm=intel.
3404 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3405 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3406 of %3 for -masm=intel.
3407 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3409 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3411 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3413 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3414 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3416 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3418 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3419 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3420 %g1 and one with %0 and %1.
3421 (avx512er_vmrcp28<mode><round_saeonly_name>,
3422 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3424 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3425 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3426 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3427 of %0 and %{%4%} for -masm=intel.
3428 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3429 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3430 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3431 order of %0 and %{%5%}%{z%} for -masm=intel.
3433 2018-04-17 Jan Hubicka <jh@suse.cz>
3436 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3438 2018-04-17 Martin Liska <mliska@suse.cz>
3441 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3442 for target_clone error message. Make default implementation
3443 clone to be a local declaration.
3444 (separate_attrs): Add new argument and check for an empty
3446 (expand_target_clones): Handle it.
3447 (ipa_target_clone): Make redirection just for target_clones
3450 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3451 Tom de Vries <tom@codesourcery.com>
3454 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3455 tiled basic blocks without omp continue statements.
3457 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3460 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3461 vec_extract expression as having side effects to make sure it gets
3464 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3467 * config/i386/i386.c (get_builtin_code_for_version): Check
3470 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3473 * gcc/config.gcc: Support "skylake".
3474 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3476 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3477 (processor_target_table): Add "skylake".
3478 (ix86_option_override_internal): Add "skylake".
3479 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3480 PROCESSOR_CANNONLAKE.
3481 (get_builtin_code_for_version): Fix priority for
3482 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3483 PROCESSOR_SKYLAKE-AVX512.
3484 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3485 (processor_type): Add PROCESSOR_SKYLAKE.
3487 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3488 Jason Merrill <jason@redhat.com>
3491 * convert.c (convert_to_integer_1): Use direct recursion for
3492 enumeral types and types with a precision less than the number
3493 of bits in their mode.
3495 2018-04-16 Julia Koval <julia.koval@intel.com>
3498 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3499 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3501 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3504 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3505 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3506 and -mno-direct-move.
3508 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3511 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3512 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3513 Ensure negative shifts result in {0}.
3515 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3517 PR rtl-optimization/79916
3518 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3519 regs (if any) to define how to gnerate SD moves when LRA is in
3522 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3524 PR rtl-optimization/85393
3525 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3526 * except.c (expand_dw2_landing_pad_for_region): Make static.
3527 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3528 a label and unconditional jump to old_bb, rather than
3529 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3532 PR rtl-optimization/85376
3533 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3534 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3535 instead of a specific value.
3537 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3538 Bin Cheng <bin.cheng@arm.com>
3540 PR tree-optimization/82965
3541 PR tree-optimization/83991
3542 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3543 by_profile_only parameter.
3544 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3545 information if the loop was predicted to iterate too many times.
3546 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3548 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3551 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3554 2018-04-13 Martin Liska <mliska@suse.cz>
3555 Jakub Jelinek <jakub@redhat.com>
3558 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3559 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3560 * builtins.c (expand_builtin_memory_copy_args): Use
3561 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3562 handle dest_addr == pc_rtx.
3564 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3567 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3568 asked to not generate direct moves.
3569 (fix_trunc<mode>si2_stfiwx): Similar.
3570 (fix_trunc<mode>si2_internal): Similar.
3572 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3575 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3576 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3577 lookup if dest in some wider mode is known to be const0_rtx and
3578 if so, record permanent equivalence for it to be ZERO_EXTEND of
3579 the narrower mode destination.
3581 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3583 * lto-streamer-out.c (output_function): Revert 259346.
3584 * omp-expand.c (expand_oacc_for): Likewise.
3586 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3588 PR rtl-optimization/85354
3589 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3590 * sel-sched.c (sel_global_init): ... here.
3592 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3595 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3596 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3597 mode for PE-COFF targets.
3598 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3599 (i386_pe_asm_lto_end): Likewise.
3600 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3601 (TARGET_ASM_LTO_END): Likewise.
3602 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3603 (i386_pe_asm_lto_start): New function.
3604 (i386_pe_asm_lto_end): Likewise.
3606 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3607 Richard Biener <rguenther@suse.de>
3610 * lto-streamer-out.c (output_function): Fix CFG loop state before
3612 * omp-expand.c (expand_oacc_for): Handle calls to internal
3613 functions like regular functions.
3615 2018-04-12 Richard Biener <rguenther@suse.de>
3618 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3619 for the early LTO debug to properly generate references to it
3620 during DIE emission. Do not re-use that for the skeleton for
3622 (dwarf2out_early_finish): Likewise.
3624 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3627 * config/i386/sse.md
3628 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3629 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3630 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3631 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3632 and output is a reg, avoid creating invalid lowpart subreg, but
3633 instead split into a 512-bit move. Don't split if not AVX512VL,
3634 input is xmm16+ reg and output is a mem.
3635 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3636 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3637 xmm16+ reg and output is a mem.
3639 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3641 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3642 also for flag_dwarf2_cfi_asm.
3644 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3646 PR rtl-optimization/85342
3647 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3648 a bool scalar var inside of the loop instead. Don't try to update
3649 recog_data.operand after failed apply_change_group.
3651 2018-04-12 Tom de Vries <tom@codesourcery.com>
3654 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3655 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3656 array with flexible array member as array without given dimension.
3657 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3658 argument for undefined param to true.
3660 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3663 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3664 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3665 from PowerPC section.
3666 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3667 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3668 help text that is too long.
3669 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3670 help text that is too long.
3671 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3672 help text that is too long.
3674 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3676 * config/alpha/alpha.md (stack_probe_internal): Rename
3677 from "probe_stack". Update all callers.
3679 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3681 PR rtl-optimization/84566
3682 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3683 sched_macro_fuse_insns.
3685 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3688 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3689 (compute_block_dependences): ... from here.
3691 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3693 PR tree-optimization/85331
3694 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3695 from int to HOST_WIDE_INT.
3697 2018-04-11 Martin Jambor <mjambor@suse.cz>
3700 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3701 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3702 not the same as the source val.
3703 (cgraph_edge_brings_value_p): New parameter.
3704 (gather_edges_for_value): Pass destination value to
3705 cgraph_edge_brings_value_p.
3706 (perhaps_add_new_callers): Likewise.
3707 (get_info_about_necessary_edges): Likewise and exclude values brought
3708 only by self-recursive edges.
3709 (create_specialized_node): Redirect only clones of self-calling edges.
3710 (+self_recursive_pass_through_p): New function.
3711 (find_more_scalar_values_for_callers_subset): Use it.
3712 (find_aggregate_values_for_callers_subset): Likewise.
3713 (known_aggs_to_agg_replacement_list): Removed.
3714 (decide_whether_version_node): Re-calculate known constants for all
3715 remaining context clones.
3717 2018-04-11 Richard Biener <rguenther@suse.de>
3720 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3721 from early DWARF output.
3722 (dwarf2out_early_finish): Output line info unconditionally into
3723 early DWARF and add reference to it.
3725 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3728 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3729 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3730 other than V2DFmode using iptr mode attribute.
3731 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3733 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3735 PR rtl-optimization/84659
3736 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3738 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3741 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3743 (output_loc_list): Pass address of a dummy size variable even in the
3744 locview handling loop.
3745 (index_location_lists): Add comment on why skip_loc_list_entry can't
3748 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3751 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3754 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3757 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3758 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3759 and -mstring-compare-inline-limit.
3761 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3764 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3765 for stack clash protection in a register whenever we need it to be in
3768 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3770 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3771 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3773 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3776 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3778 (mlong-double-): Ditto.
3779 * config/rs6000/sysv4.opt (msdata=): Ditto.
3780 (mtls-size=): Ditto.
3782 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3784 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3785 erroneous entries for
3786 "vector int vec_ldl (int, long int *)", and
3787 "vector unsigned int vec_ldl (int, unsigned long int *)".
3788 Add comments and entries for
3789 "vector bool char vec_ldl (int, bool char *)",
3790 "vector bool short vec_ldl (int, bool short *)",
3791 "vector bool int vec_ldl (int, bool int *)",
3792 "vector bool long long vec_ldl (int, bool long long *)",
3793 "vector pixel vec_ldl (int, pixel *)",
3794 "vector long long vec_ldl (int, long long *)",
3795 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3796 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3797 type tree bool_long_long_type_node and correct definition of
3798 bool_V2DI_type_node to make reference to this new type tree.
3799 (rs6000_mangle_type): Replace erroneous reference to
3800 bool_long_type_node with bool_long_long_type_node.
3801 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3802 comments to emphasize sign distinctions for char and int types and
3803 replace RS6000_BTI_bool_long constant with
3804 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3805 use of RS6000_BTI_pixel.
3806 (bool_long_type_node): Remove this macro definition.
3807 (bool_long_long_type_node): New macro definition
3809 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3811 PR rtl-optimization/85300
3812 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3813 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3814 simplify_unary_operation fails.
3816 2018-04-10 Martin Liska <mliska@suse.cz>
3818 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3819 cgraph_edge and ipa_ref.
3821 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3825 * config/i386/sse.md
3826 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3827 computation of the VEC_MERGE selector from mask.
3828 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3829 Fix decoding of the VEC_MERGE selector into mask.
3831 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3833 PR tree-optimization/85286
3834 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3836 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3838 * final.c (final_1): Set insn_last_address as well as
3839 insn_current_address.
3841 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3844 * explow.c (emit_stack_probe): Call validize_mem on memory location
3845 before passing it to gen_probe_stack. Create address operand and
3846 legitimize it for the probe_stack_address case.
3848 2018-04-09 Jan Hubicka <jh@suse.cz>
3851 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3852 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3853 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3855 (free_lang_data): Rebuild type inheritance graph.
3857 2018-04-09 Martin Sebor <msebor@redhat.com>
3859 * invoke.texi (-finline-small-functions): Mention other optimization
3861 (-findirect-inlining, -fpartial-inlining): Same.
3862 (-finline-functions-called-once): Same.
3863 (-freorder-blocks-and-partition): Same.
3865 2018-04-09 Jan Hubicka <jh@suse.cz>
3868 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3869 jumps; choose last target that matches the criteria (i.e.
3870 no partition changes for non-crossing jumps).
3871 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3872 support for redirecting crossing jumps to non-crossing.
3874 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3876 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3877 also for naked functions.
3879 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3881 * config/arc/arc.md (add_shift): New pattern.
3882 (add_shift2): Likewise.
3883 (sub_shift): Likewise.
3884 (sub_shift_cmp0_noout): Likewise.
3885 (compare_si_ashiftsi): Likewise.
3886 (xbfu_cmp0_noout): New combine pattern.
3887 (xbfu_cmp0"): Likewise.
3888 (movsi_set_cc_insn): Place the predicable variant first.
3889 (commutative_binary_cmp0_noout): Remove clobber.
3890 (commutative_binary_cmp0): New pattern.
3891 (noncommutative_binary_cmp0): Likewise.
3892 (noncommutative_binary_cmp0_noout): Likewise.
3893 (noncommutative_binary_comparison_result_used): Removed.
3894 (rsub_cmp0): New pattern.
3895 (rsub_cmp0_noout): Likewise.
3896 (extzvsi): Changed, keep only meaningful variants.
3897 (SQH, SEZ): New iterators.
3898 (SQH_postfix): New mode attribute.
3899 (SEZ_prefix): New code attribute.
3900 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3901 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3902 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3904 (noncommutative_operator): Check the availability of barrel
3907 2018-04-09 Richard Biener <rguenther@suse.de>
3909 PR tree-optimization/85284
3910 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3911 Only use the niter constraining form of simple_iv when the exit
3914 2018-04-09 Tom de Vries <tom@codesourcery.com>
3917 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3918 (define_expand "*memory_barrier"): New define_expand.
3919 (define_insn "memory_barrier"): New insn.
3921 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3923 PR rtl-optimization/80463
3924 PR rtl-optimization/83972
3925 PR rtl-optimization/83480
3927 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3928 correct producer for the insn.
3929 (tidy_control_flow): Fixup seqnos in case of debug insns.
3931 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3933 PR rtl-optimization/83913
3935 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3936 different sched-times when merging exprs.
3938 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3940 PR rtl-optimization/83962
3942 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3943 tidy_fallthru_edge and tidy_control_flow.
3945 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3947 PR rtl-optimization/83530
3949 * sel-sched.c (force_next_insn): New global variable.
3950 (remove_insn_for_debug): When force_next_insn is true, also leave only
3951 next insn in the ready list.
3952 (sel_sched_region): When the region wasn't scheduled, make another pass
3953 over it with force_next_insn set to 1.
3955 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3957 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3959 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3960 for interrupt control.
3961 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3962 functions for interrupt control.
3963 * config/nds32/nds32-intrinsic.md: Likewise.
3964 * config/nds32/nds32_intrinsic.h: Likewise.
3965 * config/nds32/nds32.h (nds32_builtins): Likewise.
3967 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3969 * config/nds32/nds32.c (nds32_init_machine_status,
3970 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3971 strict_aligned_p field.
3972 (nds32_expand_to_rtl_hook): New function.
3973 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3974 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3976 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3977 Chung-Ju Wu <jasonwucj@gmail.com>
3979 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3980 * config/nds32/nds32-n7.md: New file.
3981 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3982 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3984 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3985 * config/nds32/nds32.md (pipeline_model): Add n7.
3986 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3987 * config/nds32/pipelines.md: Include n7 settings.
3989 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3990 Chung-Ju Wu <jasonwucj@gmail.com>
3992 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3993 * config/nds32/nds32-e8.md: New file.
3994 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3995 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3997 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3998 * config/nds32/nds32.md (pipeline_model): Add e8.
3999 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4000 * config/nds32/pipelines.md: Include e8 settings.
4002 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4003 Chung-Ju Wu <jasonwucj@gmail.com>
4005 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4006 * config/nds32/nds32-n8.md: New file.
4007 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4008 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4010 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4011 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4012 * config/nds32/nds32.md (pipeline_model): Add n8.
4013 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4014 * config/nds32/pipelines.md: Include n8 settings.
4016 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4017 Chung-Ju Wu <jasonwucj@gmail.com>
4019 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4020 * config/nds32/nds32-n9-2r1w.md: New file.
4021 * config/nds32/nds32-n9-3r2w.md: New file.
4022 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4023 nds32_register_ports): New or modify for cpu n9.
4024 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4026 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4027 * config/nds32/nds32-utils.c: New file.
4028 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4029 TARGET_MUL_SLOW): Define.
4030 * config/nds32/nds32.md (pipeline_model): New attribute.
4031 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4032 New options that support cpu n9.
4033 * config/nds32/pipelines.md: Include n9 settings.
4034 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4036 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4038 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4039 information if necessary.
4040 (output_cond_branch_compare_zero): Likewise.
4041 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4042 (nds32_target_alignment): Refine for alignment.
4043 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4044 (FUNCTION_BOUNDARY): Modify.
4045 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4047 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4049 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4051 * config/nds32/constants.md (unspec_volatile_element): Add values for
4052 TLB operation and data prefetch.
4053 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4054 functions for TLB operation and data prefetch.
4055 * config/nds32/nds32-intrinsic.md: Likewise.
4056 * config/nds32/nds32_intrinsic.h: Likewise.
4057 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4058 (nds32_print_operand): Likewise.
4059 * config/nds32/nds32.h (nds32_builtins): Likewise.
4061 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4062 Andrew Pinski <pinsika@gcc.gnu.org>
4065 * match.pd: Use constant_boolean_node of correct type instead of
4066 boolean_true_node or boolean_false_node for simplifying
4067 pointer comparisons to zero.
4069 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4071 PR tree-optimization/80021
4072 * tree.c (verify_type_variant): Make error call in verify_variant_match
4073 translatable and remove final full stop.
4075 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4077 * config/nds32/constants.md (unspec_volatile_element): Add
4078 UNSPEC_VOLATILE_EH_RETURN.
4079 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4080 nds32_output_stack_pop): Support dwarf exception handling process.
4081 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4082 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4083 exception handling process.
4084 (nds32_compute_stack_frame): Likewise.
4085 (nds32_return_addr_rtx): Likewise.
4086 (nds32_initial_elimination_offset): Likewise.
4087 (nds32_expand_prologue): Likewise.
4088 (nds32_expand_epilogue): Likewise.
4089 (nds32_dynamic_chain_address): New function.
4090 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4092 (DYNAMIC_CHAIN_ADDRESS): Define.
4093 (EH_RETURN_DATA_REGNO): Define.
4094 (EH_RETURN_STACKADJ_RTX): Define.
4095 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4096 patterns for dwarf exception handling.
4098 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4100 * config/nds32/nds32.h: Clean up obsolete macros.
4102 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4104 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4105 Add enum values for particular instructions.
4106 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4107 particular intrinsic functions.
4108 * config/nds32/nds32-intrinsic.md: Likewise.
4109 * config/nds32/nds32_intrinsic.h: Likewise.
4110 * config/nds32/nds32.h (nds32_builtins): Likewise.
4111 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4112 (btst, ave): New patterns for particular instructions.
4114 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4116 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4117 Add enum values for atomic load/store and memory sync.
4118 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4120 * config/nds32/nds32-intrinsic.md: Likewise.
4121 * config/nds32/nds32_intrinsic.h: Likewise.
4122 * config/nds32/nds32.h (nds32_builtins): Likewise.
4124 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4126 PR tree-optimization/85257
4127 * fold-const.c (native_encode_vector): If not all elts could fit
4128 and off is -1, return 0 rather than offset.
4129 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4130 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4131 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4132 adjust buffer in native_interpret_expr call.
4134 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4136 * config/nds32/constants.md (unspec_volatile_element): Add cache
4137 control enum values.
4138 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4139 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4140 * config/nds32/nds32.c (nds32_cctl_names): New.
4141 (nds32_print_operand): Handle cache control register names.
4142 * config/nds32/nds32.h (nds32_builtins): New enum values.
4143 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4145 * config/nds32/nds32.md (type): Add mmu.
4146 * config/nds32/pipelines.md (simple_insn): Add mmu.
4148 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4150 * config/nds32/nds32.md (type): Remove call.
4151 * config/nds32/pipelines.md (simple_insn): Likewise.
4153 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4155 * config/nds32/constants.md (unspec_volatile_element): Add
4156 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4157 UNSPEC_VOLATILE_FMFCFG.
4158 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4159 description for fmfcfg and fmfcsr.
4160 (bdesc_1arg): Add fmtcsr.
4161 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4162 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4163 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4164 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4165 unspec_fmfcfg): New patterns.
4166 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4167 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4168 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4169 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4170 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4171 __nds32__fmfcfg): Define.
4173 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4175 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4176 intrinsic register names.
4177 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4178 intrinsic register enum values and macros.
4180 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4182 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4183 for load/store addressing form.
4184 (nds32_print_operand_address): Likewise.
4186 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4189 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4190 based on LABEL_REF. Remove useless assertion.
4191 (pic_address_needs_scratch): Fix formatting.
4192 (sparc_legitimize_pic_address): Minor tweaks.
4193 (sparc_delegitimize_address): Adjust assertion accordingly.
4194 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4195 into symbolic_operand.
4196 (movsi_high_pic_label_ref): Likewise.
4197 (movsi_lo_sum_pic_label_ref): Likewise.
4198 (movdi_pic_label_ref): Likewise.
4199 (movdi_high_pic_label_ref): Likewise.
4200 (movdi_lo_sum_pic_label_ref): Likewise.
4202 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4204 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4205 custom LIB_SPEC setup.
4207 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4208 Kito Cheng <kito.cheng@gmail.com>
4210 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4211 * config/riscv/freebsd.h: New.
4213 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4215 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4216 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4219 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4220 Kito Cheng <kito.cheng@gmail.com>
4222 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4223 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4224 * config/nds32/nds32-protos.h (nds32_output_call,
4225 nds32_output_return): Declare.
4226 * config/nds32/nds32.md: Refine all the call and return patterns.
4228 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4231 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4232 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4234 PR rtl-optimization/84872
4235 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4236 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4239 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4241 * expr.c (copy_blkmode_to_reg): Revert 254862.
4242 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4244 2018-04-06 Richard Biener <rguenther@suse.de>
4247 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4248 after seeing a component reference with an adjacent field. Treat
4249 refs to arrays at struct end of external decls similar to
4250 refs to unconstrained commons.
4252 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4255 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4256 look through SAVE_EXPRs with non-side-effects argument. Adjust
4258 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4261 2018-04-06 Richard Biener <rguenther@suse.de>
4264 * alias.c (find_base_term): New wrapper around find_base_term
4265 unwinding CSELIB_VAL_PTR changes.
4266 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4269 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4271 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4273 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4274 constant definitions.
4275 ("nop"): lr 0,0 -> nopr r0
4276 ("nop_lr0", "nop_lr1"): New insn definitions.
4278 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4280 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4281 NDS32_V3PUSH_AVAILABLE_P macro.
4283 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4284 Chung-Ju Wu <jasonwucj@gmail.com>
4286 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4287 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4288 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4289 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4290 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4291 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4292 * config/nds32/constraints.md: New constraints and checking for hard
4293 float configuration.
4294 * config/nds32/iterators.md: New mode iterator and attribute for hard
4295 float configuration.
4296 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4298 * config/nds32/nds32-fpu.md: New file.
4299 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4300 deal with hard float code generation.
4301 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4303 (abi_type, float_reg_number): New enum type.
4304 * config/nds32/nds32-predicates.c: New predicates for hard float.
4305 * config/nds32/nds32-protos.h: Declare functions for hard float.
4306 * config/nds32/nds32.c: Implementation for hard float configuration.
4307 * config/nds32/nds32.h: Definitions for hard float configuration.
4308 * config/nds32/nds32.md: Include hard float machine description and
4309 modify patterns for hard float configuration.
4310 * config/nds32/nds32.opt: New options for hard float configuration.
4311 * config/nds32/predicates.md: New predicates for hard float
4314 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4316 * common/config/nds32/nds32-common.c
4317 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4319 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4322 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4323 CONSTRUCTOR_ELT (ctor, ...)->value.
4325 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4328 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4330 2018-04-05 Tom de Vries <tom@codesourcery.com>
4333 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4336 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4337 Kito Cheng <kito.cheng@gmail.com>
4339 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4340 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4341 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4342 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4344 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4345 Kito Cheng <kito.cheng@gmail.com>
4347 * config/nds32/constraints.md (Ufe): New memory constraint.
4348 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4349 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4350 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4352 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4353 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4355 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4357 * config/nds32/nds32.md: Use optimize_size in the condition for
4358 alu-shift instructions.
4360 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4362 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4364 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4366 * config/nds32/nds32.md (negsi2): Refine pattern.
4368 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4369 Chung-Ju Wu <jasonwucj@gmail.com>
4371 * config/nds32/iterators.md (shift_rotate): New code iterator.
4372 (shift): New code attribute.
4373 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4374 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4375 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4376 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4377 bit-wise operations.
4378 (andsi3, *andsi3): Ditto.
4379 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4380 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4381 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4382 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4383 nds32_ior_operand, nds32_xor_operand): New predicates.
4385 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4387 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4388 (addsi3, subsi3): ... this.
4390 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4392 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4394 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4396 * config/nds32/nds32.md: Adjust indention.
4398 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4400 * config/nds32/nds32.md (feature): New attribute.
4402 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4404 * config/nds32/nds32.md (subtype): New attribute.
4406 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4409 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4410 expansion to perform a bitwise AND of the argument followed by a
4411 boolean negation of the result.
4413 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4415 PR rtl-optimization/84878
4416 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4417 the basic block. Assert the use reference is not artificial and that
4418 it has an associated insn.
4420 2018-04-04 Michael Matz <matz@suse.de>
4422 * builtins.c (compute_objsize): Pass correct operand
4423 to array_at_struct_end_p.
4425 2018-04-04 Richard Biener <rguenther@suse.de>
4428 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4429 from contexts for DINFO_LEVEL_TERSE and below.
4431 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4433 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4434 resiter_operand condition.
4435 * config/nds32/nds32.md (*move<mode>): Ditto.
4437 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4438 Monk Chiang <sh.chiang04@gmail.com>
4440 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4442 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4444 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4446 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4447 Kito Cheng <kito.cheng@gmail.com>
4449 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4450 nds32_cond_code_str, output_cond_branch,
4451 output_cond_branch_compare_zero, nds32_expand_cbranch,
4452 nds32_expand_cstore, nds32_expand_movcc,
4453 nds32_output_cbranchsi4_equality_zero,
4454 nds32_output_cbranchsi4_equality_reg,
4455 nds32_output_cbranchsi4_equality_reg_or_const_int,
4456 nds32_output_cbranchsi4_greater_less_zero: New functions.
4457 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4458 nds32_expand_cstore, nds32_expand_movcc,
4459 nds32_output_cbranchsi4_equality_zero,
4460 nds32_output_cbranchsi4_equality_reg,
4461 nds32_output_cbranchsi4_equality_reg_or_const_int,
4462 nds32_output_cbranchsi4_greater_less_zero): Declare.
4463 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4464 nds32_rimm11s_operand): New predicates.
4465 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4466 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4469 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4471 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4472 * config/nds32/nds32.md: Ditto.
4473 * config/nds32/pipelines.md: Ditto.
4475 2018-04-04 Richard Biener <rguenther@suse.de>
4477 PR tree-optimization/85168
4478 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4479 propagating abnormals.
4481 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4483 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4485 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4486 Kito Cheng <kito.cheng@gmail.com>
4488 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4489 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4490 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4491 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4492 * config/nds32/nds32.md (sibcall_internal): New.
4493 (sibcall_register): Remove.
4494 (sibcall_immediate): Remove.
4495 (sibcall_value_internal): New.
4496 (sibcall_value_register): Remove.
4497 (sibcall_value_immediate): Remove.
4498 * config/nds32/predicates.md (nds32_general_register_operand): New.
4499 (nds32_call_address_operand): New.
4501 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4503 PR rtl-optimization/85167
4504 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4505 bb_defs if *split_p, instead preinitialize it to NULL.
4507 PR tree-optimization/85156
4508 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4509 evaluating the argument multiple times.
4511 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4513 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4515 (_mm_cvtpd_ps): Likewise.
4516 (_mm_cvttpd_epi32): Likewise.
4517 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4518 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4519 vector, pixel, and bool following altivec.h include.
4521 2018-04-03 Martin Sebor <msebor@redhat.com>
4523 * doc/extend.texi (Common Function Attributes): Clarify.
4524 (const attribute): Likewise.
4525 (pure attribute): Likewise.
4527 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4530 * config/i386/i386.c (ix86_expand_vector_set): Use
4531 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4533 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4535 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4536 instructions when changing rounding bits to preserve precision bits
4537 in the x87 control word.
4539 2018-04-03 Martin Liska <mliska@suse.cz>
4541 PR tree-optimization/82491
4542 * rtl.h (strip_offset_and_add): Replace += suboffset with
4543 poly_uint64 () + suboffset.
4545 2018-03-29 Martin Liska <mliska@suse.cz>
4546 Martin Jambor <mjambor@suse.cz>
4549 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4550 param_type is not an integral or pointer type.
4552 2018-04-03 Richard Biener <rguenther@suse.de>
4554 * sese.h (recompute_all_dominators): Remove.
4556 2018-04-02 Martin Sebor <msebor@redhat.com>
4558 * doc/invoke.texi (-Wrestrict): Fix typos.
4560 2018-04-02 Jim Wilson <jimw@sifive.com>
4562 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4563 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4564 (<optab>di3, <optab>si3_extend): Likewise.
4565 (<optab>si3_mask, <optab>si3_mask_1): New.
4566 (<optab>di3_mask, <optab>di3_mask_1): New.
4567 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4568 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4569 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4571 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4573 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4576 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4578 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4579 (nds32_canonicalize_comparison): New function.
4581 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4582 Kito Cheng <kito.cheng@gmail.com>
4583 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4585 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4586 * config/nds32/constants.md (unspec_volatile_element): Add
4587 UNSPEC_VOLATILE_RELAX_GROUP.
4588 * config/nds32/nds32-relax-opt.c: New file.
4589 * config/nds32/nds32-predicates.c
4590 (nds32_symbol_load_store_p): New function.
4591 * config/nds32/nds32-protos.h
4592 (nds32_symbol_load_store_p): Declare function.
4593 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4594 * config/nds32/nds32.c
4595 (nds32_register_pass): New function to register pass.
4596 (nds32_register_passes): New function to register passes.
4597 * config/nds32/nds32.md (relax_group): New pattern.
4598 * config/nds32/nds32.opt (mrelax-hint): New option.
4599 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4601 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4603 * config/nds32/t-nds32: Modify files dependency.
4605 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4607 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4608 (PROFILE_HOOK): Define its implementation.
4610 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4612 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4613 type and 32-bit size.
4615 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4618 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4619 (V_128_256): New mode iterator.
4620 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4621 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4622 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4624 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4625 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4627 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4630 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4631 NaN inputs correctly.
4633 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4636 * config/rs6000/vsx.md (??r): New mode attribute.
4637 (*vsx_mov<mode>_64bit): Use it.
4638 (*vsx_mov<mode>_32bit): Likewise.
4640 2018-03-30 Martin Sebor <msebor@redhat.com>
4642 PR tree-optimization/84818
4643 * builtins.c (check_access): Use warning_n.
4645 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4648 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4650 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4653 2018-03-30 Julia Koval <julia.koval@intel.com>
4656 * x86-tune.def (movx, partial_reg_dependency): Enable for
4659 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4662 * lra-constraints.c (process_alt_operands): Move setting
4663 this_alternative_matches below.
4665 2018-03-29 Martin Liska <mliska@suse.cz>
4668 * doc/invoke.texi: Document how LTO works with debug info.
4669 Describe auto-load support of binutils. Mention 'x86-64'
4670 as valid option value of -march option.
4672 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4674 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4677 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4678 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4679 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4682 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4685 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4686 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4687 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4688 for RS6000_BTM_POWERPC64.
4689 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4690 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4691 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4696 2018-03-28 Carl Love <cel@us.ibm.com>
4699 2017-09-27 Carl Love <cel@us.ibm.com>
4701 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4702 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4703 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4706 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4708 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4709 instead of __vector bool.
4710 (_mm_max_pu8): Likewise.
4711 (_mm_min_pi16): Likewise.
4713 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4716 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4717 (DIVWEUO): Likewise.
4719 (DIVDEUO): Likewise.
4720 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4721 DIVWEUO and DIVDEUO.
4722 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4723 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4724 (div_extend): Likewise.
4725 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4727 (__builtin_divweuo): Likewise.
4728 (__builtin_divdeo): Likewise.
4729 (__builtin_divdeuo): Likewise.
4731 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4734 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4735 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4737 PR tree-optimization/82004
4738 * gimple-match-head.c (optimize_pow_to_exp): New function.
4739 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4740 Don't fold to exp if optimize_pow_to_exp is false.
4742 2018-03-28 Martin Liska <mliska@suse.cz>
4745 * calls.c (initialize_argument_information): Fix trailing space.
4746 * common.opt: Fix typo and provide better explanation for
4747 -fsanitize-coverage option.
4748 * config/i386/i386.opt: Fix typo.
4750 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4751 Martin Liska <mliska@suse.cz>
4754 * gimplify.c (asan_poison_variable): Don't do the check for
4755 gimplify_omp_ctxp here.
4756 (gimplify_decl_expr): Do it here.
4757 (gimplify_target_expr): Likewise.
4759 2018-03-28 Martin Liska <mliska@suse.cz>
4762 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4763 chkp_type_bounds_count if MPX is not enabled.
4765 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4767 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4769 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4772 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4773 function to create the function decl for complex long double
4774 multiply and divide for -mabi=ieeelongdouble.
4775 (init_float128_ieee): Call it.
4777 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4780 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4781 -fcf-protection=branch -mibt.
4782 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4784 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4787 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4789 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4792 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4793 extern array declarations.
4795 2018-03-27 Richard Biener <rguenther@suse.de>
4798 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4799 explicit single_use checks.
4801 2018-03-27 Richard Biener <rguenther@suse.de>
4803 PR tree-optimization/85082
4804 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4807 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4809 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4810 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4811 Turn on fasynchronous-unwind-tables and funwind-tables.
4813 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4816 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4817 (*bmi_blsr_<mode>_ccz): Ditto.
4819 2018-03-26 Tom de Vries <tom@codesourcery.com>
4821 PR tree-optimization/85063
4822 * omp-general.c (offloading_function_p): New function. Factor out
4824 * omp-offload.c (pass_omp_target_link::gate): ... here.
4825 * omp-general.h (offloading_function_p): Declare.
4826 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4827 with attribute omp declare target for offloading functions.
4829 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4831 PR tree-optimization/84005
4832 * tree-data-ref.h (get_base_for_alignment): Declare.
4833 * tree-data-ref.c (get_base_for_alignment_1): New function.
4834 (get_base_for_alignment): Likewise.
4835 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4836 get_base_for_alignment to find a suitable base object, instead
4837 of always using drb->base_address.
4839 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4842 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4843 known size by default.
4845 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4848 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4849 and non BLKmode operands.
4851 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4854 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4855 Clean up attributes.
4857 2018-03-23 Richard Biener <rguenther@suse.de>
4860 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4861 we are going to emit early debug for LTO.
4863 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4866 * function.c (match_asm_constraints_1): Don't optimize if input
4867 doesn't satisfy general_operand predicate for output's mode.
4870 * alias.c (write_dependence_p): Don't require for x_canonicalized
4871 non-VOIDmode if x has VOIDmode.
4874 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4875 just don't try to optimize it rather than assert it never happens.
4877 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4879 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4880 macro expansions for definition of ST_INTERNAL_<mode> and
4881 LD_INTERNAL_<mode> builtins.
4882 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4884 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4886 (altivec_expand_st_builtin): Likewise.
4887 (altivec_expand_builtin): Remove calls to deleted functions.
4888 (rs6000_address_for_altivec): Delete this function.
4889 * config/rs6000/vector.md: Remove expands for
4890 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4892 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4895 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4896 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4897 re-computing once computed.
4898 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4899 (arm_init_machine_status): Initialize
4900 machine->static_chain_stack_bytes.
4902 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4905 * doc/extend.texi: Add four new prototypes for vec_ld.
4906 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4907 definitions for more logical presentation.
4908 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4909 entries for V1TI variants of __builtin_altivec_ld builtin.
4910 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4911 handling of V1TI variant of LVX icode pattern.
4912 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4913 (rs6000_gimple_fold_builtin): Likewise.
4914 (altivec_init_builtins): Add code to define
4915 __builtin_altivec_lvx_v1ti function.
4917 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4920 * function.c (match_asm_constraints_1): Don't do the optimization
4921 if input isn't a REG, SUBREG, MEM or constant.
4923 2018-03-22 Tom de Vries <tom@codesourcery.com>
4925 PR tree-optimization/84956
4926 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4927 bb_has_abnormal_pred.
4929 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4932 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4933 DECL_INITIAL (decl) to decl at the end.
4934 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4937 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4939 * doc/extend.texi (__builtin_tgmath): Document when complex
4940 integer types are treated as _Complex _Float64.
4942 2018-03-21 Tom de Vries <tom@codesourcery.com>
4944 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4946 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4948 PR tree-optimization/84960
4949 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4950 if it is ENTRY block, move them into single succ of ENTRY in that case.
4952 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4954 PR tree-optimization/84811
4955 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4956 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4957 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4958 (known_subrange_p): Update accordingly. Cast each value involved
4959 in the size comparison, rather than casting the result of the
4962 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4964 PR tree-optimization/84982
4965 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4966 by flipping the least significant bit rather than all bits from
4967 bitpos to bitpos + bitsize - 1.
4969 2018-03-21 Nathan Sidwell <nathan@acm.org>
4971 * doc/extend.texi (Deprecated Features): Remove mention of
4972 long-deleted deprecations.
4974 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4977 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4978 * configure: Regenerate.
4980 2018-03-21 Tom de Vries <tom@codesourcery.com>
4982 PR tree-optimization/83126
4983 * tree-parloops.c (num_phis): New function.
