1 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
4 (optab, sve_int_op): Handle div and udiv.
5 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
7 (*<optab><mode>3): New insn for the same.
9 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11 * tree-vect-patterns.c: Include predict.h.
12 (vect_recog_divmod_pattern): Restrict check for division support
13 to when optimizing for size.
15 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
17 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
18 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
19 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
20 (gimple_match_op::set_op): Likewise.
21 (gimple_resimplify4): Declare.
22 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
23 (expr::gen_transform): Likewise.
24 (decision_tree::gen): Generate a simplification routine for 4 operands.
25 * gimple-match-head.c (gimple_simplify): Add an overload for
26 4 operands. In the top-level function, handle up to 4 call
27 arguments and call gimple_resimplify4.
28 (gimple_resimplify4): New function.
29 (build_call_internal): Pass a fourth operand.
30 (maybe_push_to_seq): Likewise.
31 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
32 Fold VEC_COND_EXPRs of an operation and a default value into
33 an IFN_COND_* function if possible.
34 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
36 (SVE_COND_FP_BINARY): Include them.
37 (optab, sve_fp_op): Handle them.
38 (SVE_INT_BINARY_REV): New code iterator.
39 (SVE_COND_FP_BINARY_REV): New int iterator.
40 (commutative): New int attribute.
41 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
43 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
45 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
46 (*cond_<optab><mode>): New patterns for reversed operands.
48 2018-05-25 Richard Biener <rguenther@suse.de>
50 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
51 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
52 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
53 (STMT_VINFO_GROUPED_ACCESS): Adjust.
54 * tree-vect-data-refs.c (everywhere): Adjust users.
55 * tree-vect-loop.c (everywhere): Likewise.
56 * tree-vect-slp.c (everywhere): Likewise.
57 * tree-vect-stmts.c (everywhere): Likewise.
58 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
60 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
62 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
64 (gcc_cv_as_section_exclude): ... this.
65 Try Solaris as #exclude syntax.
66 * configure: Regenerate.
67 * config.in: Regenerate.
68 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
70 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
71 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
73 * varasm.c (default_elf_asm_named_section): Don't check if
74 HAVE_GAS_SECTION_EXCLUDE is defined.
76 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
78 * doc/md.texi: Update the documentation of the cond_* optabs
79 to mention the new final operand. Fix GET_MODE_NUNITS call.
80 Describe the scalar case too.
81 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
82 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
84 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
85 (get_conditional_internal_fn): Update comment.
86 * tree-vect-loop.c (vectorizable_reduction): Pass the original
87 accumulator value as a final argument to conditional functions.
88 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
89 a define_expand and add an "else" operand. Assert for now that
90 the else operand is equal to operand 2. Use SVE_INT_BINARY and
91 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
92 (*cond_<optab><mode>): New patterns.
93 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
94 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
95 (UNSPEC_COND_EOR): Delete.
96 (optab): Remove associated mappings.
97 (SVE_INT_BINARY): New code iterator.
98 (sve_int_op): Remove int attribute and add "minus" to the code
100 (SVE_COND_INT_OP): Delete.
101 (SVE_COND_FP_OP): Rename to...
102 (SVE_COND_FP_BINARY): ...this.
104 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
106 * optabs.c (can_reuse_operands_p): New function.
107 (maybe_legitimize_operands): Try to reuse the results for
110 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
112 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
113 Add {q} suffix to insn mnemonic.
115 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
117 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
118 (msp430_warn_func_return): New.
120 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
122 * fold-const.c (tree_nonzero_bits): New function.
123 * fold-const.h (tree_nonzero_bits): Likewise.
124 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
125 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
127 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
131 * varasm.c (assemble_alias): Check ifunc_resolver only on
134 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
137 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
138 when memory input operand is handled.
140 2018-05-24 Luis Machado <luis.machado@linaro.org>
142 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
144 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
146 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
148 * match.pd: Delay FMA folds until after vectorization.
150 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
153 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
154 address check not strict.
156 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
158 * gimple-match.h (gimple_match_op): New class.
159 (mprts_hook): Replace parameters with a gimple_match_op *.
160 (maybe_build_generic_op): Likewise.
161 (gimple_simplified_result_is_gimple_val): Replace parameters with
162 a const gimple_match_op *.
163 (gimple_simplify): Replace code_helper * and tree * parameters with
164 a gimple_match_op * parameter.
165 (gimple_resimplify1): Replace code_helper *, tree and tree *
166 parameters with a gimple_match_op * parameter.
167 (gimple_resimplify2): Likewise.
168 (gimple_resimplify3): Likewise.
169 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
170 parameters with a gimple_match_op * parameter.
171 * gimple-match-head.c (gimple_simplify): Change prototypes of
172 auto-generated functions to take a gimple_match_op * instead of
173 separate code_helper * and tree * parameters. Make the same
174 change in the top-level overload and update calls to the
175 gimple_resimplify routines. Update calls to the auto-generated
176 functions and to maybe_push_res_to_seq in the publicly-facing
177 operation-specific gimple_simplify overloads.
178 (gimple_match_op::MAX_NUM_OPS): Define.
179 (gimple_resimplify1): Replace rcode and ops with a single res_op
180 parameter. Update call to gimple_simplify.
181 (gimple_resimplify2): Likewise.
182 (gimple_resimplify3): Likewise.
183 (mprts_hook): Replace parameters with a gimple_match_op *.
184 (maybe_build_generic_op): Likewise.
185 (build_call_internal): Replace type, nargs and ops with
187 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
188 with a single gimple_match_op *. Update calls to mprts_hook,
189 build_call_internal and gimple_simplified_result_is_gimple_val.
190 Factor out code that is common to the tree_code and combined_fn cases.
191 * genmatch.c (expr::gen_transform): Replace tem_code and
192 tem_ops with a gimple_match_op called tem_op. Update calls
193 to the gimple_resimplify functions and maybe_push_res_to_seq.
194 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
195 res_ops. Update call to the gimple_resimplify functions.
196 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
197 (decision_tree::gen): Make the functions take a gimple_match_op *
198 called res_op instead of separate res_code and res_ops parameters.
199 Update call accordingly.
200 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
201 and ops with a single res_op parameter. Update calls to
202 maybe_build_generic_op and maybe_push_res_to_seq.
203 (fold_stmt_1): Update calls to gimple_simplify and
204 replace_stmt_with_simplification.
205 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
206 and gimple_simplified_result_is_gimple_val.
207 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
209 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
210 with a gimple_match_op *.
211 (vn_nary_build_or_lookup): Likewise. Update call to
212 vn_nary_build_or_lookup_1.
213 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
214 gimple_match_op *. Update calls to the gimple_resimplify routines
215 and to gimple_simplified_result_is_gimple_val.
216 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
217 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
218 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
219 (visit_nary_op): Likewise.
220 (visit_reference_op_load): Likewise.
222 2018-05-23 Luis Machado <luis.machado@linaro.org>
224 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
225 modifier for printing the step amount.
227 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
230 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
233 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
235 * doc/sourcebuild.texi (Endianness): New subsubsection.
237 2018-05-23 Luis Machado <luis.machado@linaro.org>
239 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
240 <prefetch_dynamic_strides>: New const bool field.
241 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
242 prefetch_dynamic_strides.
243 (exynosm1_prefetch_tune): Likewise.
244 (thunderxt88_prefetch_tune): Likewise.
245 (thunderx_prefetch_tune): Likewise.
246 (thunderx2t99_prefetch_tune): Likewise.
247 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
249 (aarch64_override_options_internal): Update to set
250 PARAM_PREFETCH_DYNAMIC_STRIDES.
251 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
252 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
253 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
254 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
255 prefetch-dynamic-strides setting.
257 2018-05-23 Luis Machado <luis.machado@linaro.org>
259 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
260 <minimum_stride>: New const int field.
261 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
262 minimum_stride field defaulting to -1.