4984 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4986 2018-03-21 Nathan Sidwell <nathan@acm.org>
4988 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4989 mention anon-struct/union members and trailing attributes.
4991 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4993 PR tree-optimization/84969
4994 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4995 builtin memset partitions if they set different rhs values.
4997 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4999 PR rtl-optimization/84989
5000 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5001 VEC_DUPLICATE with scalar result mode.
5003 2018-03-21 Martin Liska <mliska@suse.cz>
5006 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5007 not intended return statement.
5009 2018-03-21 Martin Liska <mliska@suse.cz>
5012 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5013 (chkp_find_bound_slots_1): Limit number of iterations.
5015 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5018 * Minor grammar fixes for x86 options.
5020 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5023 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5024 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5026 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5029 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5030 (altivec_lvx_<mode>_1op): Likewise.
5031 (altivec_stvx_<mode>_2op): Likewise.
5032 (altivec_stvx_<mode>_1op): Likewise.
5033 (altivec_lvx_<VM2:mode>): New define_expand.
5034 (altivec_stvx_<VM2:mode>): Likewise.
5035 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5036 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5037 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5038 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5039 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5040 (rs6000_gen_lvx): Likewise.
5041 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5042 (altivec_expand_stv_builtin): Likewise.
5043 (altivec_expand_builtin): Likewise.
5044 * config/rs6000/vector.md: Likewise.
5046 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5049 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5052 2018-03-20 Richard Biener <rguenther@suse.de>
5055 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5056 sign-conversions as zero, fall back to standard scalar_stmt
5059 2018-03-20 Martin Liska <mliska@suse.cz>
5062 * predict.c (rebuild_frequencies): Handle case when we have
5063 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5065 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5068 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5069 flag_section_anchors.
5070 * varasm.c (use_blocks_for_decl_p): Remove hack for
5071 dw2_force_const_mem.
5074 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5076 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5077 be created, use lowpart_subreg of operands[0] rather than operands[0]
5079 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5080 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5081 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5082 and n constraint instead of aarch64_shift_imm_di and Usd.
5083 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5084 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5086 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5089 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5090 to favor GPR over NEON registers.
5091 (<shift>di3_neon): Likewise.
5093 2018-03-20 Tom de Vries <tom@codesourcery.com>
5096 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5097 (nvptx_process_pars): Emit bar.sync asap and alap.
5099 2018-03-20 Tom de Vries <tom@codesourcery.com>
5102 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5103 seen_label if seen_label is already set.
5105 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5108 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5109 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5110 Use 1U instead of 1. Formatting fixes.
5113 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5114 instead of TREE_TYPE (s1) for the return value.
5116 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5118 PR tree-optimization/84946
5119 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5120 bitsize + bitsize in poly_uint64 rather than poly_int64.
5123 * dwarf2asm.c: Include fold-const.c.
5124 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5125 of decl rather than decl itself.
5127 PR rtl-optimization/84643
5128 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5130 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5133 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5134 calling assemble_variable.
5136 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5139 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5140 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5142 2018-03-19 Jim Wilson <jimw@sifive.com>
5145 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5146 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5147 (riscv_first_stack_step): Likewise.
5148 (riscv_option_override): Use STACK_BOUNDARY instead of
5150 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5152 (BIGGEST_ALIGNMENT): Set to 128.
5153 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5154 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5157 2018-03-19 Richard Biener <rguenther@suse.de>
5159 PR tree-optimization/84933
5160 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5161 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5163 2018-03-19 Richard Biener <rguenther@suse.de>
5165 PR tree-optimization/84859
5166 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5167 (cond_if_else_store_replacement): Perform sinking operation on
5168 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5169 Generalize what a BB with a single eligible store is.
5171 2018-03-19 Richard Biener <rguenther@suse.de>
5173 PR tree-optimization/84929
5174 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5175 chrec_is_positive against non-chrec arg.
5177 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5180 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5182 2018-03-18 Martin Liska <mliska@suse.cz>
5184 PR rtl-optimization/84635
5185 * regrename.c (build_def_use): Use matches_mode only when
5188 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5190 PR tree-optimization/84913
5191 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5192 vectorize chains of COND_EXPRs.
5194 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5196 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5198 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5200 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5202 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5204 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5206 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5207 Kito Cheng <kito.cheng@gmail.com>
5209 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5210 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5211 (nds32_adjust_reg_alloc_order): New function.
5212 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5214 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5216 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5217 nds32_print_operand, nds32_print_operand_address): Use
5218 HOST_WIDE_INT_PRINT_DEC instead.
5220 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5222 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5224 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5227 * config/i386/i386.c (initial_ix86_tune_features,
5228 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5230 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5231 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5232 rather than 1u << ix86_tune. Formatting fix.
5233 (ix86_option_override_internal): Change ix86_arch_mask from
5234 unsigned int to unsigned HOST_WIDE_INT, initialize to
5235 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5236 (ix86_function_specific_restore): Likewise.
5238 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5241 * postreload.c (reload_combine_recognize_pattern): Perform
5242 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5243 truncate_int_for_mode the result for the destination's mode.
5246 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5247 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5249 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5252 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5255 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5256 regno_allocno_class_array and sorted_pseudos.
5257 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5258 insns where regno is used.
5260 2018-03-16 Martin Liska <mliska@suse.cz>
5263 * multiple_target.c (create_dispatcher_calls): Redirect
5264 reference in the symbol table.
5266 2018-03-16 Martin Liska <mliska@suse.cz>
5269 * multiple_target.c (create_dispatcher_calls): Redirect also
5272 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5276 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5277 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5278 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5280 2018-03-16 Julia Koval <julia.koval@intel.com>
5282 * doc/invoke.texi (Skylake Server): Add CLWB.
5283 Cannonlake): Remove CLWB.
5285 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5287 PR tree-optimization/84841
5288 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5290 (FLOAT_ONE_CONST_TYPE): Define.
5291 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5292 (sort_by_operand_rank): Put entries with higher constant_type last
5293 rather than first to match comments.
5295 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5297 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5300 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5303 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5304 check and use address of target always.
5306 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5309 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5310 (indirect_thunk_bnd_needed): Likewise.
5311 (indirect_thunks_used): Likewise.
5312 (indirect_thunks_bnd_used): Likewise.
5313 (indirect_return_needed): New.
5314 (indirect_return_bnd_needed): Likewise.
5315 (output_indirect_thunk_function): Add a bool argument for
5317 (output_indirect_thunk_function): Don't generate alias for
5318 function return thunk.
5319 (ix86_code_end): Call output_indirect_thunk_function to generate
5320 function return thunks.
5321 (ix86_output_function_return): Set indirect_return_bnd_needed
5322 and indirect_return_needed instead of indirect_thunk_bnd_needed
5323 and indirect_thunk_needed.
5325 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5327 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5328 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5329 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5331 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5332 Paul Hua <paul.hua.gm@gmail.com>
5335 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5337 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5339 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5340 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5343 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5346 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5347 instead of GET_MODE_SIZE when comparing Units.
5349 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5352 * varasm.c (hash_section): Return an unchangeble hash value
5353 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5354 Return !aarch64_can_use_per_function_literal_pools_p ().
5356 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5359 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5360 rather than address of cmode as last argument to prepare_cmp_insn.
5362 2018-03-15 Julia Koval <julia.koval@intel.com>
5364 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5365 F_AVX512VNNI, F_AVX512BITALG): New.
5367 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5370 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5371 insn for floating-point loads and stores.
5373 2018-03-14 Carl Love <cel@us.ibm.com>
5375 * config/rs6000/rs6000-c.c: Add macro definitions for
5376 ALTIVEC_BUILTIN_VEC_PERMXOR.
5377 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5378 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5379 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5380 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5382 * config/doc/extend.texi: Add prototypes for vec_permxor.
5384 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5387 * diagnostic-show-locus.c (class layout_point): Convert m_line
5388 from int to linenum_type.
5389 (line_span::comparator): Use linenum "compare" function when
5390 comparing line numbers.
5391 (test_line_span): New function.
5392 (layout_range::contains_point): Convert param "row" from int to
5394 (layout_range::intersects_line_p): Likewise.
5395 (layout::will_show_line_p): Likewise.
5396 (layout::print_source_line): Likewise.
5397 (layout::should_print_annotation_line_p): Likewise.
5398 (layout::print_annotation_line): Likewise.
5399 (layout::print_leading_fixits): Likewise.
5400 (layout::annotation_line_showed_range_p): Likewise.
5401 (struct line_corrections): Likewise for field m_row.
5402 (line_corrections::line_corrections): Likewise for param "row".
5403 (layout::print_trailing_fixits): Likewise.
5404 (layout::get_state_at_point): Likewise.
5405 (layout::get_x_bound_for_row): Likewise.
5406 (layout::print_line): Likewise.
5407 (diagnostic_show_locus): Likewise for locals "last_line" and
5409 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5410 * input.c (selftest::test_linenum_comparisons): New function.
5411 (selftest::input_c_tests): Call it.
5412 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5413 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5414 * selftest.h (ASSERT_GT): New macro.
5415 (ASSERT_GT_AT): New macro.
5416 (ASSERT_LT): New macro.
5417 (ASSERT_LT_AT): New macro.
5419 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5421 PR rtl-optimization/84780
5422 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5424 2018-03-14 Martin Liska <mliska@suse.cz>
5426 * tree.c (record_node_allocation_statistics): Use
5427 get_stats_node_kind.
5428 (get_stats_node_kind): New function extracted from
5429 record_node_allocation_statistics.
5430 (free_node): Use get_stats_node_kind.
5432 2018-03-14 Richard Biener <rguenther@suse.de>
5434 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5435 that the value-set of ANTIC_IN doesn't grow.
5438 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5440 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5441 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5443 2018-03-14 Julia Koval <julia.koval@intel.com>
5445 * config.gcc (icelake-client, icelake-server): New.
5447 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5448 (initial_ix86_arch_features): Ditto.
5449 (PTA_SKYLAKE): Add SGX.
5450 (PTA_ICELAKE): Remove.
5451 (PTA_ICELAKE_CLIENT): New.
5452 (PTA_ICELAKE_SERVER): New.
5453 (ix86_option_override_internal): Split up icelake on icelake client and
5455 (get_builtin_code_for_version): Ditto.
5456 (fold_builtin_cpu): Ditto.
5457 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5458 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5459 * config/i386/i386.h (processor_type): Ditto.
5460 * doc/invoke.texi: Ditto.
5462 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5465 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5466 INTEGER_CST offset, add it together with bitpos / 8 and
5467 sign extend based on POINTER_SIZE.
5471 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5474 * config/i386/constraints.md (Yc): New register constraint.
5475 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5476 Use Yc constraint for alternative 2 of operand 0. Remove
5477 preferred_for_speed attribute.
5479 2018-03-14 Richard Biener <rguenther@suse.de>
5481 PR tree-optimization/84830
5482 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5483 with the old one to avoid oscillations.
5485 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5488 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5490 (assign_by_spills): Return a flag of reload assignment failure.
5491 Do not process the reload assignment failures. Do not spill other
5492 reload pseudos if they has the same reg class. Update n if
5494 (lra_assign): Add a return arg. Set up from the result of
5495 assign_by_spills call.
5496 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5497 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5498 usage_insns if it is not NULL.
5499 (spill_hard_reg_in_range): New function.
5500 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5501 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5502 function prototypes.
5503 (lra_assign): Change prototype.
5504 * lra.c (lra): Add code to deal with fails by splitting hard reg
5507 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5509 * config/riscv/riscv.opt (mrelax): New option.
5510 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5511 "norelax" when riscv_mrelax is disabled.
5512 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5514 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5517 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5518 reassociation for int modes.
5520 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5522 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5523 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5525 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5526 * config/aarch64/aarch64-sve.md
5527 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5528 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5529 (*extend<mode><Vwide>2): Rename to...
5530 (aarch64_sve_extend<mode><Vwide>2): ...this.
5531 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5532 renaming the old pattern to...
5533 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5535 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5536 define_expand, renaming the old pattern to...
5537 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5538 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5539 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5540 account when deciding which SVE instruction the optab should use.
5541 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5543 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5545 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5546 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5547 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5548 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5549 (tlsdesc_small_<mode>): Turn a define_expand and use
5550 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5551 (tlsdesc_small_advsimd_<mode>): ...this.
5552 (tlsdesc_small_sve_<mode>): New pattern.
5554 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5556 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5557 (UNSPEC_UMUL_HIGHPART): New constants.
5558 (MUL_HIGHPART): New int iteraor.
5559 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5560 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5562 (*<su>mul<mode>3_highpart): New define_insn.
5564 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5567 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5570 2018-03-13 Martin Liska <mliska@suse.cz>
5573 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5575 (sem_item_optimizer::~sem_item_optimizer): Release it.
5576 (sem_item_optimizer::merge_classes): Register variable aliases.
5577 (sem_item_optimizer::fixup_pt_set): New function.
5578 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5579 * ipa-icf.h: Declare new variables and functions.
5581 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5584 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5585 integer_pow2p@2 and test integer_pow2p in condition.
5586 (A < 0 ? C : 0): Similarly for @1.
5589 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5590 characters starting at p contain '\0' character, don't look beyond
5594 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5595 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5598 * reg-stack.c (change_stack): Change update_end var from int to
5599 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5600 also call set_block_for_insn on the newly added insns and rescan.
5603 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5604 on the last operand.
5607 * tree.c (stabilize_reference_1): Return save_expr (e) for
5608 STATEMENT_LIST even if it doesn't have side-effects.
5610 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5612 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5614 2018-03-12 Renlin Li <renlin.li@arm.com>
5616 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5617 aarch64_output_scalar_simd_mov_immediate.
5619 2018-03-12 Martin Sebor <msebor@redhat.com>
5621 PR tree-optimization/83456
5622 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5623 for perfectly overlapping calls to memcpy.
5624 (gimple_fold_builtin_memory_chk): Same.
5625 (gimple_fold_builtin_strcpy): Handle no-warning.
5626 (gimple_fold_builtin_stxcpy_chk): Same.
5627 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5629 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5631 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5632 parameter. Use it for SFmode.
5633 (rs6000_function_arg_advance_1): Adjust.
5634 (rs6000_function_arg): Adjust.
5635 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5637 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5639 PR rtl-optimization/84169
5640 PR rtl-optimization/84780
5641 * combine.c (can_combine_p): Check for a 2-insn combination whether
5642 the destination register is used between the two insns, too.
5644 2018-03-12 Richard Biener <rguenther@suse.de>
5646 PR tree-optimization/84803
5647 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5648 for refs DR analysis didn't process.
5650 2018-03-12 Richard Biener <rguenther@suse.de>
5652 PR tree-optimization/84777
5653 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5654 force-vectorize loops ignore whether we are optimizing for size.
5656 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5658 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5659 (TARGET_MD_ASM_ADJUST): Define.
5661 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5662 Kito Cheng <kito.cheng@gmail.com>
5663 Chung-Ju Wu <jasonwucj@gmail.com>
5665 * config/nds32/nds32.c (nds32_compute_stack_frame,
5666 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5667 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5668 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5669 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5670 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5671 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5672 * config/nds32/nds32.md (prologue, epilogue): Use macro
5673 NDS32_V3PUSH_AVAILABLE_P to do checking.
5675 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5678 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5679 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5680 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5681 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5682 addition of most attributes on !orig_type_die or the attribute not
5683 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5685 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5686 Chung-Ju Wu <jasonwucj@gmail.com>
5688 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5690 * config/nds32/nds32.opt (mvh): New option.
5692 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5693 Chung-Ju Wu <jasonwucj@gmail.com>
5695 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5697 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5698 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5701 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5702 Chung-Ju Wu <jasonwucj@gmail.com>
5704 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5706 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5707 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5709 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5710 Kito Cheng <kito.cheng@gmail.com>
5711 Chung-Ju Wu <jasonwucj@gmail.com>
5713 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5714 UNSPEC_FFMISM and UNSPEC_FLMISM.
5715 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5716 for ffb, ffmism and flmism.
5717 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5718 (unspec_ffmism): Ditto.
5719 (unspec_flmism): Ditto.
5720 (nds32_expand_builtin_impl): Check if string extension is available.
5721 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5722 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5724 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5727 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5730 * lra-assigns.c (assign_by_spills): Return a flag of reload
5731 assignment failure. Do not process the reload assignment
5732 failures. Do not spill other reload pseudos if they has the same
5734 (lra_assign): Add a return arg. Set up from the result of
5735 assign_by_spills call.
5736 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5737 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5738 usage_insns if it is not NULL.
5739 (spill_hard_reg_in_range): New function.
5740 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5741 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5742 function prototypes.
5743 (lra_assign): Change prototype.
5744 * lra.c (lra): Add code to deal with fails by splitting hard reg
5747 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5750 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5752 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5755 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5756 (dw_val_node): Add val_symbolic_view.
5757 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5758 (symview_upper_bound): New.
5759 (new_line_info_table): Initialize symviews_since_reset.
5760 (dwarf2out_source_line): Count symviews_since_reset and set
5761 symview_upper_bound.
5762 (dw_val_equal_p): Handle symview.
5763 (add_AT_symview): New.
5764 (print_dw_val): Handle symview.
5765 (attr_checksum, attr_checksum_ordered): Likewise.
5766 (same_dw_val_p, size_of_die): Likewise.
5767 (value_format, output_die): Likewise.
5768 (add_high_low_attributes): Use add_AT_symview for entry_view.
5769 (dwarf2out_finish): Reset symview_upper_bound, clear
5772 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5775 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5776 Add strict argument and use it.
5777 (rs6000_split_multireg_move): Update for new strict argument.
5778 (mem_operand_gpr): Disallow all non-offsettable addresses.
5779 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5781 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5784 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5785 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5786 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5789 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5790 decl, use remap_type if we want to use the type.
5792 2018-03-09 Martin Sebor <msebor@redhat.com>
5794 PR tree-optimization/84526
5795 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5797 (builtin_access::generic_overlap): Be prepared to handle non-array
5800 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5802 PR rtl-optimization/84682
5803 * lra-constraints.c (process_address_1): Check is_address flag
5804 for address constraints.
5805 (process_alt_operands): Likewise.
5806 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5807 preprocess_constraints.
5808 * recog.h (preprocess_constraints): Add oploc parameter.
5810 * recog.c (preprocess_constraints): Test address_operand for
5811 CT_ADDRESS constraints.
5813 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5816 * lra-assigns.c (assign_by_spills): Return a flag of reload
5817 assignment failure. Do not process the reload assignment
5818 failures. Do not spill other reload pseudos if they has the same
5820 (lra_assign): Add a return arg. Set up from the result of
5821 assign_by_spills call.
5822 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5823 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5824 usage_insns if it is not NULL.
5825 (spill_hard_reg_in_range): New function.
5826 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5827 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5828 function prototypes.
5829 (lra_assign): Change prototype.
5830 * lra.c (lra): Add code to deal with fails by splitting hard reg
5833 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5836 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5837 Accept complain bool parameter. Only emit errors if it is true.
5838 (arm_parse_cpu_option_name): Likewise.
5839 (arm_target_thumb_only): Adjust callers of the above.
5840 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5841 prototype to take a default true bool parameter.
5842 (arm_parse_arch_option_name): Likewise.
5844 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5845 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5849 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5850 * configure: Regenerate.
5851 * configure.ac ("linker --version-script option"): New.
5852 ("linker soname option"): New.
5854 2018-03-09 Richard Biener <rguenther@suse.de>
5856 PR tree-optimization/84775
5857 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5858 immediate uses of predicate stmts and mark them modified.
5861 PR tree-optimization/84178
5862 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5864 (version_loop_for_if_conversion): Delay update_ssa call.
5865 (tree_if_conversion): Delay update_ssa until after predicate
5868 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5871 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5872 when the function accesses prior frames.
5874 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5877 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5878 gen_llsym, otherwise call maybe_gen_llsym.
5881 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5882 has ',' character inside of it.
5884 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5887 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5888 as clobbering CC_REGNUM.
5890 2018-03-08 Richard Biener <rguenther@suse.de>
5893 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5894 (follow_copies_to_constant): Do not follow SSA names registered
5897 2018-03-08 Richard Biener <rguenther@suse.de>
5899 PR tree-optimization/84178
5900 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5902 (version_loop_for_if_conversion): Delay update_ssa call.
5903 (tree_if_conversion): Delay update_ssa until after predicate
5906 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5908 PR tree-optimization/84178
5909 * tree-if-conv.c (release_bb_predicate): Remove the
5910 the assertion that the stmts have NULL use_ops.
5911 Discard the statements, asserting that they haven't
5912 yet been added to a BB.
5914 2018-03-08 Richard Biener <rguenther@suse.de>
5916 PR tree-optimization/84746
5917 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5918 (phi_translate): Pass in destination ANTIC_OUT set.
5919 (phi_translate_1): Likewise. For a simplified result lookup
5920 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5921 (phi_translate_set): Adjust.
5922 (do_pre_regular_insertion): Likewise.
5923 (do_pre_partial_partial_insertion): Likewise.
5925 2018-03-08 Martin Liska <mliska@suse.cz>
5927 PR gcov-profile/84735
5928 * doc/gcov.texi: Document usage of profile files.
5929 * gcov-io.h: Document changes in the format.
5931 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5935 * dwarf2out.c (struct dw_line_info_table): Update comments for
5937 (FORCE_RESET_NEXT_VIEW): New.
5938 (FORCE_RESETTING_VIEW_P): New.
5939 (RESETTING_VIEW_P): Check for -1 too.
5940 (ZERO_VIEW_P): Likewise.
5941 (new_line_info_table): Force-reset next view.
5942 (dwarf2out_begin_function): Likewise.
5943 (dwarf2out_source_line): Simplify zero_view_p initialization.
5944 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5945 view directly. Omit view when omitting .loc at line 0.
5947 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5949 PR tree-optimization/84740
5950 * tree-switch-conversion.c (process_switch): Call build_constructors
5951 only if info.phi_count is non-zero.
5953 PR tree-optimization/84739
5954 * tree-tailcall.c (find_tail_calls): Check call arguments against
5955 DECL_ARGUMENTS (current_function_decl) rather than
5956 DECL_ARGUMENTS (func) when checking for tail recursion.
5958 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5960 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5961 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5962 Volker Reichelt's entry and add entries for people that perform
5963 GCC fuzzy testing and report numerous bugs.
5965 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5968 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5969 readonly data in sdata, if that is disabled.
5970 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5971 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5972 -mreadonly-in-sdata option.
5974 2018-03-07 Martin Sebor <msebor@redhat.com>
5976 PR tree-optimization/84468
5977 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5978 basic block when looking for nul assignment.
5980 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5983 * except.h (output_function_exception_table): Adjust prototype.
5984 * except.c (output_function_exception_table): Remove FNNAME parameter
5985 and add SECTION parameter. Ouput one part of the table at a time.
5986 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5987 the first part of the exception table and emit unwind directives.
5988 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5989 (i386_pe_seh_cold_init): Likewise.
5990 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5991 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5992 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5993 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5994 * config/i386/winnt.c: Include except.h.
5995 (struct seh_frame_state): Add reg_offset, after_prologue and
5996 in_cold_section fields.
5997 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5998 (i386_pe_seh_cold_init): New function.
5999 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6000 to seh->in_cold_section.
6001 (seh_emit_push): Record the offset of the push.
6002 (seh_emit_save): Record the offet of the save.
6003 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6004 Test seh->after_prologue to disregard the epilogue.
6005 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6006 (i386_pe_end_cold_function): New function.
6008 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6011 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6012 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6015 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6017 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6018 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6021 * multiple_target.c: Include tree-inline.h and intl.h.
6022 (expand_target_clones): Diagnose and fail if node->definition and
6023 !tree_versionable_function_p (node->decl).
6025 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6027 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6029 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6030 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6031 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6033 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6036 * combine.c (try_combine): Use reg_or_subregno instead of handling
6037 just paradoxical SUBREGs and REGs.
6039 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6041 * config/arc/arc.c (arc_finalize_pic): Remove function.
6042 (arc_must_save_register): We use single base PIC register, remove
6043 checks to save/restore the PIC register.
6044 (arc_expand_prologue): Likewise.
6045 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6047 (arc_verify_short): Likewise.
6048 (arc_attr_type): Likewise.
6049 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6050 (walk_stores): Likewise.
6051 (arc_address_cost): Make it static.
6052 (arc_verify_short): Likewise.
6053 (branch_dest): Likewise.
6054 (arc_attr_type): Likewise.
6055 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6056 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6057 (arc_final_prescan_insn): Remove inserting the nops due to
6058 hardware hazards. It is done in reorg step.
6059 (insn_length_variant_t): Remove.
6060 (insn_length_parameters_t): Likewise.
6061 (arc_insn_length_parameters): Likewise.
6062 (arc_get_insn_variants): Likewise.
6063 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6065 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6068 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6071 PR tree-optimization/84687
6072 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6074 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6076 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6078 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6079 Rename to ppc_speculation_barrier.
6080 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6081 __builtin_ppc_speculation_barrier.
6083 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6086 * combine.c (combine_simplify_rtx): Don't try to simplify if
6087 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6090 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6092 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6093 to 32 bytes when compiling for POWER9.
6095 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6098 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6099 regparm >= 3 with no arg reg available also for calls with
6100 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6103 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6105 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6107 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6110 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6112 2018-03-05 Richard Biener <rguenther@suse.de>
6114 PR tree-optimization/84486
6115 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6116 When inserting a __builtin_assume_aligned call set the LHS
6117 SSA name alignment info accordingly.
6119 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6121 PR tree-optimization/84114
6122 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6123 Avoid reassociation of FLOAT_MODE addition.
6125 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6127 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6128 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6129 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6130 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6131 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6132 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6133 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6135 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6136 __builtin_ia32_wbinvd): New builtins.
6137 (SPECIAL_ARGS2): New.
6138 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6139 (SPECIAL_ARGS2): New.
6140 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6141 (ix86_valid_target_attribute_inner_p): Ditto.
6142 (ix86_init_mmx_sse_builtins): Add special_args2.
6143 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6144 TARGET_WBNOINVD_P): New.
6145 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6146 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6147 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6148 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6149 * config/i386/pconfigintrin.h: New file.
6150 * config/i386/wbnoinvdintrin.h: Ditto.
6151 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6152 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6154 2018-03-05 Richard Biener <rguenther@suse.de>
6156 PR tree-optimization/84670
6157 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6159 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6160 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6161 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6162 doesn't grow if all successors (recursively) were visited at least
6165 2018-03-05 Richard Biener <rguenther@suse.de>
6167 PR tree-optimization/84650
6168 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6169 if executed in the loop pipeline.
6171 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6173 * doc/configfiles.texi (Configuration Files): Move info about
6174 conditionalizing $target-protos.h to...
6175 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6176 differs from $target-protos.h.
6178 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6179 Chung-Ju Wu <jasonwucj@gmail.com>
6181 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6182 * config/nds32/nds32-multiple.md (setmemsi): Define.
6183 * config/nds32/nds32-memory-manipulation.c
6184 (nds32_gen_dup_4_byte_to_word_value): New.
6185 (emit_setmem_word_loop): New.
6186 (emit_setmem_byte_loop): New.
6187 (nds32_expand_setmem_loop): New.
6188 (nds32_expand_setmem_loop_v3m): New.
6189 (nds32_expand_setmem_unroll): New.
6190 (nds32_expand_setmem): New.
6192 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6193 Chung-Ju Wu <jasonwucj@gmail.com>
6195 * config/nds32/nds32-memory-manipulation.c
6196 (nds32_emit_load_store): New.
6197 (nds32_emit_post_inc_load_store): New.
6198 (nds32_emit_mem_move): New.
6199 (nds32_emit_mem_move_block): New.
6200 (nds32_expand_movmemsi_loop_unknown_size): New.
6201 (nds32_expand_movmemsi_loop_known_size): New.
6202 (nds32_expand_movmemsi_loop): New.
6203 (nds32_expand_movmemsi_unroll): New.
6204 (nds32_expand_movmemqi): Rename ...
6205 (nds32_expand_movmemsi): ... to this.
6206 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6207 (movmemsi): ... to this.
6208 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6209 (nds32_expand_movmemsi): ... to this.
6211 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6212 Monk Chiang <sh.chiang04@gmail.com>
6213 Chung-Ju Wu <jasonwucj@gmail.com>
6215 * config/nds32/nds32-protos.h
6216 (nds32_expand_load_multiple): New arguments.
6217 (nds32_expand_store_multiple): Ditto.
6218 (nds32_valid_multiple_load_store): Rename ...
6219 (nds32_valid_multiple_load_store_p): ... to this.
6220 * config/nds32/nds32-memory-manipulation.c
6221 (nds32_expand_load_multiple): Refine implementation.
6222 (nds32_expand_store_multiple): Ditto.
6223 * config/nds32/nds32-multiple.md
6224 (load_multiple): Update nds32_expand_load_multiple interface.
6225 (store_multiple): Update nds32_expand_store_multiple interface.
6226 * config/nds32/nds32-predicates.c
6227 (nds32_valid_multiple_load_store): Rename ...
6228 (nds32_valid_multiple_load_store_p): ... to this and refine
6230 * config/nds32/predicates.md
6231 (nds32_load_multiple_and_update_address_operation): New predicate.
6232 (nds32_store_multiple_and_update_address_operation): New predicate.
6234 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6235 Chung-Ju Wu <jasonwucj@gmail.com>
6237 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6238 (combo): New attribute.
6239 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6241 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6243 * config/nds32/nds32.opt: Change -mcmodel= default value.
6245 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6246 Monk Chiang <sh.chiang04@gmail.com>
6247 Chung-Ju Wu <jasonwucj@gmail.com>
6249 * config/nds32/constants.md (unspec_element): New enum.
6250 * config/nds32/constraints.md (Umw): New constraint.
6251 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6252 * config/nds32/nds32-intrinsic.md: Likewise.
6253 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6254 (nds32_valid_smw_lwm_base_p): New.
6255 (nds32_output_smw_single_word): New.
6256 (nds32_output_lmw_single_word): New.
6257 (nds32_expand_unaligned_load): New.
6258 (nds32_expand_unaligned_store): New.
6259 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6260 (nds32_output_smw_single_word): Declare.
6261 (nds32_output_lmw_single_word): Declare.
6262 (nds32_expand_unaligned_load): Declare.
6263 (nds32_expand_unaligned_store): Declare.
6264 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6265 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6266 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6267 NDS32_BUILTIN_UASTORE_DW.
6268 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6271 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6272 Kito Cheng <kito.cheng@gmail.com>
6273 Chung-Ju Wu <jasonwucj@gmail.com>
6275 * config/nds32/nds32-intrinsic.c
6276 (nds32_expand_builtin_null_ftype_reg): Delete.
6277 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6278 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6279 (nds32_read_argument): New.
6280 (nds32_legitimize_target): Ditto.
6281 (nds32_legitimize_argument): Ditto.
6282 (nds32_check_constant_argument): Ditto.
6283 (nds32_expand_unop_builtin): Ditto.
6284 (nds32_expand_unopimm_builtin): Ditto.
6285 (nds32_expand_binop_builtin): Ditto.
6286 (nds32_builtin_decl_impl): Ditto.
6287 (builtin_description): Ditto.
6288 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6289 (nds32_init_builtins_impl): Ditto.
6290 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6291 (nds32_builtin_decl): New.
6292 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6293 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6295 2018-03-02 Jeff Law <law@redhat.com>
6297 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6298 (redundant_insn, fill_simple_delay_slots): Likewise.
6299 (fill_slots_from_thread): Likewise.
6300 * resource.c (mark_referenced_resources): Likewise.
6301 (mark_set_resources, find_dead_or_set_registers): Likewise.
6303 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6305 * substring-locations.h (format_warning_va): Formatting fix for
6307 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6309 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6310 * substring-locations.c: Include intl.h.
6311 (format_warning_va): Turned into small wrapper around
6312 format_warning_n_va, renamed to ...
6313 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6314 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6316 (format_warning_at_substring_n): New function.
6317 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6318 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6319 format_warning_at_substring with just a shorter name instead of
6320 const function pointer.
6321 (fmtwarn_n): New function.
6322 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6323 appropriate, get rid of all the fmtstr temporaries, move conditionals
6324 with G_() wrapped string literals directly into fmtwarn arguments,
6325 cast dir.len to (int), formatting fixes.
6327 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6329 * doc/invoke.texi: Remove "Cilk Plus" references.
6331 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6332 Richard Biener <rguenther@suse.de>
6335 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6336 for error or warning attributes if CALL_FROM_THUNK_P is set.
6339 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6342 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6343 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6346 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6347 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6349 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6350 instead of -1U in last predictors element's probability member.
6352 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6355 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6356 arguments if they are comparable.
6358 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6360 PR tree-optimization/84634
6361 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6362 masks and masked_loop_p with a single loop_masks, making sure it's
6363 null for bb vectorization.
6365 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6367 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6368 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6369 loop->force_vectorize to check whether there is no alias.
6371 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6374 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6376 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6378 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6379 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6380 instead of a loop around prev_real_insn.
6381 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6385 * config/i386/i386.c (ix86_print_operand): Use conditional
6386 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6389 2018-03-02 Richard Biener <rguenther@suse.de>
6391 PR tree-optimization/84427
6392 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6393 (bitmap_set_subtract_values): Rewrite to handle multiple
6396 (prune_clobbered_mems): Likewise.
6397 (phi_translate): Take edge instead of pred/phiblock.
6398 (phi_translate_1): Likewise.
6399 (phi_translate_set): Likewise. Insert all translated
6400 exprs for a value into the set, keeping possibly multiple
6401 expressions per value.
6402 (compute_antic_aux): Adjust for phi_translate changes.
6403 When intersecting union the expressions and prune those
6404 not in the final value set, keeping possibly multiple
6405 expressions per value. Do not use value-insertion
6406 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6407 all expressions. Add verification that the value-sets
6408 only shrink during iteration.
6409 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6410 (do_pre_regular_insertion): Likewise.
6411 (do_pre_partial_partial_insertion): Likewise.
6413 2018-03-02 Richard Biener <rguenther@suse.de>
6416 * config/darwin.c (saved_debug_info_level): New static global.
6417 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6418 (darwin_asm_lto_end): Restore debug info generation settings.
6420 2018-03-01 Martin Liska <mliska@suse.cz>
6423 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6426 2018-03-01 Richard Biener <rguenther@suse.de>
6429 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6430 type annotation with LTO when debug was disabled at compile-time.
6432 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6434 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6436 (mips_final_postscan_insn): Likewise.
6438 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6440 PR rtl-optimization/84528
6441 * alias.c (init_alias_target): Add commentary.
6442 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6443 a unique base value if the frame pointer is not eliminated
6444 to the stack pointer.
6446 2018-03-01 Tom de Vries <tom@codesourcery.com>
6448 PR rtl-optimization/83327
6449 * lra-int.h (hard_regs_spilled_into): Declare.
6450 * lra.c (hard_regs_spilled_into): Define.
6451 (init_reg_info): Init hard_regs_spilled_into.
6452 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6453 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6454 (process_bb_lives): Handle hard_regs_spilled_into.
6455 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6456 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6458 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6460 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6461 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6462 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6463 * config/rs6000/aix72.h: New file.
6465 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6467 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6468 instead of warning_at with conditional singular and plural messages
6472 * stor-layout.c (update_alignment_for_field): For
6473 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6474 && !DECL_PACKED (field), do the alignment update, just use
6475 only desired_align instead of MAX (type_align, desired_align)
6477 (place_field): Don't do known_align < desired_align handling
6478 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6479 is non-NULL, instead do it after rli->prev_field handling and
6480 only if not within a bitfield word. For DECL_PACKED (field)
6481 use type_align of BITS_PER_UNIT.
6483 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6485 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6486 superfluous parentheses and trailing spaces.
6488 2018-02-28 Richard Biener <rguenther@suse.de>
6490 PR tree-optimization/84584
6491 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6492 SCoPs with fake exit edge.