263 (exynosm1_prefetch_tune): Likewise.
264 (thunderxt88_prefetch_tune): Likewise.
265 (thunderx_prefetch_tune): Likewise.
266 (thunderx2t99_prefetch_tune): Likewise.
267 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
268 <default_opt_level>: Set to 3.
269 (aarch64_override_options_internal): Update to set
270 PARAM_PREFETCH_MINIMUM_STRIDE.
271 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
272 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
273 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
274 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
275 stride is constant and is below the minimum stride threshold.
277 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
279 * config/arm/arm-cpus.in (mode26): Delete.
280 (armv4): Delete mode26 reference.
281 * config/arm/arm.c (arm_configure_build_target): Delete use of
284 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
286 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
288 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
289 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
291 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
292 (floatunsdidf2): Ditto.
294 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
296 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
297 (fixuns_trunc<mode>si2_avx512f): Ditto.
298 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
299 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
300 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
302 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
304 PR rtl-optimization/79985
305 * df-scan.c (df_insn_refs_collect): Remove special case for
306 global registers and asm statements.
308 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
310 * extend.texi (Global Register Variables): Rewrite the bullet list.
311 Note that the register is available for allocation. Note that access
312 via inline asm must use constraints. Add note about async-signal
313 handlers. Remove paragraph about automagic register selection.
315 2018-05-23 Richard Biener <rguenther@suse.de>
317 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
318 of fixed offset from memset VN.
320 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
322 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
324 (alloc_cand_and_find_basis): Initialize first_interp field.
325 (slsr_process_mul): Modify first_interp field.
326 (slsr_process_add): Likewise.
327 (slsr_process_cast): Modify first_interp field for each new
329 (slsr_process_copy): Likewise.
330 (dump_candidate): Dump first_interp field.
331 (replace_mult_candidate): Process all interpretations, not just
333 (replace_rhs_if_not_dup): Likewise.
334 (replace_one_candidate): Likewise.
336 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
338 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
340 (aarch64_needs_frame_chain): New function.
341 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
343 2018-05-23 Sudakshina Das <sudi.das@arm.com>
346 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
347 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
348 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
349 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
350 as true for strict-align.
351 (aarch64_can_inline_p): Perform checks even when callee has no
352 attributes to check for strict alignment.
353 * doc/extend.texi (AArch64 Function Attributes): Document
355 * doc/invoke.texi: (AArch64 Options): Likewise.
357 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
359 PR tree-optimization/85853
360 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
361 the handling of the root of the node to...
362 (vect_slp_analyze_node_operations_1): ...this new function,
363 and run the whole thing with the child nodes' def types
364 set according to their SLP node's def type.
366 2018-05-23 Richard Biener <rguenther@suse.de>
369 * tree-data-ref.c (create_runtime_alias_checks): Defer
370 and ignore overflow warnings.
372 2018-05-23 Yury Gribov <tetra2005@gmail.com>
374 PR tree-optimization/85822
375 * tree-vrp.c (is_masked_range_test): Fix handling of negative
378 2018-05-23 Richard Biener <rguenther@suse.de>
380 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
381 memset constants via native_interpret_expr.
383 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
386 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
388 (cgraph_node::create_alias): Likewise.
389 (cgraph_node::get_availability): Check ifunc_resolver instead
390 of looking up ifunc attribute.
391 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
392 * varasm.c (do_assemble_alias): Likewise.
393 (assemble_alias): Likewise.
394 (default_binds_local_p_3): Likewise.
395 * cgraph.h (cgraph_node): Add ifunc_resolver.
396 (cgraph_node::only_called_directly_or_aliased_p): Return false
398 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
400 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
401 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
402 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
403 instead of looking up ifunc attribute.
405 2018-05-22 Luis Machado <luis.machado@linaro.org>
407 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
409 2018-05-22 Martin Sebor <msebor@redhat.com>
412 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
413 only when expasion succeeds.
414 (expand_builtin_strcmp): Same.
415 (expand_builtin_strncmp): Same.