6494 2018-02-28 Martin Liska <mliska@suse.cz>
6497 * timevar.c (timer::print): Fix format to properly print 100%
6500 2018-02-28 Richard Biener <rguenther@suse.de>
6503 * genmatch.c (capture_info::walk_match): Do not mark
6504 captured expressions without operands as expr_p given
6505 they act more like predicates and should be subject to
6506 "lost tail" side-effect preserving.
6508 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6510 PR rtl-optimization/81611
6511 * auto-inc-dec.c (attempt_change): Move dead note from
6512 mem_insn if it's the next use of regno
6513 (find_address): Take address use of reg holding
6514 non-incremented value. Add parm to limit search to the named
6516 (merge_in_block): Attempt to use a mem insn that is the next
6517 use of the original regno.
6519 2018-02-27 Martin Sebor <msebor@redhat.com>
6522 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6523 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6525 2018-02-27 Martin Sebor <msebor@redhat.com>
6527 PR translation/84207
6528 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6529 n argument to unsigned HOST_WIDE_INT.
6530 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6531 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6532 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6533 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6535 2018-02-27 Richard Biener <rguenther@suse.de>
6537 PR tree-optimization/84512
6538 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6539 Do not use the estimate returned from record_stmt_cost for
6540 the scalar iteration cost but sum properly using add_stmt_cost.
6542 2018-02-27 Richard Biener <rguenther@suse.de>
6544 PR tree-optimization/84466
6545 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6546 Adjust last change to less strictly validate use operands.
6548 2018-02-27 Martin Liska <mliska@suse.cz>
6550 PR gcov-profile/84548
6551 * gcov.c (process_file): Allow partial overlap and consider it
6552 also as group functions.
6553 (output_lines): Properly calculate range of lines for a group.
6555 2018-02-27 Martin Liska <mliska@suse.cz>
6557 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6558 'ggc' suffixes. Change first column width.
6559 (timer::print): Fix formatting of the column.
6561 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6563 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6564 preserve inline entry blocks for the sake of debug inline
6565 entry point markers alone.
6566 (remove_unused_locals): Suggest in comments a better place to
6567 force the preservation of inline entry blocks that are
6568 otherwise unused, but do not preserve them.
6570 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6572 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6574 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6577 * config/i386/constraints.md (Bs): Replace
6578 ix86_indirect_branch_register with
6579 TARGET_INDIRECT_BRANCH_REGISTER.
6581 * config/i386/i386.md (indirect_jump): Likewise.
6582 (tablejump): Likewise.
6583 (*sibcall_memory): Likewise.
6584 (*sibcall_value_memory): Likewise.
6585 Peepholes of indirect call and jump via memory: Likewise.
6586 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6587 (*sibcall_value_GOT_32): Likewise.
6588 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6589 (GOT_memory_operand): Likewise.
6590 (call_insn_operand): Likewise.
6591 (sibcall_insn_operand): Likewise.
6592 (GOT32_symbol_operand): Likewise.
6593 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6595 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6597 PR rtl-optimization/83496
6598 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6599 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6600 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6601 redundant insn, if any.
6602 (relax_delay_slots): Likewise.
6603 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6605 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6607 PR tree-optimization/83965
6608 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6609 that grouped statements are part of a reduction chain. Return
6610 true if the statement is not marked as a reduction itself but
6612 (vect_recog_dot_prod_pattern): Don't check whether the statement
6613 is part of a group here.
6614 (vect_recog_sad_pattern): Likewise.
6615 (vect_recog_widen_sum_pattern): Likewise.
6617 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6620 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6622 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6625 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6627 (ix86_output_indirect_function_return): New prototype.
6628 (ix86_split_simple_return_pop_internal): Likewise.
6629 * config/i386/i386.c (indirect_return_via_cx): New.
6630 (indirect_return_via_cx_bnd): Likewise.
6631 (indirect_thunk_name): Handle return va CX_REG.
6632 (output_indirect_thunk_function): Create alias for
6633 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6634 (ix86_output_indirect_jmp): Remove the bool argument.
6635 (ix86_output_indirect_function_return): New function.
6636 (ix86_split_simple_return_pop_internal): Likewise.
6637 * config/i386/i386.md (*indirect_jump): Don't pass false
6638 to ix86_output_indirect_jmp.
6639 (*tablejump_1): Likewise.
6640 (simple_return_pop_internal): Change it to define_insn_and_split.
6641 Call ix86_split_simple_return_pop_internal to split it for
6643 (simple_return_indirect_internal): Call
6644 ix86_output_indirect_function_return instead of
6645 ix86_output_indirect_jmp.
6647 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6650 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6651 memset and value initialization afterwards.
6653 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6655 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6657 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6660 * common/config/aarch64/aarch64-common.c
6661 (aarch_option_optimization_table[]): Switch
6662 off fomit-frame-pointer
6664 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6665 Chung-Ju Wu <jasonwucj@gmail.com>
6667 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6669 (store_multiple): Ditto.
6671 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6673 * config.gcc: Add --with-cpu support for nds32 target.
6674 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6675 * config/nds32/nds32.opt: Add -mcpu= option.
6677 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6679 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6680 isel=yes): Warn for these deprecated options.
6682 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6684 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6685 ISA_2_5_MASKS_EMBEDDED.
6687 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6689 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6690 p->max as pointers rather than using iterative_hash_expr.
6692 2018-02-23 Carl Love <cel@us.ibm.com>
6694 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6695 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6697 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6698 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6699 P8V_BUILTIN_VEC_VUNSIGNED2.
6701 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6704 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6705 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6707 * lra-constraints.c (curr_insn_transform): Set up
6708 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6710 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6711 * lra-eliminations.c (spill_pseudos): Ditto.
6712 (process_insn_for_elimination): Ditto.
6713 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6714 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6715 LRA_NON_CLOBBERED_ALT.
6717 2018-02-22 Martin Sebor <msebor@redhat.com>
6719 PR tree-optimization/84480
6720 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6721 to maybe_diag_stxncpy_trunc. Call it.
6722 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6723 from gimple_fold_builtin_strcpy. Print inlining stack.
6724 (handle_builtin_stxncpy): Print inlining stack.
6725 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6727 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6730 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6731 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6732 and -fcheck-pointer-bounds are used together.
6733 (indirect_thunk_prefix): New enum.
6734 (indirect_thunk_need_prefix): New function.
6735 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6736 "_nt" instead of "_bnd" for NOTRACK prefix.
6737 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6738 (output_indirect_thunk_function): Likewise.
6740 (ix86_code_end): Update output_indirect_thunk_function calls.
6741 (ix86_output_indirect_branch_via_reg): Replace
6742 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6743 (ix86_output_indirect_branch_via_push): Likewise.
6744 (ix86_output_function_return): Likewise.
6745 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6746 incompatible with -fcf-protection=branch and
6747 -fcheck-pointer-bounds.
6749 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6752 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6753 Change gcc_assert call to output_operand_lossage.
6755 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6757 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6759 2018-02-22 DJ Delorie <dj@redhat.com>
6760 Sebastian Perta <sebastian.perta@renesas.com>
6761 Oleg Endo <olegendo@gcc.gnu.org>
6763 * config/rx/rx.c (rx_rtx_costs): New function.
6764 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6766 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6768 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6770 2018-02-22 Martin Liska <mliska@suse.cz>
6773 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6774 Add "native" as a possible value.
6776 2018-02-22 Martin Liska <mliska@suse.cz>
6779 * config/i386/i386.c (ix86_option_override_internal):
6780 Add "native" as a possible value for -march and -mtune.
6782 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6785 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6786 to all type variants.
6788 PR tree-optimization/84503
6789 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6790 width as info->bitpos + info->bitsize - start.
6791 (merged_store_group::merge_overlapping): Simplify width computation.
6792 (check_no_overlap): New function.
6793 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6794 start + width and last_order of the group, fail if check_no_overlap
6796 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6797 to group if check_no_overlap fails.
6799 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6801 * config/rs6000/altivec.md: Delete contraint arguments to
6802 define_expand, define_split, and define_peephole2, and in
6803 define_insn_and_split if always unused.
6804 * config/rs6000/darwin.md: Ditto.
6805 * config/rs6000/dfp.md: Ditto.
6806 * config/rs6000/rs6000.md: Ditto.
6807 * config/rs6000/sync.md: Ditto.
6808 * config/rs6000/vector.md: Ditto.
6809 * config/rs6000/vsx.md: Ditto.
6811 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6813 * config/rs6000/altivec.md: Write output control strings as braced
6814 blocks instead of double-quoted strings.
6815 * config/rs6000/darwin.md: Ditto.
6816 * config/rs6000/rs6000.md: Ditto.
6817 * config/rs6000/vector.md: Ditto.
6818 * config/rs6000/vsx.md: Ditto.
6820 2018-02-21 Jason Merrill <jason@redhat.com>
6822 PR c++/84314 - ICE with templates and fastcall attribute.
6823 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6825 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6827 * ipa-cp.c (determine_versionability): Fix comment typos.
6829 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6832 * ipa-cp.c (determine_versionability): Do not version functions caling
6835 2018-02-21 Martin Liska <mliska@suse.cz>
6838 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6839 Add "native" as a possible value.
6840 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6841 the macro when native cpu detection is available.
6843 2018-02-21 Martin Liska <mliska@suse.cz>
6846 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6847 Add "native" as a possible value.
6848 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6849 when native cpu detection is available.
6851 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6852 Martin Sebor <msebor@redhat.com>
6854 PR tree-optimization/84478
6855 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6857 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6858 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6859 support which is conservatively correct, for 2 only stay conservative
6860 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6861 argument to the 2 argument get_range_strlen, adjust 6 arg
6862 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6864 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6865 (gimple_fold_builtin_strlen): Pass true as last argument to
6868 2018-02-20 Martin Sebor <msebor@redhat.com>
6871 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6872 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6873 (builtin_memref::builtin_memref): Factor out parts into
6874 set_base_and_offset and call it.
6876 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6879 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6880 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6881 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6882 search at the associated MODE_INT.
6884 2018-02-20 Jeff Law <law@redhat.com>
6887 PR tree-optimization/81592
6889 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6890 for range data rather than using global data.
6891 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6892 range data rather than using global data.
6893 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6894 pass it to children as needed.
6895 (struct directive::fmtresult): Similarly.
6896 (struct directive::set_width): Similarly.
6897 (struct directive::set_precision): Similarly.
6898 (format_integer, format_directive, parse_directive): Similarly.
6899 (format_none): Accept unnamed vr_values parameter.
6900 (format_percent, format_floating, format_character): Similarly.
6901 (format_string, format_plain): Similarly.
6902 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6903 the EVRP range analyzer for range data rather than using global data.
6904 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6905 gimple-ssa-evrp-analyze.h
6906 (class sprintf_dom_walker): Add after_dom_children member function.
6907 Add evrp_range_analyzer member.
6908 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6909 range analyzer as needed.
6910 (sprintf_dom_walker::after_dom_children): New member function.
6911 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6913 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6914 (evrp_range_analyzer::pop_to_marker): Likewise.
6916 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6918 PR tree-optimization/84419
6919 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6920 with the required type if its current type is compatible but
6923 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6926 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6927 after vectorization.
6929 2018-02-20 Martin Liska <mliska@suse.cz>
6932 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6933 possible values if we don't have a hint.
6935 2018-02-20 Martin Liska <mliska@suse.cz>
6939 * final.c (shorten_branches): Build align_tab array with one
6941 * opts.c (finish_options): Add alignment option limit check.
6942 (MAX_CODE_ALIGN): Likewise.
6943 (MAX_CODE_ALIGN_VALUE): Likewise.
6944 * doc/invoke.texi: Document maximum allowed option value for
6945 all -falign-* options.
6947 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6950 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6951 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6952 * var-tracking.c (emit_note_insn_var_location): Remove all references
6953 to NOTE_INSN_CALL_ARG_LOCATION.
6954 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6955 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6956 Use copy_rtx_if_shared.
6957 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6958 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6959 (dwarf2out_var_location): Remove handling of
6960 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6962 * final.c (final_scan_insn): Remove all references to
6963 NOTE_INSN_CALL_ARG_LOCATION.
6964 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6965 before dumping final insns.
6966 * except.c (emit_note_eh_region_end): Remove all references to
6967 NOTE_INSN_CALL_ARG_LOCATION.
6968 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6969 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6970 * config/arc/arc.c (hwloop_optimize): Likewise.
6971 * config/arm/arm.c (create_fix_barrier): Likewise.
6972 * config/s390/s390.c (s390_chunkify_start): Likewise.
6973 * config/sh/sh.c (find_barrier): Likewise.
6974 * config/i386/i386.c (rest_of_insert_endbranch,
6975 ix86_seh_fixup_eh_fallthru): Likewise.
6976 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6977 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6978 * config/frv/frv.c (frv_function_prologue): Likewise.
6979 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6981 (note_outside_basic_block_p): Remove all references to
6982 NOTE_INSN_CALL_ARG_LOCATION.
6983 * gengtype.c (adjust_field_rtx_def): Likewise.
6984 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6986 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6987 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6990 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6993 PR tree-optimization/84452
6994 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6995 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6998 2018-02-19 Martin Liska <mliska@suse.cz>
7001 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7003 2018-02-19 Martin Liska <mliska@suse.cz>
7004 Richard Sandiford <richard.sandiford@linaro.org>
7006 PR tree-optimization/82491
7007 * gimple-fold.c (get_base_constructor): Make earlier bail out
7010 2018-02-19 Carl Love <cel@us.ibm.com>
7012 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7013 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7015 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7016 P8V_BUILTIN_VEC_NEG.
7018 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7020 * config/rl78/rl78.md (movdf): New define expand.
7022 2018-02-19 Martin Liska <mliska@suse.cz>
7025 * doc/invoke.texi: Fix typo.
7026 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7028 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7030 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7031 handle rs6000_single_float and rs6000_double_float specially for
7034 2018-02-16 Jeff Law <law@redhat.com>
7036 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7039 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7042 * ipa-inline.c (inline_small_functions): Fix a typo.
7044 2018-02-16 Nathan Sidwell <nathan@acm.org>
7046 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7048 2018-02-16 Carl Love <cel@us.ibm.com>
7050 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7051 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7052 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7053 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7054 expansion to P8V_BUILTIN_VEC_FLOAT2.
7056 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7058 PR rtl-optimization/70023
7059 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7060 src_regno into account.
7062 2018-02-16 Carl Love <cel@us.ibm.com>
7064 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7065 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7066 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7067 * config/rs6000/rs6000.c: Remove case statements for
7068 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7069 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7070 and P9V_BUILTIN_VEC_VINSERT4B.
7071 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7072 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7073 * config/rs6000/vsx.md:
7074 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7077 2018-02-16 Carl Love <cel@us.ibm.com>
7079 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7081 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7083 * config/rs6000/rs6000-c.c: Add the definitions for
7084 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7085 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7086 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7087 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7088 definition for insert4b and define insn *insert3b_internal.
7089 * doc/extend.texi: Add documentation for vec_extract4b.
7091 2018-02-16 Nathan Sidwell <nathan@acm.org>
7093 * doc/extend.texi (Backwards Compatibility): Mention friend
7094 injection. Note for-scope is deprecated.
7095 * doc/invoke.texi (-ffriend-injection): Deprecate.
7097 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7099 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7100 that moved to I2, also allow destinations that are a paradoxical
7101 subreg (instead of a normal reg).
7103 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7106 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7109 2018-02-16 Richard Biener <rguenther@suse.de>
7111 PR tree-optimization/84037
7112 PR tree-optimization/84016
7114 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7115 Adjust vec_construct for the fact we need additional higher latency
7116 128bit inserts for AVX256 and AVX512 vector builds.
7117 (ix86_add_stmt_cost): Scale vector construction cost for
7120 2018-02-16 Richard Biener <rguenther@suse.de>
7122 PR tree-optimization/84417
7123 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7124 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7125 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7127 2018-02-16 Martin Liska <mliska@suse.cz>
7130 * internal-fn.def (ASAN_CHECK): Set proper flags.
7131 (ASAN_MARK): Likewise.
7133 2018-02-16 Julia Koval <julia.koval@intel.com>
7135 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7136 from PTA_CANNONLAKE.
7138 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7141 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7142 Use ++iter rather than iter++ for std::list iterators.
7143 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7144 defer deleting them until all nodes in the forest are processed. Do
7145 free even leaf nodes. Change to_process into auto_vec.
7148 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7149 * vec.h (vec_default_construct): Use memset instead of placement new
7150 if BROKEN_VALUE_INITIALIZATION is defined.
7151 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7152 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7155 PR rtl-optimization/83723
7156 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7157 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7158 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7160 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7162 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7164 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7166 PR rtl-optimization/81443
7167 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7168 from inner REGs to paradoxical SUBREGs.
7170 2018-02-16 Richard Biener <rguenther@suse.de>
7172 PR tree-optimization/84399
7173 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7174 For operands we can analyze at their definition make sure we can
7175 analyze them at each use as well.
7177 2018-02-16 Richard Biener <rguenther@suse.de>
7179 PR tree-optimization/84190
7180 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7181 volatile accesses if the decl isn't volatile.
7183 2018-02-15 Jason Merrill <jason@redhat.com>
7185 PR c++/84314 - ICE with templates and fastcall attribute.
7186 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7187 TYPE_CANONICAL on an existing type.
7189 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7191 PR tree-optimization/84383
7192 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7193 dstoff nor call operand_equal_p if dstbase is NULL.
7195 PR tree-optimization/84334
7196 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7197 also a CONSTANT_CLASS_P, punt.
7199 2018-02-14 Jim Wilson <jimw@sifive.com>
7201 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7202 first SMALL_OPERAND check. New local min_second_step. Move assert
7203 to where locals are set. Add TARGET_RVC support.
7204 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7206 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7208 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7210 2018-02-14 Martin Sebor <msebor@redhat.com>
7212 PR tree-optimization/83698
7213 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7214 arrays constrain the offset range to their bounds.
7215 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7216 (builtin_access::overlap): Avoid setting the size of overlap if it's
7218 (maybe_diag_overlap): Also consider arrays when deciding what values
7219 of offsets to include in diagnostics.
7221 2018-02-14 Martin Sebor <msebor@redhat.com>
7224 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7225 that correspond to the kind of a declaration.
7227 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7230 * config/pa/pa.md: Load address of PIC label using the linkage table
7231 if the label is nonlocal.
7233 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7235 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7236 warning message if user requests -maltivec=be.
7237 * doc/invoke.texi: Document deprecation of -maltivec=be.
7239 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7242 * config/rs6000/rs6000-c.c: Update definitions for
7243 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7244 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7246 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7249 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7250 add _get_ssp intrinsics. Remove argument from
7251 __builtin_ia32_rdssp[d|q].
7252 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7253 * config/i386/i386-builtin.def: Remove argument from
7254 __builtin_ia32_rdssp[d|q].
7255 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7256 ix86_expand_special_args_builtin for _rdssp[d|q].
7257 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7258 Clear register before usage.
7259 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7260 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7262 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7264 PR tree-optimization/84357
7265 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7266 operand 1 of an ARRAY_REF too.
7268 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7271 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7272 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7274 (set_of_reg): New struct.
7275 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7276 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7277 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7279 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7280 Split into bitclr, bitset, bitinvert patterns if appropriate.
7281 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7282 use rx_fuse_in_memory_bitop.
7283 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7284 to named insn, correct maximum insn length.
7286 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7289 * machmode.def: Define a complex mode for PARTIAL_INT.
7290 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7292 * doc/rtl.texi: Document CSPImode.
7293 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7295 (msp430_hard_regno_nregs_with_padding): Likewise.
7297 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7300 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7302 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7304 PR rtl-optimization/84169
7305 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7306 we generated a parallel as new i3 and we split that to new i2 and i3
7307 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7308 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7309 those to i2, not i1. Partially rewrite this scan code.
7311 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7314 * stor-layout.c (place_field): For variable length fields, adjust
7315 offset_align afterwards not just based on the field's alignment,
7316 but also on the size.
7319 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7320 of exps and logs in the use_exp2 case.
7322 2018-02-13 Jeff Law <law@redhat.com>
7324 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7327 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7330 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7334 * common.opt (gas-loc-support, gas-locview-support): New.
7335 (ginline-points, ginternal-reset-location-views): New.
7336 * doc/invoke.texi: Document them. Use @itemx where intended.
7337 (gvariable-location-views): Adjust.
7338 * target.def (reset_location_view): New.
7339 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7340 (TARGET_RESET_LOCATION_VIEW): New.
7341 * doc/tm.texi: Rebuilt.
7342 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7343 (dwarf2out_default_as_locview_support): New.
7344 (output_asm_line_debug_info): Use option variables.
7345 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7346 (output_loc_list): Likewise.
7347 (add_high_low_attributes): Check option variables.
7348 Don't output entry view attribute in strict mode.
7349 (gen_inlined_subroutine_die): Check option variables.
7350 (dwarf2out_inline_entry): Likewise.
7351 (init_sections_and_labels): Likewise.
7352 (dwarf2out_early_finish): Likewise.
7353 (maybe_reset_location_view): New, from...
7354 (dwarf2out_var_location): ... here. Call it.
7355 * debug.h (dwarf2out_default_as_loc_support): Declare.
7356 (dwarf2out_default_as_locview_support): Declare.
7357 * hooks.c (hook_int_rtx_insn_0): New.
7358 * hooks.h (hook_int_rtx_insn_0): Declare.
7359 * toplev.c (process_options): Take -gas-loc-support and
7360 -gas-locview-support from dwarf2out. Enable
7361 -gvariable-location-views by default only with locview
7362 assembler support. Enable -ginternal-reset-location-views by
7363 default only if the target defines the corresponding hook.
7364 Enable -ginline-points by default if location views are
7365 enabled; force it disabled if statement frontiers are
7367 * tree-inline.c (expand_call_inline): Check option variables.
7368 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7370 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7372 PR tree-optimization/84321
7373 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7374 handling. Also check whether the anti-range contains any values
7375 that satisfy the mask; switch to a VR_RANGE if not.
7377 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7380 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7382 2018-02-13 Martin Jambor <mjambor@suse.cz>
7385 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7386 of call statements, also set location of a load to a temporary.
7388 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7390 * config/rl78/rl78.c (add_vector_labels): New function.
7391 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7392 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7393 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7394 which checks that no arguments are passed.
7395 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7396 * doc/extend.texi: Documentation for the new attribute.
7398 2018-02-13 Andreas Schwab <schwab@suse.de>
7400 * config/riscv/linux.h (CPP_SPEC): Define.
7402 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7405 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7406 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7407 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7408 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7409 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7410 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7411 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7412 temporarily for AES and PCLMUL builtins.
7414 PR tree-optimization/84339
7415 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7416 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7420 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7421 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7422 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7424 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7426 * omp-simd-clone.h: New file.
7427 * omp-simd-clone.c: Include omp-simd-clone.h.
7428 (expand_simd_clones): No longer static.
7429 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7430 cgraph.h and omp-simd-clone.h.
7431 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7432 (vect_recog_widen_shift_pattern): Formatting fix.
7433 (vect_pattern_recog_1): Don't check optab for calls.
7436 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7437 operands[2] into a REG before using gen_lowpart on it.
7439 2018-02-12 Jeff Law <law@redhat.com>
7442 * config/sh/sh.c (find_barrier): Consider a sibling call
7445 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7446 successfully back substituting a reg.
7448 2018-02-12 Richard Biener <rguenther@suse.de>
7450 PR tree-optimization/84037
7451 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7452 parameter, move visited init to caller.
7453 (vect_slp_analyze_operations): Separate cost from validity
7454 check, initialize visited once for all instances.
7455 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7457 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7459 (vect_model_promotion_demotion_cost): Likewise.
7460 (vectorizable_bswap): Guard cost modeling with !slp_node
7461 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7463 (vectorizable_call): Likewise.
7464 (vectorizable_conversion): Likewise.
7465 (vectorizable_assignment): Likewise.
7466 (vectorizable_shift): Likewise.
7467 (vectorizable_operation): Likewise.
7468 (vectorizable_store): Likewise.
7469 (vectorizable_load): Likewise.
7470 (vectorizable_condition): Likewise.
7471 (vectorizable_comparison): Likewise.
7473 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7476 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7477 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7478 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7480 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7483 * params.def (inline-min-speedup): Increase from 8 to 15.
7484 (max-inline-insns-auto): Decrease from 40 to 30.
7485 * ipa-split.c (consider_split): Add some buffer for function to
7486 be considered inlining candidate.
7487 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7490 2018-02-12 Richard Biener <rguenther@suse.de>
7492 PR tree-optimization/84037
7493 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7494 matched stmts if we cannot swap the non-matched ones.
7496 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7498 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7499 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7500 _mm_maskz_scalef_round_ss): New intrinsics.
7501 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7502 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7503 __builtin_ia32_scalefss_round): Remove.
7504 (__builtin_ia32_scalefsd_mask_round,
7505 __builtin_ia32_scalefss_mask_round): New intrinsics.
7506 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7507 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7508 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7509 "<round_constraint>")): Changed to ...
7510 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7511 "<round_scalar_constraint>")): ... this.
7512 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7513 %0, %1, %2<round_op3>}"): Changed to ...
7514 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7515 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7516 %2<round_scalar_mask_op3>}"): ... this.
7517 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7519 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7521 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7522 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7523 (_mm_maskz_sqrt_round_ss): New intrinsics.
7524 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7525 (__builtin_ia32_sqrtsd_mask_round)
7526 (__builtin_ia32_sqrtss_mask_round): New builtins.
7527 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7528 (__builtin_ia32_sqrtss_round): Remove.
7529 (__builtin_ia32_sqrtsd_mask_round)
7530 (__builtin_ia32_sqrtss_mask_round): New builtins.
7531 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7532 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7533 ((match_operand:VF_128 1 "vector_operand"
7534 "xBm,<round_constraint>")): Changed to ...
7535 ((match_operand:VF_128 1 "vector_operand"
7536 "xBm,<round_scalar_constraint>")): ... this.
7537 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7538 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7539 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7540 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7541 %<iptr>1<round_scalar_mask_op3>}): ... this.
7542 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7543 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7545 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7548 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7549 Cast vec_cmpeq result to correct type.
7550 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7551 Cast vec_cmpgt result to correct type.
7553 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7555 * final.c (final_scan_insn_1): Renamed from...
7556 (final_scan_insn): ... this. New wrapper, to recover
7557 seen from the outermost call in recursive ones.
7558 * config/sparc/sparc.c (output_return): Drop seen from call.
7559 (output_sibcall): Likewise.
7560 * config/visium/visium.c (output_branch): Likewise.
7562 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7564 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7567 2018-02-10 Alan Modra <amodra@gmail.com>
7570 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7571 Specify LR as an input.
7573 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7576 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7577 remove_member_access_dummy_vars): New functions.
7578 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7579 lower_omp_1, execute_lower_omp): Use them.
7581 PR rtl-optimization/84308
7582 * shrink-wrap.c (spread_components): Release todo vector.
7584 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7586 PR rtl-optimization/57193
7587 * ira-color.c (struct allocno_color_data): Add member
7588 conflict_allocno_hard_prefs.
7589 (update_conflict_allocno_hard_prefs): New.
7590 (bucket_allocno_compare_func): Add a preference based on
7591 conflict_allocno_hard_prefs.
7592 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7593 (color_allocnos): Remove a dead code. Initiate
7594 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7596 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7599 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7600 constraint from =wa to wa. Avoid a subreg on the output operand,
7601 instead use a pseudo and subreg it in a move.
7602 (p9_xxbrd_<mode>): Changed to ...
7603 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7604 (p9_xxbrd_v2df): New expander.
7605 (p9_xxbrw_<mode>): Changed to ...
7606 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7607 (p9_xxbrw_v4sf): New expander.
7609 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7611 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7613 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7616 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7617 multiply in 32-bit mode.
7618 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7619 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7622 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7624 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7625 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7626 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7627 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7629 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7632 * dwarf2out.c (is_trivial_indirect_ref): New function.
7633 (dwarf2out_late_global_decl): Do not generate a location
7634 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7635 and that are not defined in the current unit.
7637 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7639 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7640 instead of a libcall for UNORDERED.
7642 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7645 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7646 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7648 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7651 * config/s390/s390.c (s390_set_current_function): Invoke
7652 s390_indirect_branch_settings also if fndecl didn't change.
7654 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7656 * config/rs6000/rs6000.md (blockage): Set length to zero.
7658 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7660 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7662 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7665 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7666 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7670 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7671 PARALLEL incoming that failed vt_get_decl_and_offset check.
7674 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7675 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7677 (get_variable_section): For decls in named .bss* sections pass true as
7678 second argument to bss_initializer_p.
7680 2018-02-09 Marek Polacek <polacek@redhat.com>
7681 Jakub Jelinek <jakub@redhat.com>
7684 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7685 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7686 Sync some changes from cxx_fold_indirect_ref.
7688 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7690 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7692 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7693 (BLOCK_INLINE_ENTRY_LABEL): New.
7694 (dwarf2out_var_location): Disregard inline entry markers.
7695 (inline_entry_data): New struct.
7696 (inline_entry_data_hasher): New hashtable type.
7697 (inline_entry_data_hasher::hash): New.
7698 (inline_entry_data_hasher::equal): New.
7699 (inline_entry_data_table): New variable.
7700 (add_high_low_attributes): Add DW_AT_entry_pc and
7701 DW_AT_GNU_entry_view attributes if a pending entry is found
7702 in inline_entry_data_table. Add old entry_pc attribute only
7703 if debug nonbinding markers are disabled.
7704 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7705 markers are enabled.
7706 (block_within_block_p, dwarf2out_inline_entry): New.
7707 (dwarf2out_finish): Check that no entries remained in
7708 inline_entry_data_table.
7709 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7710 (final_scan_insn, notice_source_line): Likewise.
7711 (rest_of_clean_state): Skip inline entry markers.
7712 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7714 * gimple.c (gimple_build_debug_inline_entry): New.
7715 * gimple.h (enum gimple_debug_subcode): Add
7716 GIMPLE_DEBUG_INLINE_ENTRY.
7717 (gimple_build_debug_inline_entry): Declare.
7718 (gimple_debug_inline_entry_p): New.
7719 (gimple_debug_nonbind_marker_p): Adjust.
7720 * insn-notes.def (INLINE_ENTRY): New.
7721 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7722 inline entry marker notes.
7723 (print_insn): Likewise.
7724 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7725 (INSN_DEBUG_MARKER_KIND): Likewise.
7726 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7727 * tree-inline.c (expand_call_inline): Build and insert
7728 debug_inline_entry stmt.
7729 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7730 inline entry blocks early, if nonbind markers are enabled.
7731 (dump_scope_block): Dump fragment info.
7732 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7733 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7734 (gimple_build_debug_inline_entry): New.
7735 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7736 Enable/disable inline entry points too.
7737 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7738 (DEBUG_INSN): Describe inline entry markers.
7740 * common.opt (gvariable-location-views): New.
7741 (gvariable-location-views=incompat5): New.
7742 * config.in: Rebuilt.
7743 * configure: Rebuilt.
7744 * configure.ac: Test assembler for view support.
7745 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7746 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7747 * dwarf2out.c (var_loc_view): New typedef.
7748 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7749 (dwarf2out_locviews_in_attribute): New.
7750 (dwarf2out_locviews_in_loclist): New.
7751 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7752 (enum dw_line_info_opcode): Add LI_adv_address.
7753 (struct dw_line_info_table): Add view.
7754 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7755 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7756 (zero_view_p): New variable.
7757 (ZERO_VIEW_P): New macro.
7758 (output_asm_line_debug_info): New.
7759 (struct var_loc_node): Add view.
7760 (add_AT_view_list, AT_loc_list): New.
7761 (add_var_loc_to_decl): Add view param. Test it against last.
7762 (new_loc_list): Add view params. Record them.
7763 (AT_loc_list_ptr): Handle loc and view lists.
7764 (view_list_to_loc_list_val_node): New.
7765 (print_dw_val): Handle dw_val_class_view_list.
7766 (size_of_die): Likewise.
7767 (value_format): Likewise.
7768 (loc_list_has_views): New.
7769 (gen_llsym): Set vl_symbol too.
7770 (maybe_gen_llsym, skip_loc_list_entry): New.
7771 (dwarf2out_maybe_output_loclist_view_pair): New.
7772 (output_loc_list): Output view list or entries too.
7773 (output_view_list_offset): New.
7774 (output_die): Handle dw_val_class_view_list.
7775 (output_dwarf_version): New.
7776 (output_compilation_unit_header): Use it.
7777 (output_skeleton_debug_sections): Likewise.
7778 (output_rnglists, output_line_info): Likewise.
7779 (output_pubnames, output_aranges): Update version comments.
7780 (output_one_line_info_table): Output view numbers in asm comments.
7781 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7782 Call maybe_gen_llsym.
7783 (loc_list_from_tree_1): Adjust.
7784 (add_AT_location_description): Create view list attribute if
7785 needed, check it's absent otherwise.
7786 (convert_cfa_to_fb_loc_list): Adjust.
7787 (maybe_emit_file): Call output_asm_line_debug_info for test.
7788 (dwarf2out_var_location): Reset views as needed. Precompute
7789 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7790 attribute. Set view.
7791 (new_line_info_table): Reset next view.
7792 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7793 (dwarf2out_source_line): Likewise. Output view resets and labels to
7794 the assembler, or select appropriate line info opcodes.
7795 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7796 (optimize_string_length): Catch it. Adjust.
7797 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7798 dw_val_class_view_list, and remove it if no longer needed.
7799 (hash_loc_list): Hash view numbers.
7800 (loc_list_hasher::equal): Compare them.
7801 (optimize_location_lists): Check whether a view list symbol is
7802 needed, and whether the locview attribute is present, and
7803 whether they match. Remove the locview attribute if no longer
7805 (index_location_lists): Call skip_loc_list_entry for test.
7806 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7807 Use output_dwarf_version.
7808 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7809 (struct dw_val_node): Add val_view_list.
7810 * final.c (SEEN_NEXT_VIEW): New.
7811 (set_next_view_needed): New.
7812 (clear_next_view_needed): New.
7813 (maybe_output_next_view): New.
7814 (final_start_function): Rename to...
7815 (final_start_function_1): ... this. Take pointer to FIRST,
7816 add SEEN parameter. Emit param bindings in the initial view.
7817 (final_start_function): Reintroduce SEEN-less interface.
7818 (final): Rename to...
7819 (final_1): ... this. Take SEEN parameter. Output final pending
7820 next view at the end.
7821 (final): Reintroduce seen-less interface.
7822 (final_scan_insn): Output pending next view before switching
7823 sections or ending a block. Mark the next view as needed when
7824 outputting variable locations. Notify debug backend of section
7825 changes, and of location view changes.
7826 (rest_of_handle_final): Adjust.
7827 * toplev.c (process_options): Autodetect value for debug variable
7828 location views option. Warn on incompat5 without -gdwarf-5.
7829 * doc/invoke.texi (gvariable-location-views): New.
7830 (gvariable-location-views=incompat5): New.
7831 (gno-variable-location-views): New.
7833 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7835 PR tree-optimization/84136
7836 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7837 that the result of find_edge is non-NULL.
7839 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7842 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7843 storing integer register in SImode. Fix cost of 256 and 512
7844 byte aligned SSE register store.
7846 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7848 * config/i386/i386.c (ix86_multiplication_cost): Fix
7849 multiplication cost for TARGET_AVX512DQ.
7851 2018-02-08 Marek Polacek <polacek@redhat.com>
7853 PR tree-optimization/84238
7854 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7857 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7859 PR tree-optimization/84265
7860 * tree-vect-stmts.c (vectorizable_store): Don't treat
7861 VMAT_CONTIGUOUS accesses as grouped.
7862 (vectorizable_load): Likewise.
7864 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7866 PR tree-optimization/81635
7867 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7868 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7869 (test_round_for_mask): New functions.
7870 (wide_int_cc_tests): Call test_round_for_mask.
7871 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7872 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7873 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7874 range returned by get_range_info.
7876 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7879 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7880 * symtab.c: Include builtins.h
7881 (symtab_node::output_to_lto_symbol_table_p): Move here
7882 from lto-streamer-out.c:output_symbol_p.
7883 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7884 (output_symbol_p): Move all logic to symtab.c
7885 (produce_symtab): Update.
7887 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7889 * config/s390/s390-opts.h (enum indirect_branch): Define.
7890 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7891 (s390_indirect_branch_via_thunk)
7892 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7893 (enum s390_indirect_branch_type): Define.
7894 * config/s390/s390.c (struct s390_frame_layout, struct
7895 machine_function): Remove.
7896 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7897 (indirect_branch_table_label_no, indirect_branch_table_name):
7899 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7900 (enum s390_indirect_branch_option): Define.
7901 (s390_return_addr_from_memory): New function.
7902 (s390_handle_string_attribute): New function.
7903 (s390_attribute_table): Add new attribute handler.
7904 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7905 (s390_indirect_branch_via_thunk): New function.
7906 (s390_indirect_branch_via_inline_thunk): New function.
7907 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7908 sibling call optimization for non z10 compiles.
7909 (s390_emit_call): Force indirect branch target to be a single
7910 register. Add r1 clobber for non-z10 compiles.
7911 (s390_emit_epilogue): Emit return jump via return_use expander.
7912 (s390_reorg): Handle JUMP_INSNs as execute targets.
7913 (s390_option_override_internal): Perform validity checks for the
7914 new command line options.
7915 (s390_indirect_branch_attrvalue): New function.
7916 (s390_indirect_branch_settings): New function.
7917 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7918 (s390_output_indirect_thunk_function): New function.
7919 (s390_code_end): Implement target hook.
7920 (s390_case_values_threshold): Implement target hook.
7921 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7923 * config/s390/s390.h (struct s390_frame_layout)
7924 (struct machine_function): Move here from s390.c.
7925 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7926 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7927 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7928 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7929 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7930 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7931 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7932 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7933 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7934 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7935 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7936 (mnemonic attribute): Add values which aren't recognized
7938 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7939 pattern for branch conversion. Fix mnemonic attribute.
7940 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7941 indirect branch via thunk if requested.
7942 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7943 ("*indirect_jump"): Disable for branch conversion using out of
7945 ("indirect_jump_via_thunk<mode>_z10")
7946 ("indirect_jump_via_thunk<mode>")
7947 ("indirect_jump_via_inlinethunk<mode>_z10")
7948 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7949 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7950 ("casesi_jump_via_inlinethunk<mode>_z10")
7951 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7952 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7953 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7954 ("*indirect2_jump"): Disable for branch conversion.
7955 ("casesi_jump"): Turn into expander and expand patterns for branch
7957 ("return_use"): New expander.
7958 ("*return"): Emit return via thunk and rename it to ...
7959 ("*return<mode>"): ... this one.
7960 * config/s390/s390.opt: Add new options and and enum for the
7963 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7965 * lra-constraints.c (match_reload): Unconditionally use
7966 gen_lowpart_SUBREG, rather than selecting between that
7967 and equivalent gen_rtx_SUBREG code.
7969 2018-02-08 Richard Biener <rguenther@suse.de>
7971 PR tree-optimization/84233
7972 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7973 changed flag instead of boguously re-using phi_inserted.
7975 2018-02-08 Martin Jambor <mjambor@suse.cz>
7977 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7978 static local variables.
7980 2018-02-08 Richard Biener <rguenther@suse.de>
7982 PR tree-optimization/84278
7983 * tree-vect-stmts.c (vectorizable_store): When looking for
7984 smaller vector types to perform grouped strided loads/stores
7985 make sure the mode is supported by the target.
7986 (vectorizable_load): Likewise.
7988 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7990 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7991 Increase LDP/STP opportunities by adding adjacent callee-saves.
7993 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7995 PR rtl-optimization/84068
7996 PR rtl-optimization/83459
7997 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7999 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8001 PR tree-optimization/84224
8002 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8003 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8006 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8009 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8010 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8011 count, remove one USE.
8013 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8015 * doc/install.texi (Configuration): Document the
8016 --with-long-double-format={ibm,ieee} PowerPC configuration
8020 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8021 Convert from define_expand to be define_insn_and_split. Rework
8022 float/double/_Float128 conversions to QI/HI/SImode to work with
8023 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8024 conversions to QI/HImode types did a store and then a load to
8025 truncate the value. For conversions to VSX registers, don't split
8026 the insn, instead emit the code directly. Use the code iterator
8027 any_fix to combine signed and unsigned conversions.
8028 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8029 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8030 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8031 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8032 (fix_<mode>di2_hw): Likewise.
8033 (fixuns_<mode>di2_hw): Likewise.
8034 (fix_<mode>si2_hw): Likewise.
8035 (fixuns_<mode>si2_hw): Likewise.
8036 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8037 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8038 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8039 fix<uns>_trunc<SFDF:mode>si2_p8.
8040 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8042 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8043 (fix<uns>_<mode>_mem): Likewise.
8044 (fctiw<u>z_<mode>_mem): Likewise.
8045 (fix<uns>_<mode>_mem): Likewise.
8046 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8047 the register allocator from doing a direct move to the GPRs to do
8048 a store, and instead use the ISA 3.0 store byte/half-word from
8049 vector register instruction. For IEEE 128-bit floating point,
8050 also optimize stores of 32-bit ints.
8051 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8053 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8055 * genextract.c (push_pathstr_operand): New function to support
8057 (walk_rtx): Call push_pathstr_operand.
8058 (print_path): Support [a-zA-Z].
8060 2018-02-07 Richard Biener <rguenther@suse.de>
8062 PR tree-optimization/84037
8063 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8064 (cse_and_gimplify_to_preheader): Declare.
8065 (vect_get_place_in_interleaving_chain): Likewise.
8066 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8068 (_loop_vec_info::~_loop_vec_info): Delete it.
8069 (cse_and_gimplify_to_preheader): New function.
8070 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8071 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8072 (vectorizable_load): Likewise. For grouped stores always base
8073 the IV on the first element.
8074 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8075 condition before gimplifying.
8077 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8079 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8080 *DIV_EXPR and *MOD_EXPR.
8082 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8085 * config/i386/i386.c (ix86_option_override_internal): Mask out
8086 the CF_SET bit when checking -fcf-protection.
8088 2018-02-07 Tom de Vries <tom@codesourcery.com>
8091 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8094 2018-02-07 Richard Biener <rguenther@suse.de>
8096 PR tree-optimization/84204
8097 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8100 PR tree-optimization/84205
8101 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8102 special-case isl_ast_op_zdiv_r.
8104 PR tree-optimization/84223
8105 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8106 Only add conditions from within the region.
8107 (gather_bbs::after_dom_children): Adjust.
8109 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8112 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8113 * config/avr/avr.md: Only post-reload split REG-REG moves if
8114 either register is GENERAL_REG_P.
8116 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8118 PR tree-optimization/84235
8119 * tree-ssa-scopedtables.c
8120 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8121 if the subtraction is performed in floating point type where NaNs are
8122 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8123 build 1. Formatting fix.
8125 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8128 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8129 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8130 and skip it regardless of bb boundaries. Use CALL_P macro,
8131 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8132 unnecessarily, formatting fix.
8134 2018-02-06 Michael Collison <michael.collison@arm.com>
8136 * config/arm/thumb2.md:
8137 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8138 (*thumb_mov_notscc): Ditto.
8140 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8143 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8144 unsigned_fix, not "s".
8146 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8148 * configure.ac (gcc_fn_eh_frame_ro): New function.
8149 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8150 correct .eh_frame permissions.
8151 * configure: Regenerate.
8153 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8155 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8158 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8160 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8161 Display warning message for -mno-speculate-indirect-jumps.
8163 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8165 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8167 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8169 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8171 PR tree-optimization/84225
8172 * tree-eh.c (find_trapping_overflow): Only call
8173 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8175 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8178 * config/i386/i386.c: Reimplement the check of possible options
8179 -mibt/-mshstk conbination. Change error messages.
8180 * doc/invoke.texi: Fix a typo: remove extra '='.
8182 2018-02-06 Marek Polacek <polacek@redhat.com>
8184 PR tree-optimization/84228
8185 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8187 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8190 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8191 emitted arch directives.
8192 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8193 __ARM_FEATURE_COPROC before changing architectures.
8195 2018-02-06 Richard Biener <rguenther@suse.de>
8197 * config/i386/i386.c (print_reg): Fix typo.
8198 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8200 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8202 * configure: Regenerate.
8204 2018-02-05 Martin Sebor <msebor@redhat.com>
8206 PR tree-optimization/83369
8207 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8210 2018-02-05 Martin Liska <mliska@suse.cz>
8212 * doc/invoke.texi: Cherry-pick upstream r323995.
8214 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8216 * ira.c (ira_init_register_move_cost): Adjust comment.
8218 2018-02-05 Martin Liska <mliska@suse.cz>
8220 PR gcov-profile/84137
8221 * doc/gcov.texi: Fix typo in documentation.
8223 2018-02-05 Martin Liska <mliska@suse.cz>
8225 PR gcov-profile/83879
8226 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8227 using dlopen functionality.
8229 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8231 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8232 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8233 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8234 _mm_maskz_range_round_ss): New intrinsics.
8235 (__builtin_ia32_rangesd128_round)
8236 (__builtin_ia32_rangess128_round): Remove.
8237 (__builtin_ia32_rangesd128_mask_round,
8238 __builtin_ia32_rangess128_mask_round): New builtins.
8239 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8240 __builtin_ia32_rangess128_round): Remove.
8241 (__builtin_ia32_rangesd128_mask_round,
8242 __builtin_ia32_rangess128_mask_round): New builtins.
8243 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8244 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8245 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8246 "<round_saeonly_constraint>")): Changed to ...
8247 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8248 "<round_saeonly_scalar_constraint>")): ... this.
8249 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8250 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8251 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8252 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8253 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8255 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8257 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8259 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8260 Remove all values except native, 8540 and 8548.
8262 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8264 * config/i386/i386.c (ix86_output_function_return): Pass
8265 INVALID_REGNUM, instead of -1, as invalid register number to
8266 indirect_thunk_name and output_indirect_thunk.
8268 2018-02-02 Julia Koval <julia.koval@intel.com>
8270 * config.gcc: Add -march=icelake.
8271 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8272 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8273 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8274 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8275 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8276 (processor_target_table): Add icelake.
8277 (ix86_option_override_internal): Handle new PTAs.
8278 (get_builtin_code_for_version): Handle icelake.
8279 (M_INTEL_COREI7_ICELAKE): New.
8280 (fold_builtin_cpu): Handle icelake.
8281 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8282 * doc/invoke.texi: Add -march=icelake.
8284 2018-02-02 Julia Koval <julia.koval@intel.com>
8286 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8287 to wide_int_bitmask.
8288 * wide-int-bitmask.h: New.
8290 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8293 * config/i386/i386.md: Replace Pmode with word_mode in
8294 builtin_setjmp_setup and builtin_longjmp to support x32.
8296 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8300 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8302 (rs6000_supported_cpu_names): New static variable.
8303 (linux_cpu_translation_table): Likewise.
8304 (elf_platform) <cpu>: Define new static variable and use it.
8305 Translate kernel AT_PLATFORM name to canonical name if needed.
8306 Error if platform name is unknown.
8308 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8311 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8313 2018-02-01 Jeff Law <law@redhat.com>
8316 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8317 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8318 the scratch if RELEASE_VIA_POP is false.
8319 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8320 If we have to save a temporary register, decrement SIZE appropriately.
8321 Pass new arguments to release_scratch_register_on_entry.
8322 (ix86_adjust_stack_and_probe): Likewise.
8323 (ix86_emit_probe_stack_range): Pass new arguments to
8324 release_scratch_register_on_entry.
8326 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8328 PR rtl-optimization/84157
8329 * combine.c (change_zero_ext): Use REG_P predicate in
8330 front of HARD_REGISTER_P predicate.
8332 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8334 * config/avr/avr.c (avr_option_override): Move disabling of
8335 -fdelete-null-pointer-checks to...
8336 * common/config/avr/avr-common.c (avr_option_optimization_table):
8339 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8341 PR tree-optimization/81635
8342 * tree-data-ref.c (split_constant_offset_1): For types that
8343 wrap on overflow, try to use range info to prove that wrapping
8346 2018-02-01 Renlin Li <renlin.li@arm.com>
8349 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8351 (aarch64_register_move_cost): Likewise.
8352 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8354 (REG_CLASS_NAMES): Likewise.
8355 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8356 TAILCALL_ADDR_REGS. Remove IP registers.
8357 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8359 2018-02-01 Richard Biener <rguenther@suse.de>
8361 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8362 for specifying RPO order and allow NULL for that.
8363 * domwalk.c (dom_walker::dom_walker): Likewise.
8364 (dom_walker::walk): Handle NULL RPO order.
8365 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8367 (rewrite_update_dom_walker): Likewise.
8368 (mark_def_dom_walker): Likewise.
8370 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8372 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8373 (aarch64_maybe_expand_sve_subreg_move): Declare.
8374 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8375 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8377 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8378 that are semantically a reverse operation.
8379 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8380 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8381 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8383 (aarch64_can_change_mode_class): For big-endian, forbid changes
8384 between two SVE modes if they have different element sizes.
8386 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8388 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8389 the TImode handling for big-endian targets.
8391 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8393 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8394 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8396 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8397 Remove BSWAP handing for big-endian targets and use the form of
8398 LD1RQ appropariate for the mode.
8400 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8402 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8403 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8406 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8409 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8410 check for operands that need to go through aarch64_sve_reload_be.
8412 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8414 PR tree-optimization/81661
8415 PR tree-optimization/84117
8416 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8417 * tree-eh.c: Include gimplify.h.
8418 (find_trapping_overflow, replace_trapping_overflow,
8419 rewrite_to_non_trapping_overflow): New functions.
8420 * tree-vect-loop.c: Include tree-eh.h.
8421 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8422 * tree-data-ref.c: Include tree-eh.h.
8423 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8425 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8427 PR rtl-optimization/84123
8428 * combine.c (change_zero_ext): Check if hard register satisfies
8429 can_change_dest_mode before calling gen_lowpart_SUBREG.
8431 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8434 * ira.c (ira_init_register_move_cost): Remove assert.
8436 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8438 PR rtl-optimization/84071
8439 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8440 * doc/tm.texi: Regenerate.
8442 2018-01-31 Richard Biener <rguenther@suse.de>
8444 PR tree-optimization/84132
8445 * tree-data-ref.c (analyze_miv_subscript): Properly
8446 check whether evolution_function_is_affine_multivariate_p
8447 before calling gcd_of_steps_may_divide_p.
8449 2018-01-31 Julia Koval <julia.koval@intel.com>
8452 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8453 * config/i386/i386.md (rdpid_rex64) New.
8454 (rdpid): Make 32bit only.
8456 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8459 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8460 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8462 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8465 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8467 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8469 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8471 PR rtl-optimization/84071
8472 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8473 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8475 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8477 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8478 (arc_attribute_table): Add 'aux' attribute.
8479 (arc_in_small_data_p): Consider aux like variables.
8480 (arc_is_aux_reg_p): New function.
8481 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8482 (arc_get_aux_arg): New function.
8483 (prepare_move_operands): Handle aux-register access.
8484 (arc_handle_aux_attribute): New function.
8485 * doc/extend.texi (ARC Variable attributes): Add subsection.
8487 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8489 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8490 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8491 (arc_attribute_table): Add 'uncached' attribute.
8492 (arc_print_operand): Print '.di' flag for uncached memory
8494 (arc_in_small_data_p): Do not consider for small data the uncached
8496 (arc_is_uncached_mem_p): New function.
8497 * config/arc/predicates.md (compact_store_memory_operand): Check
8498 for uncached memory accesses.
8499 (nonvol_nonimm_operand): Likewise.
8500 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8502 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8505 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8506 falign-loops=): Add Optimization flag.
8508 2018-01-30 Jeff Law <law@redhat.com>
8511 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8512 INT_REGISTERS_SAVED. Check it prior to calling
8513 get_scratch_register_on_entry.
8514 (ix86_adjust_stack_and_probe): Similarly.
8515 (ix86_emit_probe_stack_range): Similarly.
8516 (ix86_expand_prologue): Corresponding changes.
8518 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8521 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8522 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8524 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8527 * lra-constraints.c (curr_insn_transform): Process AND in the
8530 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8532 PR rtl-optimization/83986
8533 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8534 dependence against last_pending_memory_flush in addition to
8537 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8539 PR tree-optimization/81611
8540 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8543 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8546 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8549 2018-01-30 Richard Biener <rguenther@suse.de>
8550 Jakub Jelinek <jakub@redhat.com>
8552 PR tree-optimization/84111
8553 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8554 inner loops added during recursion, as they don't have up-to-date
8557 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8560 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8561 (can_inline_edge_by_limits_p): ... here.
8562 (can_early_inline_edge_p, check_callers,
8563 update_caller_keys, update_callee_keys, recursive_inlining,
8564 add_new_edges_to_heap, speculation_useful_p,
8565 inline_small_functions,
8566 inline_small_functions, flatten_function,
8567 inline_to_all_callers_1): Update.
8569 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8571 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8574 2018-01-30 Richard Biener <rguenther@suse.de>
8576 PR tree-optimization/83008
8577 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8578 invariant and constant vector uses in stmts when they need
8581 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8584 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8585 * configure: Regenerate.
8587 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8589 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8591 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8592 Use gen_rtx_REG rather than gen_lowpart.
8594 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8596 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8597 rather than 0 when creating partial subregs.
8599 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8601 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8604 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8607 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8608 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8609 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8610 flags. This restores the settings used before the 2017-07-24.
8611 Turning off pre increment/decrement/modify allows IVOPTS to
8612 optimize DF/SF loops where the index is an int.
8614 2018-01-29 Richard Biener <rguenther@suse.de>
8615 Kelvin Nilsen <kelvin@gcc.gnu.org>
8618 * tree-vect-stmts.c (vectorizable_call): Don't call
8619 targetm.vectorize_builtin_md_vectorized_function if callee is
8622 2018-01-22 Carl Love <cel@us.ibm.com>
8624 * doc/extend.tex: Fix typo in second arg in
8625 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8627 2018-01-29 Richard Biener <rguenther@suse.de>
8629 PR tree-optimization/84086
8630 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8631 (flush_ssaname_freelist): When SSA names were released reset
8632 the SCEV hash table.
8634 2018-01-29 Richard Biener <rguenther@suse.de>
8636 PR tree-optimization/84057
8637 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8638 removed paths when removing edges.
8640 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8642 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8643 -mfunction-return=@var{choice}.
8645 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8648 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8650 (layout::print_source_line): Likewise.
8651 (test_get_line_width_without_trailing_whitespace): Add test cases.
8653 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8656 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8659 2018-01-26 Jim Wilson <jimw@sifive.com>
8661 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8663 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8666 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8668 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8669 and CMP + SUB-immediate -> SUBS.
8671 2018-01-26 Martin Sebor <msebor@redhat.com>
8673 PR tree-optimization/83896
8674 * tree-ssa-strlen.c (get_string_len): Rename...
8675 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8676 Avoid assuming length is constant.
8677 (handle_char_store): Use HOST_WIDE_INT for string length.
8679 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8682 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8683 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8685 2018-01-26 Richard Biener <rguenther@suse.de>
8687 PR rtl-optimization/84003
8688 * dse.c (record_store): Only record redundant stores when
8689 the earlier store aliases at least all accesses the later one does.
8691 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8693 PR rtl-optimization/83985
8694 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8695 REG_CFA_RESTORE insns.
8696 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8697 REG_CFA_RESTORE insns here.
8700 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8701 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8703 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8705 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8706 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8707 (arc_init): Likewise.
8708 (arc_override_options): Likewise.
8709 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8711 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8713 * config/arc/arc.h (TARGET_DBNZ): Define.
8714 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8715 properly set the tune attribute.
8716 (dbnz): Use TARGET_DBNZ guard.
8717 * config/arc/arc.opt (mtune): Add core3 option.
8719 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8721 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8722 recognize new pic like addresses.
8723 (arc_delegitimize_address): Clean up.
8725 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8727 * config/arc/arc-arches.def: Option mrf16 valid for all
8729 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8730 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8731 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8732 * config/arc/arc-tables.opt: Regenerate.
8733 * config/arc/arc.c (arc_conditional_register_usage): Handle
8734 reduced register file case.
8735 (arc_file_start): Set must have build attributes.
8736 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8738 * config/arc/arc.opt (mrf16): Add new option.
8739 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8740 * config/arc/genmultilib.awk: Handle new mrf16 option.
8741 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8742 * config/arc/t-multilib: Regenerate.
8743 * doc/invoke.texi (ARC Options): Document mrf16 option.
8745 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8747 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8748 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8749 (arc_attribute_table): Add 'secure_call' attribute.
8750 (arc_print_operand): Print secure call operand.
8751 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8753 (arc_is_secure_call_p): New function. * config/arc/arc.md
8754 (call_i): Add support for sjli instruction.
8755 (call_value_i): Likewise.
8756 * config/arc/constraints.md (Csc): New constraint.
8758 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8759 John Eric Martin <John.Martin@emmicro-us.com>
8761 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8762 * config/arc/arc.c (_arc_jli_section): New struct.
8763 (arc_jli_section): New type.
8764 (rc_jli_sections): New static variable.
8765 (arc_handle_jli_attribute): New function.
8766 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8767 (arc_file_end): New function.
8768 (TARGET_ASM_FILE_END): Define.
8769 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8770 (arc_add_jli_section): New function.
8771 (jli_call_scan): Likewise.
8772 (arc_reorg): Call jli_call_scan.
8773 (arc_output_addsi): Remove 'S' from printing asm operand.
8774 (arc_is_jli_call_p): New function.
8775 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8777 (movhi_insn): Likewise.
8778 (movsi_insn): Likewise.
8779 (movsi_set_cc_insn): Likewise.
8780 (loadqi_update): Likewise.
8781 (load_zeroextendqisi_update): Likewise.
8782 (load_signextendqisi_update): Likewise.
8783 (loadhi_update): Likewise.
8784 (load_zeroextendhisi_update): Likewise.
8785 (load_signextendhisi_update): Likewise.
8786 (loadsi_update): Likewise.
8787 (loadsf_update): Likewise.
8788 (movsicc_insn): Likewise.
8789 (bset_insn): Likewise.
8790 (bxor_insn): Likewise.
8791 (bclr_insn): Likewise.
8792 (bmsk_insn): Likewise.
8793 (bicsi3_insn): Likewise.
8794 (cmpsi_cc_c_insn): Likewise.
8795 (movsi_ne): Likewise.
8796 (movsi_cond_exec): Likewise.
8797 (clrsbsi2): Likewise.
8808 (call_i): Remove 'S' asm letter, add jli instruction.
8809 (call_value_i): Likewise.
8810 * config/arc/arc.op (mjli-always): New option.
8811 * config/arc/constraints.md (Cji): New constraint.
8812 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8814 (subsf3_fpx): Likewise.
8815 (mulsf3_fpx): Likewise.
8816 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8818 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8820 * doc/invoke.texi (ARC): Document mjli-always option.
8822 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8824 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8825 avoid addition with 0 and use incw and decw where possible.
8827 2018-01-26 Richard Biener <rguenther@suse.de>
8829 PR tree-optimization/81082
8830 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8831 association if it requires casting to unsigned.
8832 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8833 from fold_plusminus_mult_expr to catch important cases late when
8834 range info is available.
8836 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8838 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8839 * configure.ac (hidden_linkonce): New test.
8840 * configure: Regenerate.
8841 * config.in: Regenerate.
8843 2018-01-26 Julia Koval <julia.koval@intel.com>
8845 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8846 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8847 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8848 _mm_mask_bitshuffle_epi64_mask): Fix type.
8849 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8850 USI_FTYPE_V4DI_V4DI_USI): Remove.
8851 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8852 __builtin_ia32_vpshufbitqmb256_mask,
8853 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8854 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8855 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8857 2018-01-26 Alan Modra <amodra@gmail.com>
8860 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8861 UNSPEC_VBPERMQ. Sort other unspecs.
8863 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8865 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8867 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8870 * predict.c (drop_profile): Do not push/pop cfun; update also
8872 (handle_missing_profiles): Fix logic looking for zero profiles.
8874 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8877 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8878 on functions with #pragma omp declare simd or functions with simd
8880 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8881 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8882 Remove trailing \n from warning_at calls.
8884 2018-01-25 Tom de Vries <tom@codesourcery.com>
8887 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8888 for neutered workers.
8890 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8893 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8894 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8896 2018-01-24 Jeff Law <law@redhat.com>
8899 * i386.c (get_probe_interval): Move to earlier point.
8900 (ix86_compute_frame_layout): If -fstack-clash-protection and
8901 the frame is larger than the probe interval, then use pushes
8902 to save registers rather than reg->mem moves.
8903 (ix86_expand_prologue): Remove conditional for int_registers_saved
8906 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8909 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8910 min/max for never referenced object.
8912 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8915 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8917 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8918 attributes from DECL_ATTRIBUTES (decl) without affecting
8919 DECL_ATTRIBUTES (current_function_decl).
8920 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8921 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8923 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8925 PR tree-optimization/83979
8926 * fold-const.c (fold_comparison): Use constant_boolean_node
8927 instead of boolean_{true,false}_node.
8929 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8931 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8934 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8936 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8937 Simplify the clause that sets the length attribute.
8938 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8939 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8940 clause that sets the length attribute.
8941 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8943 2018-01-24 Tom de Vries <tom@codesourcery.com>
8946 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8947 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8948 Add strict parameter.
8949 (prevent_branch_around_nothing): Insert dummy insn between branch to
8950 label and label with no ptx insn inbetween.
8951 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8953 2018-01-24 Tom de Vries <tom@codesourcery.com>
8956 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8957 for neutered threads in warp.
8958 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8960 2018-01-24 Richard Biener <rguenther@suse.de>
8962 PR tree-optimization/83176
8963 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8966 2018-01-24 Richard Biener <rguenther@suse.de>
8968 PR tree-optimization/82819
8969 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8970 code generating pluses that are no-ops in the target precision.
8972 2018-01-24 Richard Biener <rguenther@suse.de>
8975 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8977 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8979 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8980 to merge probabilities.
8981 * predict.c (probably_never_executed): Also mark as cold functions
8982 with global 0 profile and guessed local profile.
8983 * profile-count.c (profile_probability::combine_with_count): New
8985 * profile-count.h (profile_probability::operator*,
8986 profile_probability::operator*=, profile_probability::operator/,
8987 profile_probability::operator/=): Reduce precision to adjusted
8988 and set value to guessed on contradictory divisions.
8989 (profile_probability::combine_with_freq): Remove.
8990 (profile_probability::combine_wiht_count): Declare.
8991 (profile_count::force_nonzero):: Set to adjusted.
8992 (profile_count::probability_in):: Set quality to adjusted.
8993 * tree-ssa-tail-merge.c (replace_block_by): Use
8996 2018-01-23 Andrew Waterman <andrew@sifive.com>
8997 Jim Wilson <jimw@sifive.com>
8999 * config/riscv/riscv.c (riscv_stack_boundary): New.
9000 (riscv_option_override): Set riscv_stack_boundary. Handle
9001 riscv_preferred_stack_boundary_arg.
9002 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9003 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9004 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9005 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9006 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9007 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9009 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9012 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9013 of struct ix86_frame.
9014 (ix86_expand_epilogue): Likewise. Add a local variable for
9015 the reg_save_offset field in struct ix86_frame.
9017 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9019 PR tree-optimization/82604
9020 * tree-loop-distribution.c (enum partition_kind): New enum item
9021 PKIND_PARTIAL_MEMSET.
9022 (partition_builtin_p): Support above new enum item.
9023 (generate_code_for_partition): Ditto.
9024 (compute_access_range): Differentiate cases that equality can be
9025 proven at all loops, the innermost loops or no loops.
9026 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9027 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9028 (finalize_partitions, distribute_loop): Don't fuse partition of
9029 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9030 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9033 2018-01-23 Martin Liska <mliska@suse.cz>
9035 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9036 order to ignore the predictor.
9037 (PRED_POLYMORPHIC_CALL): Likewise.
9038 (PRED_RECURSIVE_CALL): Likewise.
9040 2018-01-23 Martin Liska <mliska@suse.cz>
9042 * tree-profile.c (tree_profiling): Print function header to
9043 aware reader which function we are working on.
9044 * value-prof.c (gimple_find_values_to_profile): Do not print
9045 not interesting value histograms.
9047 2018-01-23 Martin Liska <mliska@suse.cz>
9049 * profile-count.h (enum profile_quality): Add
9050 profile_uninitialized as the first value. Do not number values
9051 as they are zero based.
9052 (profile_count::verify): Update sanity check.
9053 (profile_probability::verify): Likewise.
9055 2018-01-23 Nathan Sidwell <nathan@acm.org>
9057 * doc/invoke.texi (ffor-scope): Deprecate.
9059 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9061 PR tree-optimization/83510
9062 * domwalk.c (set_all_edges_as_executable): New function.
9063 (dom_walker::dom_walker): Convert bool param
9064 "skip_unreachable_blocks" to enum reachability. Move setup of
9065 edge flags to set_all_edges_as_executable and only do it when
9066 reachability is REACHABLE_BLOCKS.
9067 * domwalk.h (enum dom_walker::reachability): New enum.
9068 (dom_walker::dom_walker): Convert bool param
9069 "skip_unreachable_blocks" to enum reachability.
9070 (set_all_edges_as_executable): New decl.
9071 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9072 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9074 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9075 but converting true to REACHABLE_BLOCKS.
9076 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9078 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9079 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9080 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9082 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9083 if check_all_array_refs will be called.
9085 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9087 * tree.c (selftest::test_location_wrappers): Add more test
9090 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9092 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9093 (selftest::test_bit_in_range): Likewise.
9095 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9098 * doc/sourcebuild.texi (vect_float): Say that the selector
9099 only describes the situation when -funsafe-math-optimizations is on.
9100 (vect_float_strict): Document.
9102 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9104 PR tree-optimization/83965
9105 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9106 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9107 instead of checking only for a reduction.
9108 (vect_recog_widen_sum_pattern): Likewise.
9110 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9112 * predict.c (probably_never_executed): Only use precise profile info.
9113 (compute_function_frequency): Skip after inlining hack since we now
9114 have quality checking.
9116 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9118 * profile-count.h (profile_probability::very_unlikely,
9119 profile_probability::unlikely, profile_probability::even): Set
9120 precision to guessed.
9122 2018-01-23 Richard Biener <rguenther@suse.de>
9124 PR tree-optimization/83963
9125 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9126 Properly terminate dominator walk when crossing the exit edge not
9127 when visiting its source block.
9129 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9132 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9133 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9135 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9137 PR tree-optimization/83957
9138 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9139 semicolon after for body surrounded by braces.
9141 PR tree-optimization/83081
9142 * profile-count.h (profile_probability::split): New method.
9143 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9144 Use profile_probability::split.
9145 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9146 when splitting a single conditional jump into 2.
9148 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9150 PR tree-optimization/69452
9151 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9154 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9156 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9157 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9158 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9160 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9162 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9163 * config/rl78/rl78.md (movdi): New define_expand.
9164 * config/rl78/rl78.c (rl78_split_movdi): New function.
9166 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9169 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9171 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9172 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9173 128-bit to produce an UNSPEC move to get the double word with the
9174 signbit and then a shift directly to do signbit.
9175 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9176 implementation with a new version that just does either a direct
9177 move or a regular move. Move memory interface to separate insns.
9178 Move insns so they are next to the expander.
9179 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9180 with signbit move. Split big and little endian case.
9181 (signbit<mode>2_dm_mem_le): Likewise.
9182 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9183 (signbit<mode>2_dm2): Likewise.
9185 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9187 * config/rl78/rl78.md (anddi3): New define_expand.
9189 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9191 * config/rl78/rl78.md (umindi3): New define_expand.
9193 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9195 * config/rl78/rl78.md (smindi3): New define_expand.
9197 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9199 * config/rl78/rl78.md (smaxdi3): New define_expand.
9201 2018-01-22 Carl Love <cel@us.ibm.com>
9203 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9204 LVX_V1TI): Add macro expansion.
9205 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9206 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9207 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9208 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9209 Change check to determine if the instruction is a byte reversing
9210 entry. Fix typo in comment.
9211 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9212 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9213 Add def_builtin calls for new builtins.
9214 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9215 Add define_insn expansion.
9217 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9219 * config/rl78/rl78.md (umaxdi3): New define_expand.
9221 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9223 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9224 for non-QImode registers.
9226 2018-01-22 Richard Biener <rguenther@suse.de>
9228 PR tree-optimization/83963
9229 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9230 including the loop exit block.
9231 (scop_detection::merge_sese): Likewise.
9232 (scop_detection::add_scop): Do it here instead.
9234 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9236 * doc/sourcebuild.texi (arm_softfloat): Document.
9238 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9241 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9242 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9243 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9245 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9246 David Edelsohn <dje.gcc@gmail.com>
9249 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9250 Change "crset eq" to "crset 2".
9251 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9252 (*call_indirect_aix<mode>_nospec): Likewise.
9253 (*call_value_indirect_aix<mode>_nospec): Likewise.
9254 (*call_indirect_elfv2<mode>_nospec): Likewise.
9255 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9256 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9257 change assembly output from . to $.
9258 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9259 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9260 (*tablejump<mode>_internal1_nospec): Likewise.
9262 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9265 * config/sh/sh_optimize_sett_clrt.cc:
9266 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9268 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9270 PR tree-optimization/83940
9271 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9272 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9273 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9274 use it to pass back the definition type.
9275 (vect_check_store_rhs): Likewise rhs_dt_out.
9276 (vect_build_gather_load_calls): Add a mask_dt argument and use
9277 it instead of a call to vect_is_simple_use.
9278 (vectorizable_store): Update calls to vect_check_load_store_mask
9279 and vect_check_store_rhs. Use the dt returned by the latter instead
9280 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9281 instead of calls to vect_is_simple_use. Pass the scalar rather
9282 than the vector operand to vect_is_simple_use when handling
9283 second and subsequent copies of an rhs value.
9284 (vectorizable_load): Update calls to vect_check_load_store_mask
9285 and vect_build_gather_load_calls. Use the cached mask_dt and
9286 gs_info.offset_dt instead of calls to vect_is_simple_use.
9288 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9291 * tree-emutls.c: Include gimplify.h.
9292 (lower_emutls_2): New function.
9293 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9294 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9295 it before further processing.
9298 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9299 UINTVAL (trueop1) instead of INTVAL (op1).
9301 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9305 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9306 INCOMING_FRAME_SP_OFFSET if not defined.
9307 (scan_trace): Add ENTRY argument. If true and
9308 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9309 emit a note to adjust the CFA offset.
9310 (create_cfi_notes): Adjust scan_trace callers.
9311 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9312 INCOMING_FRAME_SP_OFFSET in the CIE.
9313 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9314 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9316 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9317 * doc/tm.texi: Regenerated.
9319 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9321 PR rtl-optimization/83147
9322 * lra-constraints.c (remove_inheritance_pseudos): Use
9323 lra_substitute_pseudo_within_insn.
9325 2018-01-19 Tom de Vries <tom@codesourcery.com>
9326 Cesar Philippidis <cesar@codesourcery.com>
9329 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9331 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9334 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9335 spaces for function labels.
9337 2018-01-19 Martin Liska <mliska@suse.cz>
9339 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9340 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9341 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9342 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9343 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9344 (PRED_CONST_RETURN): Change from 69 to 65.
9345 (PRED_NULL_RETURN): Change from 91 to 71.
9346 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9347 (PRED_LOOP_GUARD): Change from 66 to 73.
9349 2018-01-19 Martin Liska <mliska@suse.cz>
9351 * predict.c (predict_insn_def): Add new assert.
9352 (struct branch_predictor): Change type to signed integer.
9353 (test_prediction_value_range): Amend test to cover
9355 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9356 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9357 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9358 (PRED_LOOP_IV_COMPARE): Likewise.
9359 * predict.h (PROB_UNINITIALIZED): Define new constant.
9361 2018-01-19 Martin Liska <mliska@suse.cz>
9363 * predict.c (dump_prediction): Add new format for
9364 analyze_brprob.py script which is enabled with -details
9366 * profile-count.h (precise_p): New function.
9368 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9370 PR tree-optimization/83922
9371 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9372 there are no statements that need masking.
9373 (vect_active_double_reduction_p): New function.
9374 (vect_analyze_loop_operations): Use it when handling phis that
9375 are not in the loop header.
9377 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9379 PR tree-optimization/83914
9380 * tree-vect-loop.c (vectorizable_induction): Don't convert
9381 init_expr or apply the peeling adjustment for inductions
9382 that are nested within the vectorized loop.
9384 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9386 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9389 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9393 * function.h (gimplify_parameters): Add gimple_seq * argument.
9394 * function.c: Include gimple.h and options.h.
9395 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9396 for the added local temporaries if needed.
9397 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9398 if there are any parameter cleanups, wrap whole body into a
9399 try/finally with the cleanups.
9401 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9404 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9405 Use GET_MODE_CLASS for scalar floating point.
9407 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9411 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9412 Fix call of call_cgraph_insertion_hooks.
9414 2018-01-18 Martin Sebor <msebor@redhat.com>
9416 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9418 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9421 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9424 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9427 * common.opt: (-ffile-prefix-map): New option.
9428 * opts.c (common_handle_option): Defer it.
9429 * opts-global.c (handle_common_deferred_options): Handle it.
9430 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9431 * file-prefix-map.h: New file.
9432 (remap_debug_filename, add_debug_prefix_map): ...here.
9433 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9434 * final.c (debug_prefix_map, add_debug_prefix_map
9435 remap_debug_filename): Move to...
9436 * file-prefix-map.c: New file.
9437 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9438 generalize, get rid of alloca(), use strrchr() instead of strchr().
9439 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9440 Implement in terms of add_prefix_map().
9441 (remap_macro_filename, remap_debug_filename): Implement in term of
9443 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9444 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9445 * dbxout.c: Include file-prefix-map.h.
9446 * varasm.c: Likewise.
9447 * vmsdbgout.c: Likewise.
9448 * xcoffout.c: Likewise.
9449 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9450 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9451 * doc/invoke.texi (-ffile-prefix-map): Document.
9452 (-fdebug-prefix-map): Update description.
9454 2018-01-18 Martin Liska <mliska@suse.cz>
9456 * config/i386/i386.c (indirect_thunk_name): Document that also
9458 (output_indirect_thunk): Document why both instructions
9459 (pause and lfence) are generated.
9461 2018-01-18 Richard Biener <rguenther@suse.de>
9463 PR tree-optimization/83887
9464 * graphite-scop-detection.c
9465 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9466 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9467 (scop_detection::merge_sese): Re-implement with a flood-fill
9468 algorithm that properly finds a SESE region if it exists.
9470 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9473 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9474 pointer_diff optimizations use view_convert instead of convert.
9476 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9478 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9479 Generate different code for -mno-speculate-indirect-jumps.
9480 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9481 (*call_indirect_aix<mode>): Disable for
9482 -mno-speculate-indirect-jumps.
9483 (*call_indirect_aix<mode>_nospec): New define_insn.
9484 (*call_value_indirect_aix<mode>): Disable for
9485 -mno-speculate-indirect-jumps.
9486 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9487 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9488 -mno-speculate-indirect-jumps.
9489 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9491 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9493 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9494 long double type, set the flags for noting the default long double
9495 type, even if we don't pass or return a long double type.
9497 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9500 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9503 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9505 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9506 support for merge[hl].
9507 (fold_mergehl_helper): New helper function.
9508 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9509 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9510 (altivec_vmrglw_direct): Add xxmrglw insn.
9512 2018-01-17 Andrew Waterman <andrew@sifive.com>
9514 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9515 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9517 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9520 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9521 call the lto_location_cache before reading the
9522 DECL_SOURCE_LOCATION of the types.
9524 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9525 Richard Sandiford <richard.sandiford@linaro.org>
9527 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9528 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9529 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9530 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9531 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9533 * config/aarch64/constraints.md (aarch64_movti_operand):
9535 * config/aarch64/predicates.md (Uti): Add new constraint.
9537 2018-01-17 Carl Love <cel@us.ibm.com>
9539 * config/rs6000/vsx.md (define_expand xl_len_r,
9540 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9541 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9543 (define_expand, define_insn): Move the shift left from the
9544 define_insn to the define_expand for lxvl and stxvl instructions.
9545 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9546 and XL_LEN_R definitions to PURE.
9548 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9550 * config/i386/i386.c (indirect_thunk_name): Declare regno
9551 as unsigned int. Compare regno with INVALID_REGNUM.
9552 (output_indirect_thunk): Ditto.
9553 (output_indirect_thunk_function): Ditto.
9554 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9555 in the call to output_indirect_thunk_function.
9557 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9560 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9561 rather than the size of inner_type to determine the stack slot size
9562 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9564 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9567 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9570 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9572 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9573 endian Linux systems to optionally enable multilibs for selecting
9574 the long double type if the user configured an explicit type.
9575 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9576 have no long double multilibs if not defined.
9577 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9578 warn if the user used -mabi={ieee,ibm}longdouble and we built
9579 multilibs for long double.
9580 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9581 appropriate multilib option.
9582 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9584 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9585 for building long double multilibs.
9586 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9588 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9590 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9593 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9595 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9598 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9601 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9604 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9606 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9607 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9609 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9611 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9612 different rtl trees depending on TARGET_64BIT.
9613 (rs6000_gen_lvx): Likewise.
9615 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9617 * config/visium/visium.md (nop): Tweak comment.
9618 (hazard_nop): Likewise.
9620 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9622 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9623 -mspeculate-indirect-jumps.
9624 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9625 for -mno-speculate-indirect-jumps.
9626 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9627 (*call_value_indirect_elfv2<mode>): Disable for
9628 -mno-speculate-indirect-jumps.
9629 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9630 (indirect_jump): Emit different RTL for
9631 -mno-speculate-indirect-jumps.
9632 (*indirect_jump<mode>): Disable for
9633 -mno-speculate-indirect-jumps.
9634 (*indirect_jump<mode>_nospec): New define_insn.
9635 (tablejump): Emit different RTL for
9636 -mno-speculate-indirect-jumps.
9637 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9638 (tablejumpsi_nospec): New define_expand.
9639 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9640 (tablejumpdi_nospec): New define_expand.
9641 (*tablejump<mode>_internal1): Disable for
9642 -mno-speculate-indirect-jumps.
9643 (*tablejump<mode>_internal1_nospec): New define_insn.
9644 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9647 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9649 * caller-save.c (insert_save): Drop unnecessary parameter. All
9652 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9653 Richard Biener <rguenth@suse.de>
9656 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9657 return early, inline manually is_gimple_sizepos. Make sure if we
9658 call gimplify_expr we don't end up with a gimple constant.
9659 * tree.c (variably_modified_type_p): Don't return true for
9660 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9661 * gimplify.h (is_gimple_sizepos): Remove.
9663 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9665 PR tree-optimization/83857
9666 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9667 vectorizable_live_operation for pure SLP statements.
9668 (vectorizable_live_operation): Handle PHIs.
9670 2018-01-16 Richard Biener <rguenther@suse.de>
9672 PR tree-optimization/83867
9673 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9674 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9676 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9679 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9680 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9681 If off is not INTEGER_CST, issue a may not be aligned warning
9682 rather than isn't aligned. Use isn%'t rather than isn't.
9683 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9685 <case MULT_EXPR>: Improve the case when bottom and one of the
9686 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9687 operand, in that case check if the other operand is multiple of
9688 bottom divided by the INTEGER_CST operand.
9690 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9693 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9694 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9695 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9696 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9697 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9698 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9699 (pa_function_arg_size): New function.
9701 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9703 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9704 in a separate statement.
9706 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9708 PR tree-optimization/83847
9709 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9710 group gathers and scatters.
9712 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9714 PR rtl-optimization/86620
9715 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9717 PR rtl-optimization/83213
9718 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9719 to last if both are JUMP_INSNs.
9721 PR tree-optimization/83843
9722 * gimple-ssa-store-merging.c
9723 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9724 store_immediate_info for bswap/nop orig_stores.
9726 2018-01-15 Andrew Waterman <andrew@sifive.com>
9728 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9730 <UDIV>: Increase cost if !TARGET_DIV.
9732 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9734 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9735 (define_attr "cr_logical_3op"): New.
9736 (cceq_ior_compare): Adjust.
9737 (cceq_ior_compare_complement): Adjust.
9738 (*cceq_rev_compare): Adjust.
9739 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9740 (is_cracked_insn): Adjust.
9741 (insn_must_be_first_in_group): Adjust.
9742 * config/rs6000/40x.md: Adjust.
9743 * config/rs6000/440.md: Adjust.
9744 * config/rs6000/476.md: Adjust.
9745 * config/rs6000/601.md: Adjust.
9746 * config/rs6000/603.md: Adjust.
9747 * config/rs6000/6xx.md: Adjust.
9748 * config/rs6000/7450.md: Adjust.
9749 * config/rs6000/7xx.md: Adjust.
9750 * config/rs6000/8540.md: Adjust.
9751 * config/rs6000/cell.md: Adjust.
9752 * config/rs6000/e300c2c3.md: Adjust.
9753 * config/rs6000/e500mc.md: Adjust.
9754 * config/rs6000/e500mc64.md: Adjust.
9755 * config/rs6000/e5500.md: Adjust.
9756 * config/rs6000/e6500.md: Adjust.
9757 * config/rs6000/mpc.md: Adjust.
9758 * config/rs6000/power4.md: Adjust.
9759 * config/rs6000/power5.md: Adjust.
9760 * config/rs6000/power6.md: Adjust.
9761 * config/rs6000/power7.md: Adjust.
9762 * config/rs6000/power8.md: Adjust.
9763 * config/rs6000/power9.md: Adjust.
9764 * config/rs6000/rs64.md: Adjust.
9765 * config/rs6000/titan.md: Adjust.
9767 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9769 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9770 ix86_indirect_branch_register logic.
9772 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9774 * config/i386/constraints.md (Bs): Update
9775 ix86_indirect_branch_register check. Don't check
9776 ix86_indirect_branch_register with GOT_memory_operand.
9778 * config/i386/predicates.md (GOT_memory_operand): Don't check
9779 ix86_indirect_branch_register here.
9780 (GOT32_symbol_operand): Likewise.
9782 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9784 * config/i386/predicates.md (constant_call_address_operand):
9785 Rewrite ix86_indirect_branch_register logic.
9786 (sibcall_insn_operand): Likewise.
9788 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9790 * config/i386/constraints.md (Bs): Replace
9791 ix86_indirect_branch_thunk_register with
9792 ix86_indirect_branch_register.
9794 * config/i386/i386.md (indirect_jump): Likewise.
9795 (tablejump): Likewise.
9796 (*sibcall_memory): Likewise.
9797 (*sibcall_value_memory): Likewise.
9798 Peepholes of indirect call and jump via memory: Likewise.
9799 * config/i386/i386.opt: Likewise.
9800 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9801 (GOT_memory_operand): Likewise.
9802 (call_insn_operand): Likewise.
9803 (sibcall_insn_operand): Likewise.
9804 (GOT32_symbol_operand): Likewise.
9806 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9809 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9810 type rather than type addr's type points to.
9811 (expand_omp_atomic_mutex): Likewise.
9812 (expand_omp_atomic): Likewise.
9814 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9817 * config/i386/i386.c (output_indirect_thunk_function): Use
9818 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9819 for __x86_return_thunk.
9821 2018-01-15 Richard Biener <rguenther@suse.de>
9824 * expmed.c (extract_bit_field_1): Fix typo.
9826 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9829 * config/arm/iterators.md (VF): New mode iterator.
9830 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9831 Remove integer-related logic from pattern.
9832 (neon_vabd<mode>_3): Likewise.
9834 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9837 * common.opt (fstrict-overflow): No longer an alias.
9838 (fwrapv-pointer): New option.
9839 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9840 also for pointer types based on flag_wrapv_pointer.
9841 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9842 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9843 opts->x_flag_wrapv got set.
9844 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9845 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9846 POINTER_TYPE_OVERFLOW_UNDEFINED.
9847 * match.pd: Likewise in address comparison pattern.
9848 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9850 2018-01-15 Richard Biener <rguenther@suse.de>
9853 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9854 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9855 Reset type names to their identifier if their TYPE_DECL doesn't
9856 have linkage (and thus is used for ODR and devirt).
9857 (save_debug_info_for_decl): Remove.
9858 (save_debug_info_for_type): Likewise.
9859 (add_tree_to_fld_list): Adjust.
9860 * tree-pretty-print.c (dump_generic_node): Make dumping of
9861 type names more robust.
9863 2018-01-15 Richard Biener <rguenther@suse.de>
9865 * BASE-VER: Bump to 8.0.1.
9867 2018-01-14 Martin Sebor <msebor@redhat.com>
9870 * builtins.c (check_access): Avoid warning when the no-warning bit
9873 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9875 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9876 * ira-color (allocno_hard_regs_compare): Likewise.
9878 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9881 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9882 Use .pushsection/.popsection.
9884 2018-01-14 Martin Sebor <msebor@redhat.com>
9887 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9889 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9891 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9892 entry from extra_headers.
9893 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9894 extra_headers, make the list bitwise identical to the i?86-*-* one.
9896 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9898 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9899 -mcmodel=large with -mindirect-branch=thunk,
9900 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9901 -mfunction-return=thunk-extern.
9902 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9903 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9904 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9906 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9908 * config/i386/i386.c (print_reg): Print the name of the full
9909 integer register without '%'.
9910 (ix86_print_operand): Handle 'V'.
9911 * doc/extend.texi: Document 'V' modifier.
9913 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9915 * config/i386/constraints.md (Bs): Disallow memory operand for
9916 -mindirect-branch-register.
9918 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9919 (GOT_memory_operand): Likewise.
9920 (call_insn_operand): Likewise.
9921 (sibcall_insn_operand): Likewise.
9922 (GOT32_symbol_operand): Likewise.
9923 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9924 for -mindirect-branch-register.
9925 (tablejump): Likewise.
9926 (*sibcall_memory): Likewise.
9927 (*sibcall_value_memory): Likewise.
9928 Disallow peepholes of indirect call and jump via memory for
9929 -mindirect-branch-register.
9930 (*call_pop): Replace m with Bw.
9931 (*call_value_pop): Likewise.
9932 (*sibcall_pop_memory): Replace m with Bs.
9933 * config/i386/i386.opt (mindirect-branch-register): New option.
9934 * doc/invoke.texi: Document -mindirect-branch-register option.
9936 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9938 * config/i386/i386-protos.h (ix86_output_function_return): New.
9939 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9940 set function_return_type.
9941 (indirect_thunk_name): Add ret_p to indicate thunk for function
9943 (output_indirect_thunk_function): Pass false to
9944 indirect_thunk_name.
9945 (ix86_output_indirect_branch_via_reg): Likewise.
9946 (ix86_output_indirect_branch_via_push): Likewise.
9947 (output_indirect_thunk_function): Create alias for function
9948 return thunk if regno < 0.
9949 (ix86_output_function_return): New function.
9950 (ix86_handle_fndecl_attribute): Handle function_return.
9951 (ix86_attribute_table): Add function_return.
9952 * config/i386/i386.h (machine_function): Add
9953 function_return_type.
9954 * config/i386/i386.md (simple_return_internal): Use
9955 ix86_output_function_return.
9956 (simple_return_internal_long): Likewise.
9957 * config/i386/i386.opt (mfunction-return=): New option.
9958 (indirect_branch): Mention -mfunction-return=.
9959 * doc/extend.texi: Document function_return function attribute.
9960 * doc/invoke.texi: Document -mfunction-return= option.
9962 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9964 * config/i386/i386-opts.h (indirect_branch): New.
9965 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9966 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9967 with local indirect jump when converting indirect call and jump.
9968 (ix86_set_indirect_branch_type): New.
9969 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9970 (indirectlabelno): New.
9971 (indirect_thunk_needed): Likewise.
9972 (indirect_thunk_bnd_needed): Likewise.
9973 (indirect_thunks_used): Likewise.
9974 (indirect_thunks_bnd_used): Likewise.
9975 (INDIRECT_LABEL): Likewise.
9976 (indirect_thunk_name): Likewise.
9977 (output_indirect_thunk): Likewise.
9978 (output_indirect_thunk_function): Likewise.
9979 (ix86_output_indirect_branch_via_reg): Likewise.
9980 (ix86_output_indirect_branch_via_push): Likewise.
9981 (ix86_output_indirect_branch): Likewise.
9982 (ix86_output_indirect_jmp): Likewise.
9983 (ix86_code_end): Call output_indirect_thunk_function if needed.
9984 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9986 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9987 (ix86_attribute_table): Add indirect_branch.
9988 * config/i386/i386.h (machine_function): Add indirect_branch_type
9989 and has_local_indirect_jump.
9990 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9992 (tablejump): Likewise.
9993 (*indirect_jump): Use ix86_output_indirect_jmp.
9994 (*tablejump_1): Likewise.
9995 (simple_return_indirect_internal): Likewise.
9996 * config/i386/i386.opt (mindirect-branch=): New option.
9997 (indirect_branch): New.
10000 (thunk-inline): Likewise.
10001 (thunk-extern): Likewise.
10002 * doc/extend.texi: Document indirect_branch function attribute.
10003 * doc/invoke.texi: Document -mindirect-branch= option.
10005 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10008 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10010 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10012 * ipa-inline.c (want_inline_small_function_p): Return false if
10013 inlining has already failed with CIF_FINAL_ERROR.
10014 (update_caller_keys): Call want_inline_small_function_p before
10016 (update_callee_keys): Likewise.
10018 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10020 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10022 (rs6000_quadword_masked_address_p): Likewise.
10023 (quad_aligned_load_p): Likewise.
10024 (quad_aligned_store_p): Likewise.
10025 (const_load_sequence_p): Add comment to describe the outer-most loop.
10026 (mimic_memory_attributes_and_flags): New function.
10027 (rs6000_gen_stvx): Likewise.
10028 (replace_swapped_aligned_store): Likewise.
10029 (rs6000_gen_lvx): Likewise.
10030 (replace_swapped_aligned_load): Likewise.
10031 (replace_swapped_load_constant): Capitalize argument name in
10032 comment describing this function.
10033 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10034 and stores that access quad-word aligned addresses and replace
10035 with stvx or lvx instructions when appropriate.
10036 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10037 New function prototype.
10038 (rs6000_quadword_masked_address_p): Likewise.
10039 (rs6000_gen_lvx): Likewise.
10040 (rs6000_gen_stvx): Likewise.
10041 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10042 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10043 when memory address is aligned.
10044 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10045 this split to select lvx instruction when memory address is aligned.
10046 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10047 instruction when memory address is aligned.
10048 (*vsx_le_perm_load_v16qi): Likewise.
10049 (four unnamed splitters): Modify to select the stvx instruction
10050 when memory is aligned.
10052 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10054 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10055 which appears in the queue multiple times.
10057 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10058 Alan Hayward <alan.hayward@arm.com>
10059 David Sherwood <david.sherwood@arm.com>
10061 * tree-vectorizer.h (vec_lower_bound): New structure.
10062 (_loop_vec_info): Add check_nonzero and lower_bounds.
10063 (LOOP_VINFO_CHECK_NONZERO): New macro.
10064 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10065 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10066 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10067 fields. Make seg_len the distance travelled, not including the
10069 (dr_direction_indicator): Declare.
10070 (dr_zero_step_indicator): Likewise.
10071 (dr_known_forward_stride_p): Likewise.
10072 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10074 (runtime_alias_check_p): Allow runtime alias checks with
10076 (operator ==): Compare access_size and align.
10077 (prune_runtime_alias_test_list): Rework for new distinction between
10078 the access_size and seg_len.
10079 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10081 (get_segment_min_max): New function.
10082 (create_intersect_range_checks): Use it.
10083 (dr_step_indicator): New function.
10084 (dr_direction_indicator): Likewise.
10085 (dr_zero_step_indicator): Likewise.
10086 (dr_known_forward_stride_p): Likewise.
10087 * tree-loop-distribution.c (data_ref_segment_size): Return
10088 DR_STEP * (niters - 1).
10089 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10091 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10092 (vect_preserves_scalar_order_p): New function, split out from...
10093 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10094 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10095 (vect_vfa_access_size): New function.
10096 (vect_vfa_align): Likewise.
10097 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10098 (dump_lower_bound): New function.
10099 (vect_check_lower_bound): Likewise.
10100 (vect_small_gap_p): Likewise.
10101 (vectorizable_with_step_bound_p): Likewise.
10102 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10103 depencies if the vectorization factor is 1. Convert the checks
10104 for nonzero steps into checks on the bounds of DR_STEP. Try using
10105 a bunds check for variable steps if the minimum required step is
10106 relatively small. Update calls to the dr_with_seg_len
10107 constructor and to vect_compile_time_alias.
10108 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10110 (vect_loop_versioning): Call it.
10111 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10113 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10115 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10116 Alan Hayward <alan.hayward@arm.com>
10117 David Sherwood <david.sherwood@arm.com>
10119 * doc/sourcebuild.texi (vect_scatter_store): Document.
10120 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10122 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10124 * genopinit.c (main): Add supports_vec_scatter_store and
10125 supports_vec_scatter_store_cached to target_optabs.
10126 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10127 IFN_MASK_SCATTER_STORE.
10128 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10130 * internal-fn.h (internal_store_fn_p): Declare.
10131 (internal_fn_stored_value_index): Likewise.
10132 * internal-fn.c (scatter_store_direct): New macro.
10133 (expand_scatter_store_optab_fn): New function.
10134 (direct_scatter_store_optab_supported_p): New macro.
10135 (internal_store_fn_p): New function.
10136 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10137 IFN_MASK_SCATTER_STORE.
10138 (internal_fn_mask_index): Likewise.
10139 (internal_fn_stored_value_index): New function.
10140 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10141 for scatter stores.
10142 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10143 * optabs-query.c (supports_vec_scatter_store_p): New function.
10144 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10145 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10146 true for scatter stores.
10147 (vect_gather_scatter_fn_p): Handle scatter stores too.
10148 (vect_check_gather_scatter): Consider using scatter stores if
10149 supports_vec_scatter_store_p.
10150 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10151 scatter stores too.
10152 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10153 internal_fn_stored_value_index.
10154 (check_load_store_masking): Handle scatter stores too.
10155 (vect_get_store_rhs): Make public.
10156 (vectorizable_call): Use internal_store_fn_p.
10157 (vectorizable_store): Handle scatter store internal functions.
10158 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10159 when deciding whether the end of the group has been reached.
10160 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10161 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10162 (mask_scatter_store<mode>): New insns.
10164 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10165 Alan Hayward <alan.hayward@arm.com>
10166 David Sherwood <david.sherwood@arm.com>
10168 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10169 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10170 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10172 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10173 Use vect_truncate_gather_scatter_offset if we can't treat the
10174 operation as a normal gather load or scatter store.
10175 (get_group_load_store_type): Take the gather_scatter_info
10176 as argument. Try using a gather load or scatter store for
10177 single-element groups.
10178 (get_load_store_type): Update calls to get_group_load_store_type
10179 and vect_use_strided_gather_scatters_p.
10181 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10182 Alan Hayward <alan.hayward@arm.com>
10183 David Sherwood <david.sherwood@arm.com>
10185 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10186 optional tree argument.
10187 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10189 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10190 but continue to use the current value as a fallback.
10191 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10192 to compare the updates.
10193 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10194 (get_load_store_type): Use it when handling a strided access.
10195 (vect_get_strided_load_store_ops): New function.
10196 (vect_get_data_ptr_increment): Likewise.
10197 (vectorizable_load): Handle strided gather loads. Always pass
10198 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10200 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10201 Alan Hayward <alan.hayward@arm.com>
10202 David Sherwood <david.sherwood@arm.com>
10204 * doc/md.texi (gather_load@var{m}): Document.
10205 (mask_gather_load@var{m}): Likewise.
10206 * genopinit.c (main): Add supports_vec_gather_load and
10207 supports_vec_gather_load_cached to target_optabs.
10208 * optabs-tree.c (init_tree_optimization_optabs): Use
10209 ggc_cleared_alloc to allocate target_optabs.
10210 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10211 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10213 * internal-fn.h (internal_load_fn_p): Declare.
10214 (internal_gather_scatter_fn_p): Likewise.
10215 (internal_fn_mask_index): Likewise.
10216 (internal_gather_scatter_fn_supported_p): Likewise.
10217 * internal-fn.c (gather_load_direct): New macro.
10218 (expand_gather_load_optab_fn): New function.
10219 (direct_gather_load_optab_supported_p): New macro.
10220 (direct_internal_fn_optab): New function.
10221 (internal_load_fn_p): Likewise.
10222 (internal_gather_scatter_fn_p): Likewise.
10223 (internal_fn_mask_index): Likewise.
10224 (internal_gather_scatter_fn_supported_p): Likewise.
10225 * optabs-query.c (supports_at_least_one_mode_p): New function.
10226 (supports_vec_gather_load_p): Likewise.
10227 * optabs-query.h (supports_vec_gather_load_p): Declare.
10228 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10229 and memory_type field.
10230 (NUM_PATTERNS): Bump to 15.
10231 * tree-vect-data-refs.c: Include internal-fn.h.
10232 (vect_gather_scatter_fn_p): New function.
10233 (vect_describe_gather_scatter_call): Likewise.
10234 (vect_check_gather_scatter): Try using internal functions for
10235 gather loads. Recognize existing calls to a gather load function.
10236 (vect_analyze_data_refs): Consider using gather loads if
10237 supports_vec_gather_load_p.
10238 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10239 (vect_get_gather_scatter_offset_type): Likewise.
10240 (vect_convert_mask_for_vectype): Likewise.
10241 (vect_add_conversion_to_patterm): Likewise.
10242 (vect_try_gather_scatter_pattern): Likewise.
10243 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10244 (vect_vect_recog_func_ptrs): Add it.
10245 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10246 internal_fn_mask_index and internal_gather_scatter_fn_p.
10247 (check_load_store_masking): Take the gather_scatter_info as an
10248 argument and handle gather loads.
10249 (vect_get_gather_scatter_ops): New function.
10250 (vectorizable_call): Check internal_load_fn_p.
10251 (vectorizable_load): Likewise. Handle gather load internal
10253 (vectorizable_store): Update call to check_load_store_masking.
10254 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10255 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10256 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10257 (aarch64_gather_scale_operand_d): New predicates.
10258 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10259 (mask_gather_load<mode>): New insns.
10261 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10262 Alan Hayward <alan.hayward@arm.com>
10263 David Sherwood <david.sherwood@arm.com>
10265 * optabs.def (fold_left_plus_optab): New optab.
10266 * doc/md.texi (fold_left_plus_@var{m}): Document.
10267 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10268 * internal-fn.c (fold_left_direct): Define.
10269 (expand_fold_left_optab_fn): Likewise.
10270 (direct_fold_left_optab_supported_p): Likewise.
10271 * fold-const-call.c (fold_const_fold_left): New function.
10272 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10273 * tree-parloops.c (valid_reduction_p): New function.
10274 (gather_scalar_reductions): Use it.
10275 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10276 (vect_finish_replace_stmt): Declare.
10277 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10278 (needs_fold_left_reduction_p): New function, split out from...
10279 (vect_is_simple_reduction): ...here. Accept reductions that
10280 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10281 (vect_force_simple_reduction): Also store the reduction type in
10282 the assignment's STMT_VINFO_REDUC_TYPE.
10283 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10284 (merge_with_identity): New function.
10285 (vect_expand_fold_left): Likewise.
10286 (vectorize_fold_left_reduction): Likewise.
10287 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10288 scalar phi in place for it. Check for target support and reject
10289 cases that would reassociate the operation. Defer the transform
10290 phase to vectorize_fold_left_reduction.
10291 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10292 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10293 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10295 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10297 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10298 call to ifc_temp_var.
10300 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10301 Alan Hayward <alan.hayward@arm.com>
10302 David Sherwood <david.sherwood@arm.com>
10304 * target.def (legitimize_address_displacement): Take the original
10305 offset as a poly_int.
10306 * targhooks.h (default_legitimize_address_displacement): Update
10308 * targhooks.c (default_legitimize_address_displacement): Likewise.
10309 * doc/tm.texi: Regenerate.
10310 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10311 as an argument, moving assert of ad->disp == ad->disp_term to...
10312 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10313 Try calling targetm.legitimize_address_displacement before expanding
10314 the address rather than afterwards, and adjust for the new interface.
10315 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10316 Match the new hook interface. Handle SVE addresses.
10317 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10318 new hook interface.
10320 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10322 * Makefile.in (OBJS): Add early-remat.o.
10323 * target.def (select_early_remat_modes): New hook.
10324 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10325 * doc/tm.texi: Regenerate.
10326 * targhooks.h (default_select_early_remat_modes): Declare.
10327 * targhooks.c (default_select_early_remat_modes): New function.
10328 * timevar.def (TV_EARLY_REMAT): New timevar.
10329 * passes.def (pass_early_remat): New pass.
10330 * tree-pass.h (make_pass_early_remat): Declare.
10331 * early-remat.c: New file.
10332 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10334 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10336 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10337 Alan Hayward <alan.hayward@arm.com>
10338 David Sherwood <david.sherwood@arm.com>
10340 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10341 vfm1 with a bound_epilog parameter.
10342 (vect_do_peeling): Update calls accordingly, and move the prologue
10343 call earlier in the function. Treat the base bound_epilog as 0 for
10344 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10345 this base when peeling for gaps.
10346 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10347 with fully-masked loops.
10348 (vect_estimate_min_profitable_iters): Handle the single peeled
10349 iteration in that case.
10351 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10352 Alan Hayward <alan.hayward@arm.com>
10353 David Sherwood <david.sherwood@arm.com>
10355 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10356 single-element interleaving even if the size is not a power of 2.
10357 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10358 accesses for single-element interleaving if the group size is
10361 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10362 Alan Hayward <alan.hayward@arm.com>
10363 David Sherwood <david.sherwood@arm.com>
10365 * doc/md.texi (fold_extract_last_@var{m}): Document.
10366 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10367 * optabs.def (fold_extract_last_optab): New optab.
10368 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10369 * internal-fn.c (fold_extract_direct): New macro.
10370 (expand_fold_extract_optab_fn): Likewise.
10371 (direct_fold_extract_optab_supported_p): Likewise.
10372 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10373 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10374 EXTRACT_LAST_REDUCTION.
10375 (get_initial_def_for_reduction): Do not create an initial vector
10376 for EXTRACT_LAST_REDUCTION reductions.
10377 (vectorizable_reduction): Leave the scalar phi in place for
10378 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10379 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10380 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10381 transform phase to vectorizable_condition.
10382 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10384 (vect_finish_stmt_generation): ...here.
10385 (vect_finish_replace_stmt): New function.
10386 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10387 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10389 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10391 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10392 Alan Hayward <alan.hayward@arm.com>
10393 David Sherwood <david.sherwood@arm.com>
10395 * doc/md.texi (extract_last_@var{m}): Document.
10396 * optabs.def (extract_last_optab): New optab.
10397 * internal-fn.def (EXTRACT_LAST): New internal function.
10398 * internal-fn.c (cond_unary_direct): New macro.
10399 (expand_cond_unary_optab_fn): Likewise.
10400 (direct_cond_unary_optab_supported_p): Likewise.
10401 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10402 loops using EXTRACT_LAST.
10403 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10404 (extract_last_<mode>): ...this optab.
10405 (vec_extract<mode><Vel>): Update accordingly.
10407 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10408 Alan Hayward <alan.hayward@arm.com>
10409 David Sherwood <david.sherwood@arm.com>
10411 * target.def (empty_mask_is_expensive): New hook.
10412 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10413 * doc/tm.texi: Regenerate.
10414 * targhooks.h (default_empty_mask_is_expensive): Declare.
10415 * targhooks.c (default_empty_mask_is_expensive): New function.
10416 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10417 if the target says that empty masks are expensive.
10418 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10420 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10422 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10423 Alan Hayward <alan.hayward@arm.com>
10424 David Sherwood <david.sherwood@arm.com>
10426 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10427 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10428 (vect_use_loop_mask_for_alignment_p): New function.
10429 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10430 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10431 niters_skip argument. Make sure that the first niters_skip elements
10432 of the first iteration are inactive.
10433 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10434 Update call to vect_set_loop_masks_directly.
10435 (get_misalign_in_elems): New function, split out from...
10436 (vect_gen_prolog_loop_niters): ...here.
10437 (vect_update_init_of_dr): Take a code argument that specifies whether
10438 the adjustment should be added or subtracted.
10439 (vect_update_init_of_drs): Likewise.
10440 (vect_prepare_for_masked_peels): New function.
10441 (vect_do_peeling): Skip prologue peeling if we're using a mask
10442 instead. Update call to vect_update_inits_of_drs.
10443 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10445 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10446 alignment. Do not include the number of peeled iterations in
10447 the minimum threshold in that case.
10448 (vectorizable_induction): Adjust the start value down by
10449 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10450 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10451 Take the number of skipped iterations into account when calculating
10453 * tree-vect-stmts.c (vect_gen_while_not): New function.
10455 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10456 Alan Hayward <alan.hayward@arm.com>
10457 David Sherwood <david.sherwood@arm.com>
10459 * doc/sourcebuild.texi (vect_fully_masked): Document.
10460 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10461 default value to 0.
10462 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10464 (vect_analyze_loop_2): ...here. Don't check the vectorization
10465 factor against the number of loop iterations if the loop is
10468 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10469 Alan Hayward <alan.hayward@arm.com>
10470 David Sherwood <david.sherwood@arm.com>
10472 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10473 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10474 (dump_groups): Update accordingly.
10475 (iv_use::mem_type): New member variable.
10476 (address_p): New function.
10477 (record_use): Add a mem_type argument and initialize the new
10479 (record_group_use): Add a mem_type argument. Use address_p.
10480 Remove obsolete null checks of base_object. Update call to record_use.
10481 (find_interesting_uses_op): Update call to record_group_use.
10482 (find_interesting_uses_cond): Likewise.
10483 (find_interesting_uses_address): Likewise.
10484 (get_mem_type_for_internal_fn): New function.
10485 (find_address_like_use): Likewise.
10486 (find_interesting_uses_stmt): Try find_address_like_use before
10487 calling find_interesting_uses_op.
10488 (addr_offset_valid_p): Use the iv mem_type field as the type
10489 of the addressed memory.
10490 (add_autoinc_candidates): Likewise.
10491 (get_address_cost): Likewise.
10492 (split_small_address_groups_p): Use address_p.
10493 (split_address_groups): Likewise.
10494 (add_iv_candidate_for_use): Likewise.
10495 (autoinc_possible_for_pair): Likewise.
10496 (rewrite_groups): Likewise.
10497 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10498 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10499 (get_alias_ptr_type_for_ptr_address): New function.
10500 (rewrite_use_address): Rewrite address uses in calls that were
10501 identified by find_address_like_use.
10503 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10504 Alan Hayward <alan.hayward@arm.com>
10505 David Sherwood <david.sherwood@arm.com>
10507 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10509 * gimple-expr.h (is_gimple_addressable: Likewise.
10510 * gimple-expr.c (is_gimple_address): Likewise.
10511 * internal-fn.c (expand_call_mem_ref): New function.
10512 (expand_mask_load_optab_fn): Use it.
10513 (expand_mask_store_optab_fn): Likewise.
10515 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10516 Alan Hayward <alan.hayward@arm.com>
10517 David Sherwood <david.sherwood@arm.com>
10519 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10520 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10521 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10522 (cond_umax@var{mode}): Document.
10523 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10524 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10525 (cond_umin_optab, cond_umax_optab): New optabs.
10526 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10527 (COND_IOR, COND_XOR): New internal functions.
10528 * internal-fn.h (get_conditional_internal_fn): Declare.
10529 * internal-fn.c (cond_binary_direct): New macro.
10530 (expand_cond_binary_optab_fn): Likewise.
10531 (direct_cond_binary_optab_supported_p): Likewise.
10532 (get_conditional_internal_fn): New function.
10533 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10534 Cope with reduction statements that are vectorized as calls rather
10536 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10537 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10538 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10539 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10540 (UNSPEC_COND_EOR): New unspecs.
10541 (optab): Add mappings for them.
10542 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10543 (sve_int_op, sve_fp_op): New int attributes.
10545 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10546 Alan Hayward <alan.hayward@arm.com>
10547 David Sherwood <david.sherwood@arm.com>
10549 * optabs.def (while_ult_optab): New optab.
10550 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10551 * internal-fn.def (WHILE_ULT): New internal function.
10552 * internal-fn.h (direct_internal_fn_supported_p): New override
10553 that takes two types as argument.
10554 * internal-fn.c (while_direct): New macro.
10555 (expand_while_optab_fn): New function.
10556 (convert_optab_supported_p): Likewise.
10557 (direct_while_optab_supported_p): New macro.
10558 * wide-int.h (wi::udiv_ceil): New function.
10559 * tree-vectorizer.h (rgroup_masks): New structure.
10560 (vec_loop_masks): New typedef.
10561 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10562 and fully_masked_p.
10563 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10564 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10565 (vect_max_vf): New function.
10566 (slpeel_make_loop_iterate_ntimes): Delete.
10567 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10568 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10569 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10570 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10571 internal-fn.h, stor-layout.h and optabs-query.h.
10572 (vect_set_loop_mask): New function.
10573 (add_preheader_seq): Likewise.
10574 (add_header_seq): Likewise.
10575 (interleave_supported_p): Likewise.
10576 (vect_maybe_permute_loop_masks): Likewise.
10577 (vect_set_loop_masks_directly): Likewise.
10578 (vect_set_loop_condition_masked): Likewise.
10579 (vect_set_loop_condition_unmasked): New function, split out from
10580 slpeel_make_loop_iterate_ntimes.
10581 (slpeel_make_loop_iterate_ntimes): Rename to..
10582 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10583 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10584 (vect_do_peeling): Update call accordingly.
10585 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10587 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10588 mask_compare_type, can_fully_mask_p and fully_masked_p.
10589 (release_vec_loop_masks): New function.
10590 (_loop_vec_info): Use it to free the loop masks.
10591 (can_produce_all_loop_masks_p): New function.
10592 (vect_get_max_nscalars_per_iter): Likewise.
10593 (vect_verify_full_masking): Likewise.
10594 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10595 retries, and free the mask rgroups before retrying. Check loop-wide
10596 reasons for disallowing fully-masked loops. Make the final decision
10597 about whether use a fully-masked loop or not.
10598 (vect_estimate_min_profitable_iters): Do not assume that peeling
10599 for the number of iterations will be needed for fully-masked loops.
10600 (vectorizable_reduction): Disable fully-masked loops.
10601 (vectorizable_live_operation): Likewise.
10602 (vect_halve_mask_nunits): New function.
10603 (vect_double_mask_nunits): Likewise.
10604 (vect_record_loop_mask): Likewise.
10605 (vect_get_loop_mask): Likewise.
10606 (vect_transform_loop): Handle the case in which the final loop
10607 iteration might handle a partial vector. Call vect_set_loop_condition
10608 instead of slpeel_make_loop_iterate_ntimes.
10609 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10610 (check_load_store_masking): New function.
10611 (prepare_load_store_mask): Likewise.
10612 (vectorizable_store): Handle fully-masked loops.
10613 (vectorizable_load): Likewise.
10614 (supportable_widening_operation): Use vect_halve_mask_nunits for
10616 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10617 (vect_gen_while): New function.
10618 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10619 (aarch64_uqdec<mode>): New insn.
10621 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10622 Alan Hayward <alan.hayward@arm.com>
10623 David Sherwood <david.sherwood@arm.com>
10625 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10626 (reduc_xor_scal_optab): New optabs.
10627 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10628 (reduc_xor_scal_@var{m}): Document.
10629 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10630 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10631 internal functions.
10632 * fold-const-call.c (fold_const_call): Handle them.
10633 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10634 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10635 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10636 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10637 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10638 (UNSPEC_XORV): New unspecs.
10639 (optab): Add entries for them.
10640 (BITWISEV): New int iterator.
10641 (bit_reduc_op): New int attributes.
10643 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10644 Alan Hayward <alan.hayward@arm.com>
10645 David Sherwood <david.sherwood@arm.com>
10647 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10648 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10649 * optabs.def (vec_shl_insert_optab): New optab.
10650 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10651 (duplicate_and_interleave): Likewise.
10652 * tree-vect-loop.c: Include internal-fn.h.
10653 (neutral_op_for_slp_reduction): New function, split out from
10654 get_initial_defs_for_reduction.
10655 (get_initial_def_for_reduction): Handle option 2 for variable-length
10656 vectors by loading the neutral value into a vector and then shifting
10657 the initial value into element 0.
10658 (get_initial_defs_for_reduction): Replace the code argument with
10659 the neutral value calculated by neutral_op_for_slp_reduction.
10660 Use gimple_build_vector for constant-length vectors.
10661 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10662 but the first group_size elements have a neutral value.
10663 Use duplicate_and_interleave otherwise.
10664 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10665 Update call to get_initial_defs_for_reduction. Handle SLP
10666 reductions for variable-length vectors by creating one vector
10667 result for each scalar result, with the elements associated
10668 with other scalar results stubbed out with the neutral value.
10669 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10670 Require IFN_VEC_SHL_INSERT for double reductions on
10671 variable-length vectors, or SLP reductions that have
10672 a neutral value. Require can_duplicate_and_interleave_p
10673 support for variable-length unchained SLP reductions if there
10674 is no neutral value, such as for MIN/MAX reductions. Also require
10675 the number of vector elements to be a multiple of the number of
10676 SLP statements when doing variable-length unchained SLP reductions.
10677 Update call to vect_create_epilog_for_reduction.
10678 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10679 and remove initial values.
10680 (duplicate_and_interleave): Make public.
10681 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10682 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10684 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10685 Alan Hayward <alan.hayward@arm.com>
10686 David Sherwood <david.sherwood@arm.com>
10688 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10689 (can_duplicate_and_interleave_p): New function.
10690 (vect_get_and_check_slp_defs): Take the vector of statements
10691 rather than just the current one. Remove excess parentheses.
10692 Restriction rejectinon of vect_constant_def and vect_external_def
10693 for variable-length vectors to boolean types, or types for which
10694 can_duplicate_and_interleave_p is false.
10695 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10696 (duplicate_and_interleave): New function.
10697 (vect_get_constant_vectors): Use gimple_build_vector for
10698 constant-length vectors and suitable variable-length constant
10699 vectors. Use duplicate_and_interleave for other variable-length
10700 vectors. Don't defer the update when inserting new statements.
10702 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10703 Alan Hayward <alan.hayward@arm.com>
10704 David Sherwood <david.sherwood@arm.com>
10706 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10707 min_profitable_iters doesn't go negative.
10709 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10710 Alan Hayward <alan.hayward@arm.com>
10711 David Sherwood <david.sherwood@arm.com>
10713 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10714 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10715 * optabs.def (vec_mask_load_lanes_optab): New optab.
10716 (vec_mask_store_lanes_optab): Likewise.
10717 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10718 (MASK_STORE_LANES): Likewise.
10719 * internal-fn.c (mask_load_lanes_direct): New macro.
10720 (mask_store_lanes_direct): Likewise.
10721 (expand_mask_load_optab_fn): Handle masked operations.
10722 (expand_mask_load_lanes_optab_fn): New macro.
10723 (expand_mask_store_optab_fn): Handle masked operations.
10724 (expand_mask_store_lanes_optab_fn): New macro.
10725 (direct_mask_load_lanes_optab_supported_p): Likewise.
10726 (direct_mask_store_lanes_optab_supported_p): Likewise.
10727 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10729 (vect_load_lanes_supported): Likewise.
10730 * tree-vect-data-refs.c (strip_conversion): New function.
10731 (can_group_stmts_p): Likewise.
10732 (vect_analyze_data_ref_accesses): Use it instead of checking
10733 for a pair of assignments.
10734 (vect_store_lanes_supported): Take a masked_p parameter.
10735 (vect_load_lanes_supported): Likewise.
10736 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10737 vect_store_lanes_supported and vect_load_lanes_supported.
10738 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10739 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10740 parameter. Don't allow gaps for masked accesses.
10741 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10742 and vect_load_lanes_supported.
10743 (get_load_store_type): Take a masked_p parameter and update
10744 call to get_group_load_store_type.
10745 (vectorizable_store): Update call to get_load_store_type.
10746 Handle IFN_MASK_STORE_LANES.
10747 (vectorizable_load): Update call to get_load_store_type.
10748 Handle IFN_MASK_LOAD_LANES.
10750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10751 Alan Hayward <alan.hayward@arm.com>
10752 David Sherwood <david.sherwood@arm.com>
10754 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10756 * config/aarch64/aarch64-protos.h
10757 (aarch64_sve_struct_memory_operand_p): Declare.
10758 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10759 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10760 (VPRED, vpred): Handle SVE structure modes.
10761 * config/aarch64/constraints.md (Utx): New constraint.
10762 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10763 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10764 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10765 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10766 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10767 structure modes. Split into pieces after RA.
10768 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10769 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10771 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10772 SVE structure modes.
10773 (aarch64_classify_address): Likewise.
10774 (sizetochar): Move earlier in file.
10775 (aarch64_print_operand): Handle SVE register lists.
10776 (aarch64_array_mode): New function.
10777 (aarch64_sve_struct_memory_operand_p): Likewise.
10778 (TARGET_ARRAY_MODE): Redefine.
10780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10781 Alan Hayward <alan.hayward@arm.com>
10782 David Sherwood <david.sherwood@arm.com>
10784 * target.def (array_mode): New target hook.
10785 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10786 * doc/tm.texi: Regenerate.
10787 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10788 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10789 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10790 targetm.array_mode.
10791 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10794 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10795 Alan Hayward <alan.hayward@arm.com>
10796 David Sherwood <david.sherwood@arm.com>
10798 * fold-const.c (fold_binary_loc): Check the argument types
10799 rather than the result type when testing for a vector operation.
10801 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10803 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10804 * doc/tm.texi: Regenerate.
10806 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10807 Alan Hayward <alan.hayward@arm.com>
10808 David Sherwood <david.sherwood@arm.com>
10810 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10811 (sve): Document new AArch64 extension.
10812 * doc/md.texi (w): Extend the description of the AArch64
10813 constraint to include SVE vectors.
10814 (Upl, Upa): Document new AArch64 predicate constraints.
10815 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10817 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10818 (msve-vector-bits=): New option.
10819 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10820 SVE when these are disabled.
10821 (sve): New extension.
10822 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10823 modes. Adjust their number of units based on aarch64_sve_vg.
10824 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10825 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10826 aarch64_addr_query_type.
10827 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10828 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10829 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10830 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10831 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10832 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10833 (aarch64_simd_imm_zero_p): Delete.
10834 (aarch64_check_zero_based_sve_index_immediate): Declare.
10835 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10836 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10837 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10838 (aarch64_sve_float_mul_immediate_p): Likewise.
10839 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10840 rather than an rtx.
10841 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10842 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10843 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10844 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10845 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10846 (aarch64_regmode_natural_size): Likewise.
10847 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10848 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10850 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10851 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10852 for VG and the SVE predicate registers.
10853 (V_ALIASES): Add a "z"-prefixed alias.
10854 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10855 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10856 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10857 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10858 (REG_CLASS_NAMES): Add entries for them.
10859 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10860 and the predicate registers.
10861 (aarch64_sve_vg): Declare.
10862 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10863 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10864 (REGMODE_NATURAL_SIZE): Define.
10865 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10867 * config/aarch64/aarch64.c: Include cfgrtl.h.
10868 (simd_immediate_info): Add a constructor for series vectors,
10869 and an associated step field.
10870 (aarch64_sve_vg): New variable.
10871 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10872 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10873 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10874 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10875 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10876 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10877 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10878 (aarch64_get_mask_mode): New functions.
10879 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10880 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10881 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10882 predicate modes and predicate registers. Explicitly restrict
10883 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10884 to store a vector mode if it is recognized by
10885 aarch64_classify_vector_mode.
10886 (aarch64_regmode_natural_size): New function.
10887 (aarch64_hard_regno_caller_save_mode): Return the original mode
10889 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10890 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10891 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10892 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10894 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10895 does not overlap dest if the function is frame-related. Handle
10897 (aarch64_split_add_offset): New function.
10898 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10899 them aarch64_add_offset.
10900 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10901 and update call to aarch64_sub_sp.
10902 (aarch64_add_cfa_expression): New function.
10903 (aarch64_expand_prologue): Pass extra temporary registers to the
10904 functions above. Handle the case in which we need to emit new
10905 DW_CFA_expressions for registers that were originally saved
10906 relative to the stack pointer, but now have to be expressed
10907 relative to the frame pointer.
10908 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10910 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10911 IP0 and IP1 values for SVE frames.
10912 (aarch64_expand_vec_series): New function.
10913 (aarch64_expand_sve_widened_duplicate): Likewise.
10914 (aarch64_expand_sve_const_vector): Likewise.
10915 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10916 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10917 into the register, rather than emitting a SET directly.
10918 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10919 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10920 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10921 (offset_9bit_signed_scaled_p): New functions.
10922 (aarch64_replicate_bitmask_imm): New function.
10923 (aarch64_bitmask_imm): Use it.
10924 (aarch64_cannot_force_const_mem): Reject expressions involving
10925 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10926 (aarch64_classify_index): Handle SVE indices, by requiring
10927 a plain register index with a scale that matches the element size.
10928 (aarch64_classify_address): Handle SVE addresses. Assert that
10929 the mode of the address is VOIDmode or an integer mode.
10930 Update call to aarch64_classify_symbol.
10931 (aarch64_classify_symbolic_expression): Update call to
10932 aarch64_classify_symbol.
10933 (aarch64_const_vec_all_in_range_p): New function.
10934 (aarch64_print_vector_float_operand): Likewise.
10935 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10936 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10937 and the FP immediates 1.0 and 0.5.
10938 (aarch64_print_address_internal): Handle SVE addresses.
10939 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10940 (aarch64_regno_regclass): Handle predicate registers.
10941 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10943 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10944 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10945 (aarch64_convert_sve_vector_bits): New function.
10946 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10947 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10948 rather than an rtx.
10949 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10950 Handle SVE vector and predicate modes. Accept VL-based constants
10951 that need only one temporary register, and VL offsets that require
10952 no temporary registers.
10953 (aarch64_conditional_register_usage): Mark the predicate registers
10954 as fixed if SVE isn't available.
10955 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10956 Return true for SVE vector and predicate modes.
10957 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10958 rather than an unsigned int. Handle SVE modes.
10959 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10961 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10963 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10964 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10965 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10966 (aarch64_sve_float_mul_immediate_p): New functions.
10967 (aarch64_sve_valid_immediate): New function.
10968 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10969 Explicitly reject structure modes. Check for INDEX constants.
10970 Handle PTRUE and PFALSE constants.
10971 (aarch64_check_zero_based_sve_index_immediate): New function.
10972 (aarch64_simd_imm_zero_p): Delete.
10973 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10974 vector modes. Accept constants in the range of CNT[BHWD].
10975 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10976 ask for an Advanced SIMD mode.
10977 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10978 (aarch64_simd_vector_alignment): Handle SVE predicates.
10979 (aarch64_vectorize_preferred_vector_alignment): New function.
10980 (aarch64_simd_vector_alignment_reachable): Use it instead of
10982 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10983 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10985 (MAX_VECT_LEN): Delete.
10986 (expand_vec_perm_d): Add a vec_flags field.
10987 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10988 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10989 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10991 (aarch64_evpc_rev): Rename to...
10992 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10993 (aarch64_evpc_rev_global): New function.
10994 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10995 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10997 (aarch64_evpc_sve_tbl): New function.
10998 (aarch64_expand_vec_perm_const_1): Update after rename of
10999 aarch64_evpc_rev. Handle SVE permutes too, trying
11000 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11001 than aarch64_evpc_tbl.
11002 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11003 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11004 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11005 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11006 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11007 (aarch64_expand_sve_vcond): New functions.
11008 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11009 of aarch64_vector_mode_p.
11010 (aarch64_dwarf_poly_indeterminate_value): New function.
11011 (aarch64_compute_pressure_classes): Likewise.
11012 (aarch64_can_change_mode_class): Likewise.
11013 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11014 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11015 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11016 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11017 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11018 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11019 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11020 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11022 (Dn, Dl, Dr): Accept const as well as const_vector.
11023 (Dz): Likewise. Compare against CONST0_RTX.
11024 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11025 of "vector" where appropriate.
11026 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11027 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11028 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11029 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11030 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11031 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11032 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11033 (v_int_equiv): Extend to SVE modes.
11034 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11036 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11037 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11038 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11039 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11040 (SVE_COND_FP_CMP): New int iterators.
11041 (perm_hilo): Handle the new unpack unspecs.
11042 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11044 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11045 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11046 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11047 (aarch64_equality_operator, aarch64_constant_vector_operand)
11048 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11049 (aarch64_sve_nonimmediate_operand): Likewise.
11050 (aarch64_sve_general_operand): Likewise.
11051 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11052 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11053 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11054 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11055 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11056 (aarch64_sve_float_arith_immediate): Likewise.
11057 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11058 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11059 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11060 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11061 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11062 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11063 (aarch64_sve_float_arith_operand): Likewise.
11064 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11065 (aarch64_sve_float_mul_operand): Likewise.
11066 (aarch64_sve_vec_perm_operand): Likewise.
11067 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11068 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11069 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11070 as well as const_vector.
11071 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11072 in file. Use CONST0_RTX and CONSTM1_RTX.
11073 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11074 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11075 Use aarch64_simd_imm_zero.
11076 * config/aarch64/aarch64-sve.md: New file.
11077 * config/aarch64/aarch64.md: Include it.
11078 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11079 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11080 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11081 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11082 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11083 (sve): New attribute.
11084 (enabled): Disable instructions with the sve attribute unless
11086 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11087 aarch64_expand_mov_immediate.
11088 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11089 CNT[BHSD] immediates.
11090 (movti): Split CONST_POLY_INT moves into two halves.
11091 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11092 Split additions that need a temporary here if the destination
11093 is the stack pointer.
11094 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11095 (*add<mode>3_poly_1): New instruction.
11096 (set_clobber_cc): New expander.
11098 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11100 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11101 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11102 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11103 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11104 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11105 Change innermode from fixed_mode_size to machine_mode.
11106 (simplify_subreg): Update call accordingly. Handle a constant-sized
11107 subreg of a variable-length CONST_VECTOR.
11109 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11110 Alan Hayward <alan.hayward@arm.com>
11111 David Sherwood <david.sherwood@arm.com>
11113 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11114 (add_offset_to_base): New function, split out from...
11115 (create_mem_ref): ...here. When handling a scale other than 1,
11116 check first whether the address is valid without the offset.
11117 Add it into the base if so, leaving the index and scale as-is.
11119 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11122 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11123 fold_for_warn before checking if arg2 is INTEGER_CST.
11125 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11127 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11128 (store_multiple_operation): Delete.
11129 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11130 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11131 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11132 guarded by TARGET_STRING.
11133 (rs6000_output_load_multiple): Delete.
11134 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11135 OPTION_MASK_STRING / TARGET_STRING handling.
11136 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11137 (const rs6000_opt_masks) <"string">: Change mask to 0.
11138 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11139 (MASK_STRING): Delete.
11140 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11142 (load_multiple): Delete.
11149 (store_multiple): Delete.
11156 (movmemsi_8reg): Delete.
11157 (corresponding unnamed define_insn): Delete.
11158 (movmemsi_6reg): Delete.
11159 (corresponding unnamed define_insn): Delete.
11160 (movmemsi_4reg): Delete.
11161 (corresponding unnamed define_insn): Delete.
11162 (movmemsi_2reg): Delete.
11163 (corresponding unnamed define_insn): Delete.
11164 (movmemsi_1reg): Delete.
11165 (corresponding unnamed define_insn): Delete.
11166 * config/rs6000/rs6000.opt (mno-string): New.
11167 (mstring): Replace by deprecation warning stub.
11168 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11170 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11172 * regrename.c (regrename_do_replace): If replacing the same
11173 reg multiple times, try to reuse last created gen_raw_REG.
11176 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11177 main to workaround a bug in GDB.
11179 2018-01-12 Tom de Vries <tom@codesourcery.com>
11182 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11184 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11186 PR rtl-optimization/80481
11187 * ira-color.c (get_cap_member): New function.
11188 (allocnos_conflict_by_live_ranges_p): Use it.
11189 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11190 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11192 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11195 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11196 (*saddl_se_1): Ditto.
11197 (*ssubsi_1): Ditto.
11198 (*ssubl_se_1): Ditto.
11200 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11202 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11203 rather than wi::to_widest for DR_INITs.
11204 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11205 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11206 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11208 (vect_analyze_group_access_1): Note that here.
11210 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11212 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11213 polynomial type sizes.
11215 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11217 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11218 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11219 (gimple_add_tmp_var): Likewise.
11221 2018-01-12 Martin Liska <mliska@suse.cz>
11223 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11224 (gimple_alloc_sizes): Likewise.
11225 (dump_gimple_statistics): Use PRIu64 in printf format.
11226 * gimple.h: Change uint64_t to int.
11228 2018-01-12 Martin Liska <mliska@suse.cz>
11230 * tree-core.h: Use uint64_t instead of int.
11231 * tree.c (tree_node_counts): Likewise.
11232 (tree_node_sizes): Likewise.
11233 (dump_tree_statistics): Use PRIu64 in printf format.
11235 2018-01-12 Martin Liska <mliska@suse.cz>
11237 * Makefile.in: As qsort_chk is implemented in vec.c, add
11238 vec.o to linkage of gencfn-macros.
11239 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11240 passing the info to record_node_allocation_statistics.
11241 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11243 * ggc-common.c (struct ggc_usage): Add operator== and use
11244 it in operator< and compare function.
11245 * mem-stats.h (struct mem_usage): Likewise.
11246 * vec.c (struct vec_usage): Remove operator< and compare
11247 function. Can be simply inherited.
11249 2018-01-12 Martin Jambor <mjambor@suse.cz>
11252 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11253 * tree-ssa-math-opts.c: Include domwalk.h.
11254 (convert_mult_to_fma_1): New function.
11255 (fma_transformation_info): New type.
11256 (fma_deferring_state): Likewise.
11257 (cancel_fma_deferring): New function.
11258 (result_of_phi): Likewise.
11259 (last_fma_candidate_feeds_initial_phi): Likewise.
11260 (convert_mult_to_fma): Added deferring logic, split actual
11261 transformation to convert_mult_to_fma_1.
11262 (math_opts_dom_walker): New type.
11263 (math_opts_dom_walker::after_dom_children): New method, body moved
11264 here from pass_optimize_widening_mul::execute, added deferring logic
11266 (pass_optimize_widening_mul::execute): Moved most of code to
11267 math_opts_dom_walker::after_dom_children.
11268 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11269 * config/i386/i386.c (ix86_option_override_internal): Added
11270 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11272 2018-01-12 Richard Biener <rguenther@suse.de>
11275 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11276 inline instance vars.
11278 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11281 * config/rx/rx.c (rx_is_restricted_memory_address):
11282 Handle SUBREG case.
11284 2018-01-12 Richard Biener <rguenther@suse.de>
11286 PR tree-optimization/80846
11287 * target.def (split_reduction): New target hook.
11288 * targhooks.c (default_split_reduction): New function.
11289 * targhooks.h (default_split_reduction): Declare.
11290 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11291 target requests first reduce vectors by combining low and high
11293 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11294 (get_vectype_for_scalar_type_and_size): Export.
11295 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11296 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11297 * doc/tm.texi: Regenerate.
11298 * config/i386/i386.c (ix86_split_reduction): Implement
11299 TARGET_VECTORIZE_SPLIT_REDUCTION.
11301 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11304 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11305 in PIC mode except for TARGET_VXWORKS_RTP.
11306 * config/sparc/sparc.c: Include cfgrtl.h.
11307 (TARGET_INIT_PIC_REG): Define.
11308 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11309 (sparc_pic_register_p): New predicate.
11310 (sparc_legitimate_address_p): Use it.
11311 (sparc_legitimize_pic_address): Likewise.
11312 (sparc_delegitimize_address): Likewise.
11313 (sparc_mode_dependent_address_p): Likewise.
11314 (gen_load_pcrel_sym): Remove 4th parameter.
11315 (load_got_register): Adjust call to above. Remove obsolete stuff.
11316 (sparc_expand_prologue): Do not call load_got_register here.
11317 (sparc_flat_expand_prologue): Likewise.
11318 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11319 (sparc_use_pseudo_pic_reg): New function.
11320 (sparc_init_pic_reg): Likewise.
11321 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11322 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11324 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11326 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11327 Add item for branch_cost.
11329 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11331 PR rtl-optimization/83565
11332 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11333 not extend the result to a larger mode for rotate operations.
11334 (num_sign_bit_copies1): Likewise.
11336 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11339 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11341 Use values-Xc.o for -pedantic.
11342 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11344 2018-01-12 Martin Liska <mliska@suse.cz>
11347 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11349 (possible_polymorphic_call_targets): Use it.
11350 (ipa_devirt): Likewise.
11352 2018-01-12 Martin Liska <mliska@suse.cz>
11354 * profile-count.h (enum profile_quality): Use 0 as invalid
11355 enum value of profile_quality.
11357 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11359 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11360 -mext-string options.
11362 2018-01-12 Richard Biener <rguenther@suse.de>
11364 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11365 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11366 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11368 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11370 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11372 * configure.ac (--with-long-double-format): Add support for the
11373 configuration option to change the default long double format on
11375 * config.gcc (powerpc*-linux*-*): Likewise.
11376 * configure: Regenerate.
11377 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11378 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11379 used without modification.
11381 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11383 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11384 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11385 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11386 MISC_BUILTIN_SPEC_BARRIER.
11387 (rs6000_init_builtins): Likewise.
11388 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11390 (speculation_barrier): New define_insn.
11391 * doc/extend.texi: Document __builtin_speculation_barrier.
11393 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11396 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11397 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11398 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11400 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11401 integral modes instead of "ss" and "sd".
11402 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11403 vectors with 32-bit and 64-bit elements.
11404 (vecdupssescalarmodesuffix): New mode attribute.
11405 (vec_dup<mode>): Use it.
11407 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11410 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11411 frame if argument is passed on stack.
11413 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11416 * ree.c (combine_reaching_defs): Optimize also
11417 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11418 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11420 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11422 PR middle-end/83189
11423 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11425 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11427 PR middle-end/83718
11428 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11429 after they are computed.
11431 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11433 PR tree-optimization/83695
11434 * gimple-loop-linterchange.cc
11435 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11436 reset cached scev information after interchange.
11437 (pass_linterchange::execute): Remove call to scev_reset_htab.
11439 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11441 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11442 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11443 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11444 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11445 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11446 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11447 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11448 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11449 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11450 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11451 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11452 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11453 (V_lane_reg): Likewise.
11454 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11456 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11457 (vfmal_lane_low<mode>_intrinsic,
11458 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11459 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11460 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11461 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11462 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11463 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11465 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11467 * config/arm/arm-cpus.in (fp16fml): New feature.
11468 (ALL_SIMD): Add fp16fml.
11469 (armv8.2-a): Add fp16fml as an option.
11470 (armv8.3-a): Likewise.
11471 (armv8.4-a): Add fp16fml as part of fp16.
11472 * config/arm/arm.h (TARGET_FP16FML): Define.
11473 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11475 * config/arm/arm-modes.def (V2HF): Define.
11476 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11477 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11478 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11479 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11480 vfmsl_low, vfmsl_high): New set of builtins.
11481 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11482 (vfml_op): New code attribute.
11483 (VFMLHALVES): New int iterator.
11484 (VFML, VFMLSEL): New mode attributes.
11485 (V_reg): Define mapping for V2HF.
11486 (V_hi, V_lo): New mode attributes.
11487 (VF_constraint): Likewise.
11488 (vfml_half, vfml_half_selector): New int attributes.
11489 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11491 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11492 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11494 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11495 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11496 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11497 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11499 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11500 Document new effective target and option set.
11502 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11504 * config/arm/arm-cpus.in (armv8_4): New feature.
11505 (ARMv8_4a): New fgroup.
11506 (armv8.4-a): New arch.
11507 * config/arm/arm-tables.opt: Regenerate.
11508 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11509 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11510 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11511 Add matching rules for -march=armv8.4-a and extensions.
11512 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11514 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11517 * config/rx/rx.md (BW): New mode attribute.
11518 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11520 2018-01-11 Richard Biener <rguenther@suse.de>
11522 PR tree-optimization/83435
11523 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11524 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11525 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11527 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11528 Alan Hayward <alan.hayward@arm.com>
11529 David Sherwood <david.sherwood@arm.com>
11531 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11533 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11534 (aarch64_print_address_internal): Use it to check for a zero offset.
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-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11541 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11542 Return a poly_int64 rather than a HOST_WIDE_INT.
11543 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11544 rather than a HOST_WIDE_INT.
11545 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11546 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11547 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11548 final_offset from HOST_WIDE_INT to poly_int64.
11549 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11550 to_constant when getting the number of units in an Advanced SIMD
11552 (aarch64_builtin_vectorized_function): Check for a constant number
11554 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11556 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11557 attribute instead of GET_MODE_NUNITS.
11558 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11559 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11560 GET_MODE_SIZE for fixed-size registers.
11561 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11562 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11563 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11564 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11565 (aarch64_print_operand, aarch64_print_address_internal)
11566 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11567 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11568 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11569 Handle polynomial GET_MODE_SIZE.
11570 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11571 wider than SImode without modification.
11572 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11573 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11574 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11575 passing and returning SVE modes.
11576 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11577 rather than GEN_INT.
11578 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11579 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11580 (aarch64_allocate_and_probe_stack_space): Likewise.
11581 (aarch64_layout_frame): Cope with polynomial offsets.
11582 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11583 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11584 polynomial offsets.
11585 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11586 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11587 poly_int64 rather than a HOST_WIDE_INT.
11588 (aarch64_get_separate_components, aarch64_process_components)
11589 (aarch64_expand_prologue, aarch64_expand_epilogue)
11590 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11591 (aarch64_anchor_offset): New function, split out from...
11592 (aarch64_legitimize_address): ...here.
11593 (aarch64_builtin_vectorization_cost): Handle polynomial
11594 TYPE_VECTOR_SUBPARTS.
11595 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11597 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11598 number of elements from the PARALLEL rather than the mode.
11599 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11600 rather than GET_MODE_BITSIZE.
11601 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11602 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11603 (aarch64_expand_vec_perm_const_1): Handle polynomial
11604 d->perm.length () and d->perm elements.
11605 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11606 Apply to_constant to d->perm elements.
11607 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11608 polynomial CONST_VECTOR_NUNITS.
11609 (aarch64_move_pointer): Take amount as a poly_int64 rather
11611 (aarch64_progress_pointer): Avoid temporary variable.
11612 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11613 the mode attribute instead of GET_MODE.
11615 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11616 Alan Hayward <alan.hayward@arm.com>
11617 David Sherwood <david.sherwood@arm.com>
11619 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11620 x exists before using it.
11621 (aarch64_add_constant_internal): Rename to...
11622 (aarch64_add_offset_1): ...this. Replace regnum with separate
11623 src and dest rtxes. Handle the case in which they're different,
11624 including when the offset is zero. Replace scratchreg with an rtx.
11625 Use 2 additions if there is no spare register into which we can
11626 move a 16-bit constant.
11627 (aarch64_add_constant): Delete.
11628 (aarch64_add_offset): Replace reg with separate src and dest
11629 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11630 Use aarch64_add_offset_1.
11631 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11632 an rtx rather than an int. Take the delta as a poly_int64
11633 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11634 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11635 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11636 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11637 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11638 and aarch64_add_sp.
11639 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11640 aarch64_add_constant.
11642 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11644 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11645 Use scalar_float_mode.
11647 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11649 * config/aarch64/aarch64-simd.md
11650 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11651 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11652 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11653 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11654 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11655 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11656 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11657 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11658 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11659 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11661 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11664 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11665 targ_options->x_arm_arch_string is non NULL.
11667 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11669 * config/aarch64/aarch64.h
11670 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11672 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11675 * expmed.c (emit_store_flag_force): Swap if const op0
11676 and change VOIDmode to mode of op0.
11678 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11680 PR rtl-optimization/83761
11681 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11682 than bytes to mode_for_size.
11684 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11686 PR middle-end/83189
11687 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11688 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11691 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11693 PR middle-end/83575
11694 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11695 when in layout mode.
11696 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11697 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11700 2018-01-10 Michael Collison <michael.collison@arm.com>
11702 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11703 * config/aarch64/aarch64-option-extension.def: Add
11704 AARCH64_OPT_EXTENSION of 'fp16fml'.
11705 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11706 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11707 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11708 * config/aarch64/constraints.md (Ui7): New constraint.
11709 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11710 (VFMLA_SEL_W): Ditto.
11713 (VFMLA16_LOW): New int iterator.
11714 (VFMLA16_HIGH): Ditto.
11715 (UNSPEC_FMLAL): New unspec.
11716 (UNSPEC_FMLSL): Ditto.
11717 (UNSPEC_FMLAL2): Ditto.
11718 (UNSPEC_FMLSL2): Ditto.
11719 (f16mac): New code attribute.
11720 * config/aarch64/aarch64-simd-builtins.def
11721 (aarch64_fmlal_lowv2sf): Ditto.
11722 (aarch64_fmlsl_lowv2sf): Ditto.
11723 (aarch64_fmlalq_lowv4sf): Ditto.
11724 (aarch64_fmlslq_lowv4sf): Ditto.
11725 (aarch64_fmlal_highv2sf): Ditto.
11726 (aarch64_fmlsl_highv2sf): Ditto.
11727 (aarch64_fmlalq_highv4sf): Ditto.
11728 (aarch64_fmlslq_highv4sf): Ditto.
11729 (aarch64_fmlal_lane_lowv2sf): Ditto.
11730 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11731 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11732 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11733 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11734 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11735 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11736 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11737 (aarch64_fmlal_lane_highv2sf): Ditto.
11738 (aarch64_fmlsl_lane_highv2sf): Ditto.
11739 (aarch64_fmlal_laneq_highv2sf): Ditto.
11740 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11741 (aarch64_fmlalq_lane_highv4sf): Ditto.
11742 (aarch64_fmlsl_lane_highv4sf): Ditto.
11743 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11744 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11745 * config/aarch64/aarch64-simd.md:
11746 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11747 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11748 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11749 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11750 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11751 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11752 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11753 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11754 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11755 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11756 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11757 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11758 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11759 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11760 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11761 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11762 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11763 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11764 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11765 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11766 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11767 (vfmlsl_low_u32): Ditto.
11768 (vfmlalq_low_u32): Ditto.
11769 (vfmlslq_low_u32): Ditto.
11770 (vfmlal_high_u32): Ditto.
11771 (vfmlsl_high_u32): Ditto.
11772 (vfmlalq_high_u32): Ditto.
11773 (vfmlslq_high_u32): Ditto.
11774 (vfmlal_lane_low_u32): Ditto.
11775 (vfmlsl_lane_low_u32): Ditto.
11776 (vfmlal_laneq_low_u32): Ditto.
11777 (vfmlsl_laneq_low_u32): Ditto.
11778 (vfmlalq_lane_low_u32): Ditto.
11779 (vfmlslq_lane_low_u32): Ditto.
11780 (vfmlalq_laneq_low_u32): Ditto.
11781 (vfmlslq_laneq_low_u32): Ditto.
11782 (vfmlal_lane_high_u32): Ditto.
11783 (vfmlsl_lane_high_u32): Ditto.
11784 (vfmlal_laneq_high_u32): Ditto.
11785 (vfmlsl_laneq_high_u32): Ditto.
11786 (vfmlalq_lane_high_u32): Ditto.
11787 (vfmlslq_lane_high_u32): Ditto.
11788 (vfmlalq_laneq_high_u32): Ditto.
11789 (vfmlslq_laneq_high_u32): Ditto.
11790 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11791 (AARCH64_FL_FOR_ARCH8_4): New.
11792 (AARCH64_ISA_F16FML): New ISA flag.
11793 (TARGET_F16FML): New feature flag for fp16fml.
11794 (doc/invoke.texi): Document new fp16fml option.
11796 2018-01-10 Michael Collison <michael.collison@arm.com>
11798 * config/aarch64/aarch64-builtins.c:
11799 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11800 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11801 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11802 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11803 (AARCH64_ISA_SHA3): New ISA flag.
11804 (TARGET_SHA3): New feature flag for sha3.
11805 * config/aarch64/iterators.md (sha512_op): New int attribute.
11806 (CRYPTO_SHA512): New int iterator.
11807 (UNSPEC_SHA512H): New unspec.
11808 (UNSPEC_SHA512H2): Ditto.
11809 (UNSPEC_SHA512SU0): Ditto.
11810 (UNSPEC_SHA512SU1): Ditto.
11811 * config/aarch64/aarch64-simd-builtins.def
11812 (aarch64_crypto_sha512hqv2di): New builtin.
11813 (aarch64_crypto_sha512h2qv2di): Ditto.
11814 (aarch64_crypto_sha512su0qv2di): Ditto.
11815 (aarch64_crypto_sha512su1qv2di): Ditto.
11816 (aarch64_eor3qv8hi): Ditto.
11817 (aarch64_rax1qv2di): Ditto.
11818 (aarch64_xarqv2di): Ditto.
11819 (aarch64_bcaxqv8hi): Ditto.
11820 * config/aarch64/aarch64-simd.md:
11821 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11822 (aarch64_crypto_sha512su0qv2di): Ditto.
11823 (aarch64_crypto_sha512su1qv2di): Ditto.
11824 (aarch64_eor3qv8hi): Ditto.
11825 (aarch64_rax1qv2di): Ditto.
11826 (aarch64_xarqv2di): Ditto.
11827 (aarch64_bcaxqv8hi): Ditto.
11828 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11829 (vsha512h2q_u64): Ditto.
11830 (vsha512su0q_u64): Ditto.
11831 (vsha512su1q_u64): Ditto.
11832 (veor3q_u16): Ditto.
11833 (vrax1q_u64): Ditto.
11834 (vxarq_u64): Ditto.
11835 (vbcaxq_u16): Ditto.
11836 * config/arm/types.md (crypto_sha512): New type attribute.
11837 (crypto_sha3): Ditto.
11838 (doc/invoke.texi): Document new sha3 option.
11840 2018-01-10 Michael Collison <michael.collison@arm.com>
11842 * config/aarch64/aarch64-builtins.c:
11843 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11844 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11845 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11846 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11847 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11848 (AARCH64_ISA_SM4): New ISA flag.
11849 (TARGET_SM4): New feature flag for sm4.
11850 * config/aarch64/aarch64-simd-builtins.def
11851 (aarch64_sm3ss1qv4si): Ditto.
11852 (aarch64_sm3tt1aq4si): Ditto.
11853 (aarch64_sm3tt1bq4si): Ditto.
11854 (aarch64_sm3tt2aq4si): Ditto.
11855 (aarch64_sm3tt2bq4si): Ditto.
11856 (aarch64_sm3partw1qv4si): Ditto.
11857 (aarch64_sm3partw2qv4si): Ditto.
11858 (aarch64_sm4eqv4si): Ditto.
11859 (aarch64_sm4ekeyqv4si): Ditto.
11860 * config/aarch64/aarch64-simd.md:
11861 (aarch64_sm3ss1qv4si): Ditto.
11862 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11863 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11864 (aarch64_sm4eqv4si): Ditto.
11865 (aarch64_sm4ekeyqv4si): Ditto.
11866 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11867 (sm3part_op): Ditto.
11868 (CRYPTO_SM3TT): Ditto.
11869 (CRYPTO_SM3PART): Ditto.
11870 (UNSPEC_SM3SS1): New unspec.
11871 (UNSPEC_SM3TT1A): Ditto.
11872 (UNSPEC_SM3TT1B): Ditto.
11873 (UNSPEC_SM3TT2A): Ditto.
11874 (UNSPEC_SM3TT2B): Ditto.
11875 (UNSPEC_SM3PARTW1): Ditto.
11876 (UNSPEC_SM3PARTW2): Ditto.
11877 (UNSPEC_SM4E): Ditto.
11878 (UNSPEC_SM4EKEY): Ditto.
11879 * config/aarch64/constraints.md (Ui2): New constraint.
11880 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11881 * config/arm/types.md (crypto_sm3): New type attribute.
11882 (crypto_sm4): Ditto.
11883 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11884 (vsm3tt1aq_u32): Ditto.
11885 (vsm3tt1bq_u32): Ditto.
11886 (vsm3tt2aq_u32): Ditto.
11887 (vsm3tt2bq_u32): Ditto.
11888 (vsm3partw1q_u32): Ditto.
11889 (vsm3partw2q_u32): Ditto.
11890 (vsm4eq_u32): Ditto.
11891 (vsm4ekeyq_u32): Ditto.
11892 (doc/invoke.texi): Document new sm4 option.
11894 2018-01-10 Michael Collison <michael.collison@arm.com>
11896 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11897 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11898 (AARCH64_FL_FOR_ARCH8_4): New.
11899 (AARCH64_FL_V8_4): New flag.
11900 (doc/invoke.texi): Document new armv8.4-a option.
11902 2018-01-10 Michael Collison <michael.collison@arm.com>
11904 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11905 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11906 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11907 * config/aarch64/aarch64-option-extension.def: Add
11908 AARCH64_OPT_EXTENSION of 'sha2'.
11909 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11910 (crypto): Disable sha2 and aes if crypto disabled.
11911 (crypto): Enable aes and sha2 if enabled.
11912 (simd): Disable sha2 and aes if simd disabled.
11913 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11915 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11916 (TARGET_SHA2): New feature flag for sha2.
11917 (TARGET_AES): New feature flag for aes.
11918 * config/aarch64/aarch64-simd.md:
11919 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11920 conditional on TARGET_AES.
11921 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11922 (aarch64_crypto_sha1hsi): Make pattern conditional
11924 (aarch64_crypto_sha1hv4si): Ditto.
11925 (aarch64_be_crypto_sha1hv4si): Ditto.
11926 (aarch64_crypto_sha1su1v4si): Ditto.
11927 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11928 (aarch64_crypto_sha1su0v4si): Ditto.
11929 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11930 (aarch64_crypto_sha256su0v4si): Ditto.
11931 (aarch64_crypto_sha256su1v4si): Ditto.
11932 (doc/invoke.texi): Document new aes and sha2 options.
11934 2018-01-10 Martin Sebor <msebor@redhat.com>
11936 PR tree-optimization/83781
11937 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11940 2018-01-11 Martin Sebor <msebor@gmail.com>
11941 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11943 PR tree-optimization/83501
11944 PR tree-optimization/81703
11946 * tree-ssa-strlen.c (get_string_cst): Rename...
11947 (get_string_len): ...to this. Handle global constants.
11948 (handle_char_store): Adjust.
11950 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11951 Jim Wilson <jimw@sifive.com>
11953 * config/riscv/riscv-protos.h (riscv_output_return): New.
11954 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11955 (riscv_attribute_table, riscv_output_return),
11956 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11957 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11958 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11960 (riscv_expand_prologue): Add early return for naked function.
11961 (riscv_expand_epilogue): Likewise.
11962 (riscv_function_ok_for_sibcall): Return false for naked function.
11963 (riscv_set_current_function): New.
11964 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11965 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11966 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11967 * doc/extend.texi (RISC-V Function Attributes): New.
11969 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11971 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11972 check for 128-bit long double before checking TCmode.
11973 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11974 128-bit long doubles before checking TFmode or TCmode.
11975 (FLOAT128_IBM_P): Likewise.
11977 2018-01-10 Martin Sebor <msebor@redhat.com>
11979 PR tree-optimization/83671
11980 * builtins.c (c_strlen): Unconditionally return zero for the empty
11982 Use -Warray-bounds for warnings.
11983 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11984 for non-constant array indices with COMPONENT_REF, arrays of
11985 arrays, and pointers to arrays.
11986 (gimple_fold_builtin_strlen): Determine and set length range for
11987 non-constant character arrays.
11989 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11991 PR middle-end/81897
11992 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11995 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11997 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11999 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12002 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12003 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12004 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12005 indexed_or_indirect_operand predicate.
12006 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12007 (*vsx_le_perm_load_v8hi): Likewise.
12008 (*vsx_le_perm_load_v16qi): Likewise.
12009 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12010 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12011 (*vsx_le_perm_store_v8hi): Likewise.
12012 (*vsx_le_perm_store_v16qi): Likewise.
12013 (eight unnamed splitters): Likewise.
12015 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12017 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12018 * config/rs6000/emmintrin.h: Likewise.
12019 * config/rs6000/mmintrin.h: Likewise.
12020 * config/rs6000/xmmintrin.h: Likewise.
12022 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12025 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12027 * tree.c (tree_nop_conversion): Return true for location wrapper
12029 (maybe_wrap_with_location): New function.
12030 (selftest::check_strip_nops): New function.
12031 (selftest::test_location_wrappers): New function.
12032 (selftest::tree_c_tests): Call it.
12033 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12034 (maybe_wrap_with_location): New decl.
12035 (EXPR_LOCATION_WRAPPER_P): New macro.
12036 (location_wrapper_p): New inline function.
12037 (tree_strip_any_location_wrapper): New inline function.
12039 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12042 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12043 stack_realign_offset for the largest alignment of stack slot
12045 (ix86_find_max_used_stack_alignment): New function.
12046 (ix86_finalize_stack_frame_flags): Use it. Set
12047 max_used_stack_alignment if we don't realign stack.
12048 * config/i386/i386.h (machine_function): Add
12049 max_used_stack_alignment.
12051 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12053 * config/arm/arm.opt (-mbranch-cost): New option.
12054 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12057 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12060 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12061 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12063 2018-01-10 Richard Biener <rguenther@suse.de>
12066 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12067 early out so it also covers the case where we have a non-NULL
12070 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12072 PR tree-optimization/83753
12073 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12074 for non-strided grouped accesses if the number of elements is 1.
12076 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12079 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12080 * i386.h (TARGET_USE_GATHER): Define.
12081 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12083 2018-01-10 Martin Liska <mliska@suse.cz>
12086 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12087 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12089 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12090 CLEANUP_NO_PARTITIONING is not set.
12092 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12094 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12095 for vectors, as a partial revert of r254296.
12096 * rtl.h (const_vec_p): Delete.
12097 (const_vec_duplicate_p): Don't test for vector CONSTs.
12098 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12099 * expmed.c (make_tree): Likewise.
12102 * common.md (E, F): Use CONSTANT_P instead of checking for
12104 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12105 checking for CONST_VECTOR.
12107 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12109 PR middle-end/83575
12110 * predict.c (force_edge_cold): Handle in more sane way edges
12111 with no prediction.
12113 2018-01-09 Carl Love <cel@us.ibm.com>
12115 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12117 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12118 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12119 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12120 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12121 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12122 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12123 * config/rs6000/rs6000-protos.h: Add extern defition for
12124 rs6000_generate_float2_double_code.
12125 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12127 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12128 (float2_v2df): Add define_expand.
12130 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12133 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12134 op_mode in the force_to_mode call.
12136 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12138 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12139 instead of checking each element individually.
12140 (aarch64_evpc_uzp): Likewise.
12141 (aarch64_evpc_zip): Likewise.
12142 (aarch64_evpc_ext): Likewise.
12143 (aarch64_evpc_rev): Likewise.
12144 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12145 instead of checking each element individually. Return true without
12147 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12148 whether all selected elements come from the same input, instead of
12149 checking each element individually. Remove calls to gen_rtx_REG,
12150 start_sequence and end_sequence and instead assert that no rtl is
12153 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12155 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12156 order of HIGH and CONST checks.
12158 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12160 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12161 if the destination isn't an SSA_NAME.
12163 2018-01-09 Richard Biener <rguenther@suse.de>
12165 PR tree-optimization/83668
12166 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12168 (canonicalize_loop_form): ... here, renamed from ...
12169 (canonicalize_loop_closed_ssa_form): ... this and amended to
12170 swap successor edges for loop exit blocks to make us use
12171 the RPO order we need for initial schedule generation.
12173 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12175 PR tree-optimization/64811
12176 * match.pd: When optimizing comparisons with Inf, avoid
12177 introducing or losing exceptions from comparisons with NaN.
12179 2018-01-09 Martin Liska <mliska@suse.cz>
12182 * asan.c (shadow_mem_size): Add gcc_assert.
12184 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12186 Don't save registers in main().
12189 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12190 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12191 * config/avr/avr.c (avr_set_current_function): Don't error if
12192 naked, OS_task or OS_main are specified at the same time.
12193 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12195 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12197 * common/config/avr/avr-common.c (avr_option_optimization_table):
12198 Switch on -mmain-is-OS_task for optimizing compilations.
12200 2018-01-09 Richard Biener <rguenther@suse.de>
12202 PR tree-optimization/83572
12203 * graphite.c: Include cfganal.h.
12204 (graphite_transform_loops): Connect infinite loops to exit
12205 and remove fake edges at the end.
12207 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12209 * ipa-inline.c (edge_badness): Revert accidental checkin.
12211 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12214 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12215 symbols; not inline clones.
12217 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12220 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12221 hard registers. Formatting fixes.
12223 PR preprocessor/83722
12224 * gcc.c (try_generate_repro): Pass
12225 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12226 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12229 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12230 Kito Cheng <kito.cheng@gmail.com>
12232 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12233 (riscv_leaf_function_p): Delete.
12234 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12236 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12238 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12240 (do_ifelse): New function.
12241 (do_isel): New function.
12242 (do_sub3): New function.
12243 (do_add3): New function.
12244 (do_load_mask_compare): New function.
12245 (do_overlap_load_compare): New function.
12246 (expand_compare_loop): New function.
12247 (expand_block_compare): Call expand_compare_loop() when appropriate.
12248 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12249 option description.
12250 (-mblock-compare-inline-loop-limit): New option.
12252 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12255 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12256 Reverse order of second and third operands in first alternative.
12257 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12258 of first and second elements in UNSPEC_VPERMR vector.
12259 (altivec_expand_vec_perm_le): Likewise.
12261 2018-01-08 Jeff Law <law@redhat.com>
12263 PR rtl-optimizatin/81308
12264 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12265 (process_switch): If group_case_labels makes a change, then set
12267 (pass_convert_switch::execute): If a switch is converted, then
12268 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12270 PR rtl-optimization/81308
12271 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12274 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12276 PR target/83663 - Revert r255946
12277 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12278 generation for cases where splatting a value is not useful.
12279 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12280 across a vec_duplicate and a paradoxical subreg forming a vector
12281 mode to a vec_concat.
12283 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12285 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12286 -march=armv8.3-a variants.
12287 * config/arm/t-multilib: Likewise.
12288 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12290 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12292 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12294 (cceq_ior_compare_complement): Give it a name so I can use it, and
12295 change boolean_or_operator predicate to boolean_operator so it can
12296 be used to generate a crand.
12297 (eqne): New code iterator.
12298 (bd/bd_neg): New code_attrs.
12299 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12300 a single define_insn.
12301 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12302 decrement (bdnzt/bdnzf/bdzt/bdzf).
12303 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12304 with the new names of the branch decrement patterns, and added the
12305 names of the branch decrement conditional patterns.
12307 2018-01-08 Richard Biener <rguenther@suse.de>
12309 PR tree-optimization/83563
12310 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12313 2018-01-08 Richard Biener <rguenther@suse.de>
12315 PR middle-end/83713
12316 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12318 2018-01-08 Richard Biener <rguenther@suse.de>
12320 PR tree-optimization/83685
12321 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12322 references to abnormals.
12324 2018-01-08 Richard Biener <rguenther@suse.de>
12327 * dwarf2out.c (output_indirect_strings): Handle empty
12328 skeleton_debug_str_hash.
12329 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12331 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12333 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12334 (emit_store_direct): Likewise.
12335 (arc_trampoline_adjust_address): Likewise.
12336 (arc_asm_trampoline_template): New function.
12337 (arc_initialize_trampoline): Use asm_trampoline_template.
12338 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12339 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12340 * config/arc/arc.md (flush_icache): Delete pattern.
12342 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12344 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12345 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12348 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12351 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12352 by not USED_FOR_TARGET.
12353 (make_pass_resolve_sw_modes): Likewise.
12355 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12357 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12360 2018-01-08 Richard Biener <rguenther@suse.de>
12362 PR middle-end/83580
12363 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12365 2018-01-08 Richard Biener <rguenther@suse.de>
12367 PR middle-end/83517
12368 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12370 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12372 PR middle-end/81897
12373 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12374 basic blocks with a small number of successors.
12375 (convert_control_dep_chain_into_preds): Improve handling of
12377 (dump_predicates): Split apart into...
12378 (dump_pred_chain): ...here...
12379 (dump_pred_info): ...and here.
12380 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12381 (can_chain_union_be_invalidated_p): Improve check for invalidation
12383 (uninit_uses_cannot_happen): Avoid unnecessary if
12384 convert_control_dep_chain_into_preds yielded nothing.
12386 2018-01-06 Martin Sebor <msebor@redhat.com>
12388 PR tree-optimization/83640
12389 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12390 subtracting negative offset from size.
12391 (builtin_access::overlap): Adjust offset bounds of the access to fall
12392 within the size of the object if possible.
12394 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12396 PR rtl-optimization/83699
12397 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12398 extract_bit_field_as_subreg to cases in which the extracted
12399 value is also a vector.
12401 * lra-constraints.c (process_alt_operands): Test for the equivalence
12402 substitutions when detecting a possible reload cycle.
12404 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12407 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12408 by default if flag_selective_schedling{,2}. Formatting fixes.
12410 PR rtl-optimization/83682
12411 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12412 if it has non-VECTOR_MODE element mode.
12413 (vec_duplicate_p): Likewise.
12415 PR middle-end/83694
12416 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12417 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12419 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12422 * config/i386/i386-builtin.def
12423 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12424 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12425 Require also OPTION_MASK_ISA_AVX512F in addition to
12426 OPTION_MASK_ISA_GFNI.
12427 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12428 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12429 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12430 to OPTION_MASK_ISA_GFNI.
12431 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12432 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12433 OPTION_MASK_ISA_AVX512BW.
12434 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12435 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12436 addition to OPTION_MASK_ISA_GFNI.
12437 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12438 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12439 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12440 to OPTION_MASK_ISA_GFNI.
12441 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12442 a requirement for all ISAs rather than any of them with a few
12444 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12446 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12447 bitmasks to be enabled with 3 exceptions, instead of requiring any
12448 enabled ISA with lots of exceptions.
12449 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12450 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12451 Change avx512bw in isa attribute to avx512f.
12452 * config/i386/sgxintrin.h: Add license boilerplate.
12453 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12454 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12455 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12456 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12458 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12459 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12460 temporarily sse2 rather than sse if not enabled already.
12463 * config/i386/sse.md (VI248_VLBW): Rename to ...
12464 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12465 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12466 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12467 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12468 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12469 mode iterator instead of VI248_VLBW.
12471 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12473 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12474 (record_modified): Skip clobbers; add debug output.
12475 (param_change_prob): Use sreal frequencies.
12477 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12479 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12480 punt for user-aligned variables.
12482 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12484 * tree-chrec.c (chrec_contains_symbols): Return true for
12487 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12490 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12491 of (x|y) == x for BICS pattern.
12493 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12495 PR tree-optimization/83605
12496 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12497 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12500 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12502 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12503 * config/epiphany/rtems.h: New file.
12505 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12506 Uros Bizjak <ubizjak@gmail.com>
12509 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12510 QIreg_operand instead of register_operand predicate.
12511 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12512 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12513 comments instead of -fmitigate[-_]rop.
12515 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12518 * cgraphunit.c (symbol_table::compile): Switch to text_section
12519 before calling assembly_start debug hook.
12520 * run-rtl-passes.c (run_rtl_passes): Likewise.
12523 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12525 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12526 range_int_cst_p rather than !symbolic_range_p before calling
12527 extract_range_from_multiplicative_op_1.
12529 2018-01-04 Jeff Law <law@redhat.com>
12531 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12532 redundant test in assertion.
12534 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12536 * doc/rtl.texi: Document machine_mode wrapper classes.
12538 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12540 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12541 using tree_to_uhwi.
12543 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12545 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12546 the VEC_PERM_EXPR fold to fail.
12548 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12551 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12552 to switched_sections.
12554 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12557 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12558 test for d.testing.
12560 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12563 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12564 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12566 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12569 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12570 is BLKmode and bitpos not zero or mode change is needed.
12572 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12575 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12578 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12581 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12582 instead of MULT rtx. Update all corresponding splitters.
12583 (*saddl_se): Ditto.
12584 (*ssub<modesuffix>): Ditto.
12585 (*ssubl_se): Ditto.
12586 (*cmp_sadd_di): Update split patterns.
12587 (*cmp_sadd_si): Ditto.
12588 (*cmp_sadd_sidi): Ditto.
12589 (*cmp_ssub_di): Ditto.
12590 (*cmp_ssub_si): Ditto.
12591 (*cmp_ssub_sidi): Ditto.
12592 * config/alpha/predicates.md (const23_operand): New predicate.
12593 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12594 Look for ASHIFT, not MULT inner operand.
12595 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12597 2018-01-04 Martin Liska <mliska@suse.cz>
12599 PR gcov-profile/83669
12600 * gcov.c (output_intermediate_file): Add version to intermediate
12602 * doc/gcov.texi: Document new field 'version' in intermediate
12603 file format. Fix location of '-k' option of gcov command.
12605 2018-01-04 Martin Liska <mliska@suse.cz>
12608 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12610 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12612 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12614 2018-01-03 Martin Sebor <msebor@redhat.com>
12616 PR tree-optimization/83655
12617 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12618 checking calls with invalid arguments.
12620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12622 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12623 (vectorizable_mask_load_store): Delete.
12624 (vectorizable_call): Return false for masked loads and stores.
12625 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12626 instead of gimple_assign_rhs1.
12627 (vectorizable_load): Handle IFN_MASK_LOAD.
12628 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12632 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12634 (vectorizable_mask_load_store): ...here.
12635 (vectorizable_load): ...and here.
12637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12639 * tree-vect-stmts.c (vect_build_all_ones_mask)
12640 (vect_build_zero_merge_argument): New functions, split out from...
12641 (vectorizable_load): ...here.
12643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12645 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12647 (vectorizable_mask_load_store): ...here.
12648 (vectorizable_store): ...and here.
12650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12652 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12654 (vectorizable_mask_load_store): ...here.
12656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12658 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12659 (vect_model_store_cost): Take a vec_load_store_type instead of a
12661 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12662 (vect_model_store_cost): Take a vec_load_store_type instead of a
12664 (vectorizable_mask_load_store): Update accordingly.
12665 (vectorizable_store): Likewise.
12666 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12670 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12671 IFN_MASK_LOAD calls here rather than...
12672 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12674 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12675 Alan Hayward <alan.hayward@arm.com>
12676 David Sherwood <david.sherwood@arm.com>
12678 * expmed.c (extract_bit_field_1): For vector extracts,
12679 fall back to extract_bit_field_as_subreg if vec_extract
12682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12683 Alan Hayward <alan.hayward@arm.com>
12684 David Sherwood <david.sherwood@arm.com>
12686 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12687 they are variable or constant sized.
12688 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12689 slots for constant-sized data.
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 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12696 handling COND_EXPRs with boolean comparisons, try to find a better
12697 basis for the mask type than the boolean itself.
12699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12701 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12702 is calculated and how it can be overridden.
12703 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12704 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12706 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12710 Alan Hayward <alan.hayward@arm.com>
12711 David Sherwood <david.sherwood@arm.com>
12713 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12714 Remove the mode argument.
12715 (aarch64_simd_valid_immediate): Remove the mode and inverse
12717 * config/aarch64/iterators.md (bitsize): New iterator.
12718 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12719 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12720 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12721 aarch64_simd_valid_immediate.
12722 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12723 (aarch64_reg_or_bic_imm): Likewise.
12724 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12725 with an insn_type enum and msl with a modifier_type enum.
12726 Replace element_width with a scalar_mode. Change the shift
12727 to unsigned int. Add constructors for scalar_float_mode and
12728 scalar_int_mode elements.
12729 (aarch64_vect_float_const_representable_p): Delete.
12730 (aarch64_can_const_movi_rtx_p)
12731 (aarch64_simd_scalar_immediate_valid_for_move)
12732 (aarch64_simd_make_constant): Update call to
12733 aarch64_simd_valid_immediate.
12734 (aarch64_advsimd_valid_immediate_hs): New function.
12735 (aarch64_advsimd_valid_immediate): Likewise.
12736 (aarch64_simd_valid_immediate): Remove mode and inverse
12737 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12738 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12739 and aarch64_float_const_representable_p on the result.
12740 (aarch64_output_simd_mov_immediate): Remove mode argument.
12741 Update call to aarch64_simd_valid_immediate and use of
12742 simd_immediate_info.
12743 (aarch64_output_scalar_simd_mov_immediate): Update call
12746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12747 Alan Hayward <alan.hayward@arm.com>
12748 David Sherwood <david.sherwood@arm.com>
12750 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12751 (mode_nunits): Likewise CONST_MODE_NUNITS.
12752 * machmode.def (ADJUST_NUNITS): Document.
12753 * genmodes.c (mode_data::need_nunits_adj): New field.
12754 (blank_mode): Update accordingly.
12755 (adj_nunits): New variable.
12756 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12758 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12759 listed in adj_nunits.
12760 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12761 listed in adj_nunits. Don't emit case statements for such modes.
12762 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12763 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12764 nothing if adj_nunits is nonnull.
12765 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12766 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12767 (emit_mode_fbit): Update use of print_maybe_const_decl.
12768 (emit_move_size): Likewise. Treat the array as non-const
12770 (emit_mode_adjustments): Handle adj_nunits.
12772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12774 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12775 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12776 (VECTOR_MODES): Use it.
12777 (make_vector_modes): Take the prefix as an argument.
12779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12780 Alan Hayward <alan.hayward@arm.com>
12781 David Sherwood <david.sherwood@arm.com>
12783 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12784 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12785 for MODE_VECTOR_BOOL.
12786 * machmode.def (VECTOR_BOOL_MODE): Document.
12787 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12788 (make_vector_bool_mode): New function.
12789 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12791 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12792 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12794 * stor-layout.c (int_mode_for_mode): Likewise.
12795 * tree.c (build_vector_type_for_mode): Likewise.
12796 * varasm.c (output_constant_pool_2): Likewise.
12797 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12798 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12799 for MODE_VECTOR_BOOL.
12800 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12801 of mode class checks.
12802 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12803 instead of a list of mode class checks.
12804 (expand_vector_scalar_condition): Likewise.
12805 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12808 Alan Hayward <alan.hayward@arm.com>
12809 David Sherwood <david.sherwood@arm.com>
12811 * machmode.h (mode_size): Change from unsigned short to
12813 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12814 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12815 or if measurement_type is not polynomial.
12816 (fixed_size_mode::includes_p): Check for constant-sized modes.
12817 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12818 return a poly_uint16 rather than an unsigned short.
12819 (emit_mode_size): Change the type of mode_size from unsigned short
12820 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12821 (emit_mode_adjustments): Cope with polynomial vector sizes.
12822 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12824 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12826 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12827 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12828 * caller-save.c (setup_save_areas): Likewise.
12829 (replace_reg_with_saved_mem): Likewise.
12830 * calls.c (emit_library_call_value_1): Likewise.
12831 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12832 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12833 (gen_lowpart_for_combine): Likewise.
12834 * convert.c (convert_to_integer_1): Likewise.
12835 * cse.c (equiv_constant, cse_insn): Likewise.
12836 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12837 (cselib_subst_to_values): Likewise.
12838 * dce.c (word_dce_process_block): Likewise.
12839 * df-problems.c (df_word_lr_mark_ref): Likewise.
12840 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12841 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12842 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12843 (rtl_for_decl_location): Likewise.
12844 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12845 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12846 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12847 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12848 (expand_expr_real_1): Likewise.
12849 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12850 (pad_below): Likewise.
12851 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12852 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12853 * ira.c (get_subreg_tracking_sizes): Likewise.
12854 * ira-build.c (ira_create_allocno_objects): Likewise.
12855 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12856 (ira_sort_regnos_for_alter_reg): Likewise.
12857 * ira-costs.c (record_operand_costs): Likewise.
12858 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12859 (resolve_simple_move): Likewise.
12860 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12861 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12862 (lra_constraints): Likewise.
12863 (CONST_POOL_OK_P): Reject variable-sized modes.
12864 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12865 (add_pseudo_to_slot, lra_spill): Likewise.
12866 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12867 * optabs-query.c (get_best_extraction_insn): Likewise.
12868 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12869 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12870 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12871 * recog.c (offsettable_address_addr_space_p): Likewise.
12872 * regcprop.c (maybe_mode_change): Likewise.
12873 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12874 * regrename.c (build_def_use): Likewise.
12875 * regstat.c (dump_reg_info): Likewise.
12876 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12877 (find_reloads, find_reloads_subreg_address): Likewise.
12878 * reload1.c (eliminate_regs_1): Likewise.
12879 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12880 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12881 (simplify_binary_operation_1, simplify_subreg): Likewise.
12882 * targhooks.c (default_function_arg_padding): Likewise.
12883 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12884 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12885 (verify_gimple_assign_ternary): Likewise.
12886 * tree-inline.c (estimate_move_cost): Likewise.
12887 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12888 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12889 (get_address_cost_ainc): Likewise.
12890 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12891 (vect_supportable_dr_alignment): Likewise.
12892 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12893 (vectorizable_reduction): Likewise.
12894 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12895 (vectorizable_operation, vectorizable_load): Likewise.
12896 * tree.c (build_same_sized_truth_vector_type): Likewise.
12897 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12898 * var-tracking.c (emit_note_insn_var_location): Likewise.
12899 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12900 (ADDR_VEC_ALIGN): Likewise.
12902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12903 Alan Hayward <alan.hayward@arm.com>
12904 David Sherwood <david.sherwood@arm.com>
12906 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12908 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12909 or if measurement_type is polynomial.
12910 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12911 * combine.c (make_extraction): Likewise.
12912 * dse.c (find_shift_sequence): Likewise.
12913 * dwarf2out.c (mem_loc_descriptor): Likewise.
12914 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12915 (extract_bit_field, extract_low_bits): Likewise.
12916 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12917 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12918 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12919 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12920 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12921 * reload.c (find_reloads): Likewise.
12922 * reload1.c (alter_reg): Likewise.
12923 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12924 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12925 * tree-if-conv.c (predicate_mem_writes): Likewise.
12926 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12927 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12928 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12929 * valtrack.c (dead_debug_insert_temp): Likewise.
12930 * varasm.c (mergeable_constant_section): Likewise.
12931 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12934 Alan Hayward <alan.hayward@arm.com>
12935 David Sherwood <david.sherwood@arm.com>
12937 * expr.c (expand_assignment): Cope with polynomial mode sizes
12938 when assigning to a CONCAT.
12940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12941 Alan Hayward <alan.hayward@arm.com>
12942 David Sherwood <david.sherwood@arm.com>
12944 * machmode.h (mode_precision): Change from unsigned short to
12946 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12948 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12949 or if measurement_type is not polynomial.
12950 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12951 in which the mode is already known to be a scalar_int_mode.
12952 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12953 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12955 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12956 for GET_MODE_PRECISION.
12957 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12958 for GET_MODE_PRECISION.
12959 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12961 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12962 (expand_field_assignment, make_extraction): Likewise.
12963 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12964 (get_last_value): Likewise.
12965 * convert.c (convert_to_integer_1): Likewise.
12966 * cse.c (cse_insn): Likewise.
12967 * expr.c (expand_expr_real_1): Likewise.
12968 * lra-constraints.c (simplify_operand_subreg): Likewise.
12969 * optabs-query.c (can_atomic_load_p): Likewise.
12970 * optabs.c (expand_atomic_load): Likewise.
12971 (expand_atomic_store): Likewise.
12972 * ree.c (combine_reaching_defs): Likewise.
12973 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12974 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12975 * tree.h (type_has_mode_precision_p): Likewise.
12976 * ubsan.c (instrument_si_overflow): Likewise.
12978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12979 Alan Hayward <alan.hayward@arm.com>
12980 David Sherwood <david.sherwood@arm.com>
12982 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12983 polynomial numbers of units.
12984 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12985 (valid_vector_subparts_p): New function.
12986 (build_vector_type): Remove temporary shim and take the number
12987 of units as a poly_uint64 rather than an int.
12988 (build_opaque_vector_type): Take the number of units as a
12989 poly_uint64 rather than an int.
12990 * tree.c (build_vector_from_ctor): Handle polynomial
12991 TYPE_VECTOR_SUBPARTS.
12992 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12993 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12994 (build_vector_from_val): If the number of units is variable,
12995 use build_vec_duplicate_cst for constant operands and
12996 VEC_DUPLICATE_EXPR otherwise.
12997 (make_vector_type): Remove temporary is_constant ().
12998 (build_vector_type, build_opaque_vector_type): Take the number of
12999 units as a poly_uint64 rather than an int.
13000 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13002 * cfgexpand.c (expand_debug_expr): Likewise.
13003 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13004 (store_constructor, expand_expr_real_1): Likewise.
13005 (const_scalar_mask_from_tree): Likewise.
13006 * fold-const-call.c (fold_const_reduction): Likewise.
13007 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13008 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13009 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13010 (fold_relational_const): Likewise.
13011 (native_interpret_vector): Likewise. Change the size from an
13012 int to an unsigned int.
13013 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13014 TYPE_VECTOR_SUBPARTS.
13015 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13016 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13017 duplicating a non-constant operand into a variable-length vector.
13018 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13019 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13020 * ipa-icf.c (sem_variable::equals): Likewise.
13021 * match.pd: Likewise.
13022 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13023 * print-tree.c (print_node): Likewise.
13024 * stor-layout.c (layout_type): Likewise.
13025 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13026 * tree-cfg.c (verify_gimple_comparison): Likewise.
13027 (verify_gimple_assign_binary): Likewise.
13028 (verify_gimple_assign_ternary): Likewise.
13029 (verify_gimple_assign_single): Likewise.
13030 * tree-pretty-print.c (dump_generic_node): Likewise.
13031 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13032 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13033 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13034 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13035 (vect_shift_permute_load_chain): Likewise.
13036 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13037 (expand_vector_condition, optimize_vector_constructor): Likewise.
13038 (lower_vec_perm, get_compute_type): Likewise.
13039 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13040 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13041 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13042 (vect_recog_mask_conversion_pattern): Likewise.
13043 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13044 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13045 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13046 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13047 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13048 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13049 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13050 (supportable_widening_operation): Likewise.
13051 (supportable_narrowing_operation): Likewise.
13052 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13054 * varasm.c (output_constant): Likewise.
13056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13057 Alan Hayward <alan.hayward@arm.com>
13058 David Sherwood <david.sherwood@arm.com>
13060 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13061 so that both the length == 3 and length != 3 cases set up their
13062 own permute vectors. Add comments explaining why we know the
13063 number of elements is constant.
13064 (vect_permute_load_chain): Likewise.
13066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13067 Alan Hayward <alan.hayward@arm.com>
13068 David Sherwood <david.sherwood@arm.com>
13070 * machmode.h (mode_nunits): Change from unsigned char to
13072 (ONLY_FIXED_SIZE_MODES): New macro.
13073 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13074 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13075 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13077 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13078 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13079 or if measurement_type is not polynomial.
13080 * genmodes.c (ZERO_COEFFS): New macro.
13081 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13083 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13084 Use ZERO_COEFFS when emitting initializers.
13085 * data-streamer.h (bp_pack_poly_value): New function.
13086 (bp_unpack_poly_value): Likewise.
13087 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13088 for GET_MODE_NUNITS.
13089 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13090 for GET_MODE_NUNITS.
13091 * tree.c (make_vector_type): Remove temporary shim and make
13092 the real function take the number of units as a poly_uint64
13093 rather than an int.
13094 (build_vector_type_for_mode): Handle polynomial nunits.
13095 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13096 * emit-rtl.c (const_vec_series_p_1): Likewise.
13097 (gen_rtx_CONST_VECTOR): Likewise.
13098 * fold-const.c (test_vec_duplicate_folding): Likewise.
13099 * genrecog.c (validate_pattern): Likewise.
13100 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13101 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13102 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13103 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13104 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13105 * rtlanal.c (subreg_get_info): Likewise.
13106 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13107 (vect_grouped_load_supported): Likewise.
13108 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13109 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13110 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13111 (simplify_const_unary_operation, simplify_binary_operation_1)
13112 (simplify_const_binary_operation, simplify_ternary_operation)
13113 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13114 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13115 instead of CONST_VECTOR_NUNITS.
13116 * varasm.c (output_constant_pool_2): Likewise.
13117 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13118 explicit-encoded elements in the XVEC for variable-length vectors.
13120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13122 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13125 Alan Hayward <alan.hayward@arm.com>
13126 David Sherwood <david.sherwood@arm.com>
13128 * coretypes.h (fixed_size_mode): Declare.
13129 (fixed_size_mode_pod): New typedef.
13130 * builtins.h (target_builtins::x_apply_args_mode)
13131 (target_builtins::x_apply_result_mode): Change type to
13132 fixed_size_mode_pod.
13133 * builtins.c (apply_args_size, apply_result_size, result_vector)
13134 (expand_builtin_apply_args_1, expand_builtin_apply)
13135 (expand_builtin_return): Update accordingly.
13137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13139 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13140 * cselib.c (cselib_hash_rtx): Likewise.
13141 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13142 CONST_VECTOR encoding.
13144 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13145 Jeff Law <law@redhat.com>
13148 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13149 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13150 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13151 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13154 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13155 explicitly probe *sp in a noreturn function if there were any callee
13156 register saves or frame pointer is needed.
13158 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13161 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13162 BLKmode for ternary, binary or unary expressions.
13165 * var-tracking.c (delete_vta_debug_insn): New inline function.
13166 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13167 insns from get_insns () to NULL instead of each bb separately.
13168 Use delete_vta_debug_insn. No longer static.
13169 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13170 delete_vta_debug_insns callers.
13171 * rtl.h (delete_vta_debug_insns): Declare.
13172 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13173 instead of variable_tracking_main.
13175 2018-01-03 Martin Sebor <msebor@redhat.com>
13177 PR tree-optimization/83603
13178 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13179 arguments past the endof the argument list in functions declared
13180 without a prototype.
13181 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13182 Avoid checking when arguments are null.
13184 2018-01-03 Martin Sebor <msebor@redhat.com>
13187 * doc/extend.texi (attribute const): Fix a typo.
13188 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13189 issuing -Wsuggest-attribute for void functions.
13191 2018-01-03 Martin Sebor <msebor@redhat.com>
13193 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13194 offset_int::from instead of wide_int::to_shwi.
13195 (maybe_diag_overlap): Remove assertion.
13196 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13197 * gimple-ssa-sprintf.c (format_directive): Same.
13198 (parse_directive): Same.
13199 (sprintf_dom_walker::compute_format_length): Same.
13200 (try_substitute_return_value): Same.
13202 2018-01-03 Jeff Law <law@redhat.com>
13204 PR middle-end/83654
13205 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13206 non-constant residual for zero at runtime and avoid probing in
13207 that case. Reorganize code for trailing problem to mirror handling
13210 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13212 PR tree-optimization/83501
13213 * tree-ssa-strlen.c (get_string_cst): New.
13214 (handle_char_store): Call get_string_cst.
13216 2018-01-03 Martin Liska <mliska@suse.cz>
13218 PR tree-optimization/83593
13219 * tree-ssa-strlen.c: Include tree-cfg.h.
13220 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13221 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13222 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13224 (strlen_dom_walker::before_dom_children): Call
13225 gimple_purge_dead_eh_edges. Dump tranformation with details
13227 (strlen_dom_walker::before_dom_children): Update call by adding
13228 new argument cleanup_eh.
13229 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13231 2018-01-03 Martin Liska <mliska@suse.cz>
13234 * cif-code.def (VARIADIC_THUNK): New enum value.
13235 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13238 2018-01-03 Jan Beulich <jbeulich@suse.com>
13240 * sse.md (mov<mode>_internal): Tighten condition for when to use
13241 vmovdqu<ssescalarsize> for TI and OI modes.
13243 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13245 Update copyright years.
13247 2018-01-03 Martin Liska <mliska@suse.cz>
13250 * ipa-visibility.c (function_and_variable_visibility): Skip
13251 functions with noipa attribure.
13253 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13255 * gcc.c (process_command): Update copyright notice dates.
13256 * gcov-dump.c (print_version): Ditto.
13257 * gcov.c (print_version): Ditto.
13258 * gcov-tool.c (print_version): Ditto.
13259 * gengtype.c (create_file): Ditto.
13260 * doc/cpp.texi: Bump @copying's copyright year.
13261 * doc/cppinternals.texi: Ditto.
13262 * doc/gcc.texi: Ditto.
13263 * doc/gccint.texi: Ditto.
13264 * doc/gcov.texi: Ditto.
13265 * doc/install.texi: Ditto.
13266 * doc/invoke.texi: Ditto.
13268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13270 * vector-builder.h (vector_builder::m_full_nelts): Change from
13271 unsigned int to poly_uint64.
13272 (vector_builder::full_nelts): Update prototype accordingly.
13273 (vector_builder::new_vector): Likewise.
13274 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13275 (vector_builder::operator ==): Likewise.
13276 (vector_builder::finalize): Likewise.
13277 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13278 Take the number of elements as a poly_uint64 rather than an
13280 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13281 from unsigned int to poly_uint64.
13282 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13283 (vec_perm_indices::new_vector): Likewise.
13284 (vec_perm_indices::length): Likewise.
13285 (vec_perm_indices::nelts_per_input): Likewise.
13286 (vec_perm_indices::input_nelts): Likewise.
13287 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13288 number of elements per input as a poly_uint64 rather than an
13289 unsigned int. Use the original encoding for variable-length
13290 vectors, rather than clamping each individual element.
13291 For the second and subsequent elements in each pattern,
13292 clamp the step and base before clamping their sum.
13293 (vec_perm_indices::series_p): Handle polynomial element counts.
13294 (vec_perm_indices::all_in_range_p): Likewise.
13295 (vec_perm_indices_to_tree): Likewise.
13296 (vec_perm_indices_to_rtx): Likewise.
13297 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13298 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13299 (tree_vector_builder::new_binary_operation): Handle polynomial
13300 element counts. Return false if we need to know the number
13301 of elements at compile time.
13302 * fold-const.c (fold_vec_perm): Punt if the number of elements
13303 isn't known at compile time.
13305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13307 * vec-perm-indices.h (vec_perm_builder): Change element type
13308 from HOST_WIDE_INT to poly_int64.
13309 (vec_perm_indices::element_type): Update accordingly.
13310 (vec_perm_indices::clamp): Handle polynomial element_types.
13311 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13312 (vec_perm_indices::all_in_range_p): Likewise.
13313 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13315 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13316 polynomial vec_perm_indices element types.
13317 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13318 * fold-const.c (fold_vec_perm): Likewise.
13319 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13320 * tree-vect-generic.c (lower_vec_perm): Likewise.
13321 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13322 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13323 element type to HOST_WIDE_INT.
13325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13326 Alan Hayward <alan.hayward@arm.com>
13327 David Sherwood <david.sherwood@arm.com>
13329 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13330 rather than an int. Use plus_constant.
13331 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13332 Take the offset "c" as a poly_int64 rather than a 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 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13339 a HOST_WIDE_INT to a poly_int64.
13341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13342 Alan Hayward <alan.hayward@arm.com>
13343 David Sherwood <david.sherwood@arm.com>
13345 * calls.c (load_register_parameters): Cope with polynomial
13346 mode sizes. Require a constant size for BLKmode parameters
13347 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13348 forces a parameter to be padded at the lsb end in order to
13349 fill a complete number of words, require the parameter size
13350 to be ordered wrt UNITS_PER_WORD.
13352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13353 Alan Hayward <alan.hayward@arm.com>
13354 David Sherwood <david.sherwood@arm.com>
13356 * reload1.c (spill_stack_slot_width): Change element type
13357 from unsigned int to poly_uint64_pod.
13358 (alter_reg): Treat mode sizes as polynomial.
13360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13361 Alan Hayward <alan.hayward@arm.com>
13362 David Sherwood <david.sherwood@arm.com>
13364 * reload.c (complex_word_subreg_p): New function.
13365 (reload_inner_reg_of_subreg, push_reload): Use it.
13367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13368 Alan Hayward <alan.hayward@arm.com>
13369 David Sherwood <david.sherwood@arm.com>
13371 * lra-constraints.c (process_alt_operands): Reject matched
13372 operands whose sizes aren't ordered.
13373 (match_reload): Refer to this check here.
13375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13376 Alan Hayward <alan.hayward@arm.com>
13377 David Sherwood <david.sherwood@arm.com>
13379 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13380 that the mode size is in the set {1, 2, 4, 8, 16}.
13382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13383 Alan Hayward <alan.hayward@arm.com>
13384 David Sherwood <david.sherwood@arm.com>
13386 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13387 Use plus_constant instead of gen_rtx_PLUS.
13389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13390 Alan Hayward <alan.hayward@arm.com>
13391 David Sherwood <david.sherwood@arm.com>
13393 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13394 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13395 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13396 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13397 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13398 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13399 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13400 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13401 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13402 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13404 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13405 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13406 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13407 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13408 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13409 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13410 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13411 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13412 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13413 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13415 * expr.c (emit_move_resolve_push): Treat the input and result
13416 of PUSH_ROUNDING as a poly_int64.
13417 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13418 (emit_push_insn): Likewise.
13419 * lra-eliminations.c (mark_not_eliminable): Likewise.
13420 * recog.c (push_operand): Likewise.
13421 * reload1.c (elimination_effects): Likewise.
13422 * rtlanal.c (nonzero_bits1): Likewise.
13423 * calls.c (store_one_arg): Likewise. Require the padding to be
13424 known at compile time.
13426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13427 Alan Hayward <alan.hayward@arm.com>
13428 David Sherwood <david.sherwood@arm.com>
13430 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13431 Use plus_constant instead of gen_rtx_PLUS.
13433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13434 Alan Hayward <alan.hayward@arm.com>
13435 David Sherwood <david.sherwood@arm.com>
13437 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13438 rather than an int.
13440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13441 Alan Hayward <alan.hayward@arm.com>
13442 David Sherwood <david.sherwood@arm.com>
13444 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13445 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13446 via stack temporaries. Treat the mode size as polynomial too.
13448 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13449 Alan Hayward <alan.hayward@arm.com>
13450 David Sherwood <david.sherwood@arm.com>
13452 * expr.c (expand_expr_real_2): When handling conversions involving
13453 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13454 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13455 as a poly_uint64 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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13464 Alan Hayward <alan.hayward@arm.com>
13465 David Sherwood <david.sherwood@arm.com>
13467 * combine.c (can_change_dest_mode): Handle polynomial
13468 REGMODE_NATURAL_SIZE.
13469 * expmed.c (store_bit_field_1): Likewise.
13470 * expr.c (store_constructor): Likewise.
13471 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13472 and polynomial REGMODE_NATURAL_SIZE.
13473 (gen_lowpart_common): Likewise.
13474 * reginfo.c (record_subregs_of_mode): Likewise.
13475 * rtlanal.c (read_modify_subreg_p): Likewise.
13477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13478 Alan Hayward <alan.hayward@arm.com>
13479 David Sherwood <david.sherwood@arm.com>
13481 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13482 numbers of elements.
13484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13485 Alan Hayward <alan.hayward@arm.com>
13486 David Sherwood <david.sherwood@arm.com>
13488 * match.pd: Cope with polynomial numbers of vector elements.
13490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13491 Alan Hayward <alan.hayward@arm.com>
13492 David Sherwood <david.sherwood@arm.com>
13494 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13495 in a POINTER_PLUS_EXPR.
13497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13498 Alan Hayward <alan.hayward@arm.com>
13499 David Sherwood <david.sherwood@arm.com>
13501 * omp-simd-clone.c (simd_clone_subparts): New function.
13502 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13503 (ipa_simd_modify_function_body): Likewise.
13505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13506 Alan Hayward <alan.hayward@arm.com>
13507 David Sherwood <david.sherwood@arm.com>
13509 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13510 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13511 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13512 (expand_vector_condition, vector_element): Likewise.
13513 (subparts_gt): New function.
13514 (get_compute_type): Use subparts_gt.
13515 (count_type_subparts): Delete.
13516 (expand_vector_operations_1): Use subparts_gt instead of
13517 count_type_subparts.
13519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13520 Alan Hayward <alan.hayward@arm.com>
13521 David Sherwood <david.sherwood@arm.com>
13523 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13524 (vect_compile_time_alias): ...this new function. Do the calculation
13525 on poly_ints rather than trees.
13526 (vect_prune_runtime_alias_test_list): Update call accordingly.
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-slp.c (vect_build_slp_tree_1): Handle polynomial
13534 (vect_schedule_slp_instance): Likewise.
13536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13537 Alan Hayward <alan.hayward@arm.com>
13538 David Sherwood <david.sherwood@arm.com>
13540 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13541 constant and extern definitions for variable-length vectors.
13542 (vect_get_constant_vectors): Note that the number of units
13543 is known to be constant.
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-stmts.c (vectorizable_conversion): Treat the number
13550 of units as polynomial. Choose between WIDE and NARROW based
13553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13554 Alan Hayward <alan.hayward@arm.com>
13555 David Sherwood <david.sherwood@arm.com>
13557 * tree-vect-stmts.c (simd_clone_subparts): New function.
13558 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13561 Alan Hayward <alan.hayward@arm.com>
13562 David Sherwood <david.sherwood@arm.com>
13564 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13565 vectors as polynomial. Use build_index_vector for
13566 IFN_GOMP_SIMD_LANE.
13568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13569 Alan Hayward <alan.hayward@arm.com>
13570 David Sherwood <david.sherwood@arm.com>
13572 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13573 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13574 for variable-length vectors.
13575 (vectorizable_mask_load_store): Treat the number of units as
13576 polynomial, asserting that it is constant if the condition has
13577 already been enforced.
13578 (vectorizable_store, vectorizable_load): Likewise.
13580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13581 Alan Hayward <alan.hayward@arm.com>
13582 David Sherwood <david.sherwood@arm.com>
13584 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13585 of units as polynomial. Punt if we can't tell at compile time
13586 which vector contains the final result.
13588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13589 Alan Hayward <alan.hayward@arm.com>
13590 David Sherwood <david.sherwood@arm.com>
13592 * tree-vect-loop.c (vectorizable_induction): Treat the number
13593 of units as polynomial. Punt on SLP inductions. Use an integer
13594 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13595 cast of such a series for variable-length floating-point
13598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13599 Alan Hayward <alan.hayward@arm.com>
13600 David Sherwood <david.sherwood@arm.com>
13602 * tree.h (build_index_vector): Declare.
13603 * tree.c (build_index_vector): New function.
13604 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13605 of units as polynomial, forcibly converting it to a constant if
13606 vectorizable_reduction has already enforced the condition.
13607 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13608 to create a {1,2,3,...} vector.
13609 (vectorizable_reduction): Treat the number of units as polynomial.
13610 Choose vectype_in based on the largest scalar element size rather
13611 than the smallest number of units. Enforce the restrictions
13614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13615 Alan Hayward <alan.hayward@arm.com>
13616 David Sherwood <david.sherwood@arm.com>
13618 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13619 number of units as polynomial.
13621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13622 Alan Hayward <alan.hayward@arm.com>
13623 David Sherwood <david.sherwood@arm.com>
13625 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13626 * target.def (autovectorize_vector_sizes): Return the vector sizes
13627 by pointer, using vector_sizes rather than a bitmask.
13628 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13629 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13630 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13632 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13633 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13634 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13635 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13636 * omp-general.c (omp_max_vf): Likewise.
13637 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13638 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13639 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13640 * tree-vect-slp.c (vect_slp_bb): Likewise.
13641 * doc/tm.texi: Regenerate.
13642 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13644 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13645 the vector size as a poly_uint64 rather than an unsigned int.
13646 (current_vector_size): Change from an unsigned int to a poly_uint64.
13647 (get_vectype_for_scalar_type): Update accordingly.
13648 * tree.h (build_truth_vector_type): Take the size and number of
13649 units as a poly_uint64 rather than an unsigned int.
13650 (build_vector_type): Add a temporary overload that takes
13651 the number of units as a poly_uint64 rather than an unsigned int.
13652 * tree.c (make_vector_type): Likewise.
13653 (build_truth_vector_type): Take the number of units as a poly_uint64
13654 rather than an unsigned int.
13656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13657 Alan Hayward <alan.hayward@arm.com>
13658 David Sherwood <david.sherwood@arm.com>
13660 * target.def (get_mask_mode): Take the number of units and length
13661 as poly_uint64s rather than unsigned ints.
13662 * targhooks.h (default_get_mask_mode): Update accordingly.
13663 * targhooks.c (default_get_mask_mode): Likewise.
13664 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13665 * doc/tm.texi: Regenerate.
13667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13668 Alan Hayward <alan.hayward@arm.com>
13669 David Sherwood <david.sherwood@arm.com>
13671 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13672 * omp-general.c (omp_max_vf): Likewise.
13673 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13674 (expand_omp_simd): Handle polynomial safelen.
13675 * omp-low.c (omplow_simd_context): Add a default constructor.
13676 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13677 (lower_rec_simd_input_clauses): Update accordingly.
13678 (lower_rec_input_clauses): Likewise.
13680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13681 Alan Hayward <alan.hayward@arm.com>
13682 David Sherwood <david.sherwood@arm.com>
13684 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13685 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13686 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13687 (vect_analyze_slp_cost): Likewise.
13688 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13689 (vect_model_load_cost): Likewise.
13691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13692 Alan Hayward <alan.hayward@arm.com>
13693 David Sherwood <david.sherwood@arm.com>
13695 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13696 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13697 from an unsigned int * to a poly_uint64_pod *.
13698 (calculate_unrolling_factor): New function.
13699 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13702 Alan Hayward <alan.hayward@arm.com>
13703 David Sherwood <david.sherwood@arm.com>
13705 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13706 from an unsigned int to a poly_uint64.
13707 (_loop_vec_info::slp_unrolling_factor): Likewise.
13708 (_loop_vec_info::vectorization_factor): Change from an int
13710 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13711 (vect_get_num_vectors): New function.
13712 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13713 (vect_get_num_copies): Use vect_get_num_vectors.
13714 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13715 to an unsigned int *.
13716 (vect_analyze_data_refs): Change min_vf from an int * to a
13718 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13719 than an unsigned HOST_WIDE_INT.
13720 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13721 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13722 to an unsigned int *.
13723 (vect_analyze_data_ref_dependences): Likewise.
13724 (vect_compute_data_ref_alignment): Handle polynomial vf.
13725 (vect_enhance_data_refs_alignment): Likewise.
13726 (vect_prune_runtime_alias_test_list): Likewise.
13727 (vect_shift_permute_load_chain): Likewise.
13728 (vect_supportable_dr_alignment): Likewise.
13729 (dependence_distance_ge_vf): Take the vectorization factor as a
13730 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13731 (vect_analyze_data_refs): Change min_vf from an int * to a
13733 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13734 vfm1 as a poly_uint64 rather than an int. Make the same change
13735 for the returned bound_scalar.
13736 (vect_gen_vector_loop_niters): Handle polynomial vf.
13737 (vect_do_peeling): Likewise. Update call to
13738 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13739 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13741 * tree-vect-loop.c (vect_determine_vectorization_factor)
13742 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13743 (vect_get_known_peeling_cost): Likewise.
13744 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13745 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13746 (vect_transform_loop): Likewise. Use the lowest possible VF when
13747 updating the upper bounds of the loop.
13748 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13749 rather than an int.
13750 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13751 polynomial unroll factors.
13752 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13753 (vect_make_slp_decision): Likewise.
13754 (vect_supported_load_permutation_p): Likewise, and polynomial
13756 (vect_analyze_slp_cost): Handle polynomial vf.
13757 (vect_slp_analyze_node_operations): Likewise.
13758 (vect_slp_analyze_bb_1): Likewise.
13759 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13760 than an unsigned HOST_WIDE_INT.
13761 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13762 (vectorizable_load): Handle polynomial vf.
13763 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13765 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13768 Alan Hayward <alan.hayward@arm.com>
13769 David Sherwood <david.sherwood@arm.com>
13771 * match.pd: Handle bit operations involving three constants
13772 and try to fold one pair.
13774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13776 * tree-vect-loop-manip.c: Include gimple-fold.h.
13777 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13778 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13779 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13780 Add a path that uses a step of VF instead of 1, but disable it
13782 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13783 and niters_no_overflow parameters. Update calls to
13784 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13785 Create a new SSA name if the latter choses to use a ste other
13786 than zero, and return it via niters_vector_mult_vf_var.
13787 * tree-vect-loop.c (vect_transform_loop): Update calls to
13788 vect_do_peeling, vect_gen_vector_loop_niters and
13789 slpeel_make_loop_iterate_ntimes.
13790 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13791 (vect_gen_vector_loop_niters): Update declarations after above changes.
13793 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13795 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13796 128-bit round to integer instructions.
13797 (ceil<mode>2): Likewise.
13798 (btrunc<mode>2): Likewise.
13799 (round<mode>2): Likewise.
13801 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13803 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13804 unaligned VSX load/store on P8/P9.
13805 (expand_block_clear): Allow the use of unaligned VSX
13806 load/store on P8/P9.
13808 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13810 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13812 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13813 swap associated with both a load and a store.
13815 2018-01-02 Andrew Waterman <andrew@sifive.com>
13817 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13818 * config/riscv/riscv.md (clear_cache): Use it.
13820 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13822 * web.c: Remove out-of-date comment.
13824 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13826 * expr.c (fixup_args_size_notes): Check that any existing
13827 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13828 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13829 (emit_single_push_insn): ...here.
13831 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13833 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13834 (const_vector_encoded_nelts): New function.
13835 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13836 (const_vector_int_elt, const_vector_elt): Declare.
13837 * emit-rtl.c (const_vector_int_elt_1): New function.
13838 (const_vector_elt): Likewise.
13839 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13840 of CONST_VECTOR_ELT.
13842 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13844 * expr.c: Include rtx-vector-builder.h.
13845 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13846 directly on the tree encoding.
13847 (const_vector_from_tree): Likewise.
13848 * optabs.c: Include rtx-vector-builder.h.
13849 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13850 sequence of "u" values.
13851 * vec-perm-indices.c: Include rtx-vector-builder.h.
13852 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13853 directly on the vec_perm_indices encoding.
13855 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13857 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13858 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13859 * rtx-vector-builder.h: New file.
13860 * rtx-vector-builder.c: Likewise.
13861 * rtl.h (rtx_def::u2): Add a const_vector field.
13862 (CONST_VECTOR_NPATTERNS): New macro.
13863 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13864 (CONST_VECTOR_DUPLICATE_P): Likewise.
13865 (CONST_VECTOR_STEPPED_P): Likewise.
13866 (CONST_VECTOR_ENCODED_ELT): Likewise.
13867 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13868 (unwrap_const_vec_duplicate): Likewise.
13869 (const_vec_series_p): Check for a non-duplicated vector encoding.
13870 Say that the function only returns true for integer vectors.
13871 * emit-rtl.c: Include rtx-vector-builder.h.
13872 (gen_const_vec_duplicate_1): Delete.
13873 (gen_const_vector): Call gen_const_vec_duplicate instead of
13874 gen_const_vec_duplicate_1.
13875 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13876 (gen_const_vec_duplicate): Use rtx_vector_builder.
13877 (gen_const_vec_series): Likewise.
13878 (gen_rtx_CONST_VECTOR): Likewise.
13879 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13880 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13881 Build a new vector rather than modifying a CONST_VECTOR in-place.
13882 (handle_special_swappables): Update call accordingly.
13883 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13884 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13885 Build a new vector rather than modifying a CONST_VECTOR in-place.
13886 (handle_special_swappables): Update call accordingly.
13888 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13890 * simplify-rtx.c (simplify_const_binary_operation): Use
13891 CONST_VECTOR_ELT instead of XVECEXP.
13893 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13895 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13896 the selector elements to be different from the data elements
13897 if the selector is a VECTOR_CST.
13898 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13899 ssizetype for the selector.
13901 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13903 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13904 before testing each element individually.
13905 * tree-vect-generic.c (lower_vec_perm): Likewise.
13907 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13909 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13910 * selftest-run-tests.c (selftest::run_tests): Call it.
13911 * vector-builder.h (vector_builder::operator ==): New function.
13912 (vector_builder::operator !=): Likewise.
13913 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13914 (vec_perm_indices::all_from_input_p): New function.
13915 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13916 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13917 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13918 instead of reading the VECTOR_CST directly. Detect whether both
13919 vector inputs are the same before constructing the vec_perm_indices,
13920 and update the number of inputs argument accordingly. Use the
13921 utility functions added above. Only construct sel2 if we need to.
13923 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13925 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13926 the broadcast of the low byte.
13927 (expand_mult_highpart): Use an explicit encoding for the permutes.
13928 * optabs-query.c (can_mult_highpart_p): Likewise.
13929 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13930 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13931 (vectorizable_bswap): Likewise.
13932 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13933 explicit encoding for the power-of-2 permutes.
13934 (vect_permute_store_chain): Likewise.
13935 (vect_grouped_load_supported): Likewise.
13936 (vect_permute_load_chain): Likewise.
13938 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13940 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13941 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13942 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13943 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13944 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13945 (vect_gen_perm_mask_any): Likewise.
13947 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13949 * int-vector-builder.h: New file.
13950 * vec-perm-indices.h: Include int-vector-builder.h.
13951 (vec_perm_indices): Redefine as an int_vector_builder.
13952 (auto_vec_perm_indices): Delete.
13953 (vec_perm_builder): Redefine as a stand-alone class.
13954 (vec_perm_indices::vec_perm_indices): New function.
13955 (vec_perm_indices::clamp): Likewise.
13956 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13957 (vec_perm_indices::new_vector): New function.
13958 (vec_perm_indices::new_expanded_vector): Update for new
13959 vec_perm_indices class.
13960 (vec_perm_indices::rotate_inputs): New function.
13961 (vec_perm_indices::all_in_range_p): Operate directly on the
13962 encoded form, without computing elided elements.
13963 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13964 encoding. Update for new vec_perm_indices class.
13965 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13966 the given vec_perm_builder.
13967 (expand_vec_perm_var): Update vec_perm_builder constructor.
13968 (expand_mult_highpart): Use vec_perm_builder instead of
13969 auto_vec_perm_indices.
13970 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13971 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13972 or double series encoding as appropriate.
13973 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13974 vec_perm_indices instead of auto_vec_perm_indices.
13975 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13976 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13977 (vect_permute_store_chain): Likewise.
13978 (vect_grouped_load_supported): Likewise.
13979 (vect_permute_load_chain): Likewise.
13980 (vect_shift_permute_load_chain): Likewise.
13981 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13982 (vect_transform_slp_perm_load): Likewise.
13983 (vect_schedule_slp_instance): Likewise.
13984 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13985 (vectorizable_mask_load_store): Likewise.
13986 (vectorizable_bswap): Likewise.
13987 (vectorizable_store): Likewise.
13988 (vectorizable_load): Likewise.
13989 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13990 vec_perm_indices instead of auto_vec_perm_indices. Use
13991 tree_to_vec_perm_builder to read the vector from a tree.
13992 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13993 vec_perm_builder instead of a vec_perm_indices.
13994 (have_whole_vector_shift): Use vec_perm_builder and
13995 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13996 truncation to calc_vec_perm_mask_for_shift.
13997 (vect_create_epilog_for_reduction): Likewise.
13998 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13999 from auto_vec_perm_indices to vec_perm_indices.
14000 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14001 instead of changing individual elements.
14002 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14003 the vector in d.perm.
14004 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14005 from auto_vec_perm_indices to vec_perm_indices.
14006 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14007 instead of changing individual elements.
14008 (arm_vectorize_vec_perm_const): Use new_vector to install
14009 the vector in d.perm.
14010 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14011 Update vec_perm_builder constructor.
14012 (rs6000_expand_interleave): Likewise.
14013 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14014 (rs6000_expand_interleave): Likewise.
14016 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14018 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14019 to qimode could truncate the indices.
14020 * optabs.c (expand_vec_perm_var): Likewise.
14022 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14024 * Makefile.in (OBJS): Add vec-perm-indices.o.
14025 * vec-perm-indices.h: New file.
14026 * vec-perm-indices.c: Likewise.
14027 * target.h (vec_perm_indices): Replace with a forward class
14029 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14030 * optabs.h: Include vec-perm-indices.h.
14031 (expand_vec_perm): Delete.
14032 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14033 (expand_vec_perm_const): Declare.
14034 * target.def (vec_perm_const_ok): Replace with...
14035 (vec_perm_const): ...this new hook.
14036 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14037 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14038 * doc/tm.texi: Regenerate.
14039 * optabs.def (vec_perm_const): Delete.
14040 * doc/md.texi (vec_perm_const): Likewise.
14041 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14042 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14043 expand_vec_perm for constant permutation vectors. Assert that
14044 the mode of variable permutation vectors is the integer equivalent
14045 of the mode that is being permuted.
14046 * optabs-query.h (selector_fits_mode_p): Declare.
14047 * optabs-query.c: Include vec-perm-indices.h.
14048 (selector_fits_mode_p): New function.
14049 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14050 is defined, instead of checking whether the vec_perm_const_optab
14051 exists. Use targetm.vectorize.vec_perm_const instead of
14052 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14053 fit in the vector mode before using a variable permute.
14054 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14055 vec_perm_indices instead of an rtx.
14056 (expand_vec_perm): Replace with...
14057 (expand_vec_perm_const): ...this new function. Take the selector
14058 as a vec_perm_indices rather than an rtx. Also take the mode of
14059 the selector. Update call to shift_amt_for_vec_perm_mask.
14060 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14061 Use vec_perm_indices::new_expanded_vector to expand the original
14062 selector into bytes. Check whether the indices fit in the vector
14063 mode before using a variable permute.
14064 (expand_vec_perm_var): Make global.
14065 (expand_mult_highpart): Use expand_vec_perm_const.
14066 * fold-const.c: Includes vec-perm-indices.h.
14067 * tree-ssa-forwprop.c: Likewise.
14068 * tree-vect-data-refs.c: Likewise.
14069 * tree-vect-generic.c: Likewise.
14070 * tree-vect-loop.c: Likewise.
14071 * tree-vect-slp.c: Likewise.
14072 * tree-vect-stmts.c: Likewise.
14073 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14075 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14076 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14077 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14078 (aarch64_vectorize_vec_perm_const): ...this new function.
14079 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14080 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14081 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14082 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14083 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14084 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14085 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14087 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14088 check for NEON modes.
14089 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14090 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14091 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14092 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14094 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14095 the old VEC_PERM_CONST conditions.
14096 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14097 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14098 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14099 (ia64_vectorize_vec_perm_const_ok): Merge into...
14100 (ia64_vectorize_vec_perm_const): ...this new function.
14101 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14102 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14103 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14104 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14105 * config/mips/mips.c (mips_expand_vec_perm_const)
14106 (mips_vectorize_vec_perm_const_ok): Merge into...
14107 (mips_vectorize_vec_perm_const): ...this new function.
14108 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14109 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14110 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14111 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14112 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14113 (rs6000_expand_vec_perm_const): Delete.
14114 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14116 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14117 (altivec_expand_vec_perm_const_le): Take each operand individually.
14118 Operate on constant selectors rather than rtxes.
14119 (altivec_expand_vec_perm_const): Likewise. Update call to
14120 altivec_expand_vec_perm_const_le.
14121 (rs6000_expand_vec_perm_const): Delete.
14122 (rs6000_vectorize_vec_perm_const_ok): Delete.
14123 (rs6000_vectorize_vec_perm_const): New function.
14124 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14125 an element count and rtx array.
14126 (rs6000_expand_extract_even): Update call accordingly.
14127 (rs6000_expand_interleave): Likewise.
14128 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14129 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14130 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14131 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14132 (rs6000_expand_vec_perm_const): Delete.
14133 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14134 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14135 (altivec_expand_vec_perm_const_le): Take each operand individually.
14136 Operate on constant selectors rather than rtxes.
14137 (altivec_expand_vec_perm_const): Likewise. Update call to
14138 altivec_expand_vec_perm_const_le.
14139 (rs6000_expand_vec_perm_const): Delete.
14140 (rs6000_vectorize_vec_perm_const_ok): Delete.
14141 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14142 reference to the SPE evmerge intructions.
14143 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14144 an element count and rtx array.
14145 (rs6000_expand_extract_even): Update call accordingly.
14146 (rs6000_expand_interleave): Likewise.
14147 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14148 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14150 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14152 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14154 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14155 vector mode and that that mode matches the mode of the data
14157 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14158 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14159 directly using expand_vec_perm_1 when forcing selectors into
14161 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14163 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14165 * optabs-query.h (can_vec_perm_p): Delete.
14166 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14167 * optabs-query.c (can_vec_perm_p): Split into...
14168 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14169 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14170 particular selector is valid.
14171 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14172 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14173 (vect_grouped_load_supported): Likewise.
14174 (vect_shift_permute_load_chain): Likewise.
14175 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14176 (vect_transform_slp_perm_load): Likewise.
14177 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14178 (vectorizable_bswap): Likewise.
14179 (vect_gen_perm_mask_checked): Likewise.
14180 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14181 implementations of variable permutation vectors into account
14182 when deciding which selector to use.
14183 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14184 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14185 with a false third argument.
14186 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14187 to test whether the constant selector is valid and can_vec_perm_var_p
14188 to test whether a variable selector is valid.
14190 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14192 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14193 * optabs-query.c (can_vec_perm_p): Likewise.
14194 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14195 instead of vec_perm_indices.
14196 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14197 (vect_gen_perm_mask_checked): Likewise,
14198 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14199 (vect_gen_perm_mask_checked): Likewise,
14201 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14203 * optabs-query.h (qimode_for_vec_perm): Declare.
14204 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14205 (qimode_for_vec_perm): ...this new function.
14206 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14208 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14210 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14211 does not have a conditional at the top.
14213 2018-01-02 Richard Biener <rguenther@suse.de>
14215 * ipa-inline.c (big_speedup_p): Fix expression.
14217 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14220 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14223 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14227 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14228 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14229 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14230 cond_taken_branch_cost 3->4.
14232 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14234 PR tree-optimization/83581
14235 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14236 TODO_cleanup_cfg if any changes have been made.
14238 PR middle-end/83608
14239 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14240 convert_modes if target mode has the right side, but different mode
14243 PR middle-end/83609
14244 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14245 last argument when extracting from CONCAT. If either from_real or
14246 from_imag is NULL, use expansion through memory. If result is not
14247 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14248 the parts directly to inner mode, if even that fails, use expansion
14251 PR middle-end/83623
14252 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14253 check for bswap in mode rather than HImode and use that in expand_unop
14256 Copyright (C) 2018 Free Software Foundation, Inc.
14258 Copying and distribution of this file, with or without modification,
14259 are permitted in any medium without royalty provided the copyright
14260 notice and this notice are preserved.