417 2018-05-22 Martin Sebor <msebor@redhat.com>
419 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
421 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
422 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
424 * config/aarch64/aarch64-ldpstp.md: Replace uses of
425 aarch64_mem_pair_operand with memory_operand and delete operand swapping
427 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
428 Add check for legitimate_address.
429 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
430 (aarch64_swap_ldrstr_operands): New.
431 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
434 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
435 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
437 * config/aarch64/aarch64.md: New patterns to generate stp
439 (store_pair_sw, store_pair_dw): New patterns to generate stp for
440 single words and double words.
441 (load_pair_sw, load_pair_dw): Likewise.
442 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
444 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
446 * config/aarch64/aarch64-ldpstp.md: Modify peephole
447 for different mode ldpstp and add peephole for merged zero stores.
449 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
451 (aarch64_gen_store_pair): Rename calls to match new patterns.
452 (aarch64_gen_load_pair): Rename calls to match new patterns.
453 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
454 (load_pair<DREG:mode><DREG2:mode>): ... This.
455 (store_pair<mode>): Rename to...
456 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
457 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
459 (V_INT_EQUIV): Handle SImode.
460 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
463 2018-05-22 Martin Sebor <msebor@redhat.com>
466 * calls.c (maybe_warn_nonstring_arg): Use string length to set
467 or ajust the presumed bound on an operation to avoid unnecessary
470 2018-05-22 Martin Sebor <msebor@redhat.com>
472 PR tree-optimization/85826
473 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
474 assuming that a DECL necesarily has a constant size.
476 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
479 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
481 2018-05-22 Richard Biener <rguenther@suse.de>
483 PR tree-optimization/85834
484 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
485 non-constant and non-zero memset arguments.
487 2018-05-22 Martin Liska <mliska@suse.cz>
490 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
492 2018-05-22 Richard Biener <rguenther@suse.de>
494 PR tree-optimization/85863
495 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
496 comparisons when vectype is specified.
497 (vectorizable_condition): Do not specify vectype for
498 vect_is_simple_cond when SLP vectorizing.
500 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
503 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
504 define __ibm128 as long double.
505 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
506 as a distinct type when IEEE 128-bit support is enabled.
507 (init_float128_ieee): Fix up conversions between IFmode and IEEE
508 128-bit types to use the correct functions.
509 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
510 convert between 128-bit floating point types that have different
511 modes but the same representation, instead of using gen_lowpart to
513 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
515 (IFKF_reg): New attributes to give the register constraints for
517 (extend<mode>tf2_internal): New insns to mark an explicit
518 conversion between 128-bit floating point types that have a
519 different mode but share the same representation.
521 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
523 PR tree-optimization/85814
524 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
525 a null return from get_strinfo when unsharing the next
526 strinfo in the chain.
528 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
531 * varasm.c (weak_finish): Clean up weak_decls.
533 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
535 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
536 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
537 UNSPEC_UADALP values.
538 * config/aarch64/iterators.md (ABAL): New int iterator.
541 (sur): Add mappings for the above.
542 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
544 (aarch64_<sur>abal<mode>_4): Likewise.
545 (aarch64_<sur>adalp<mode>_3): Likewise.
546 (<sur>sadv16qi): New define_expand.
548 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
550 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
551 (*movdf_internal): Ditto.
552 (*rcpsf2_sse): Ditto.
553 (*rsqrtsf2_sse): Ditto.
554 (*sqrt<mode>2_sse): Ditto.
556 2018-05-21 Tamar Christina <tamar.christina@arm.com>
558 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
560 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
561 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
562 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
563 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
565 * config/aarch64/arm_neon.h: Likewise.
566 * config/aarch64/iterators.md (VQ_I): New.
568 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
570 * config.gcc: Add arc/t-multilib-linux to tmake_file for
572 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
575 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
577 * config/nds32/constraints.md (S): New constraint.
578 * config/nds32/nds32.md (call_internal): Use constraint S.
579 (call_value_internal): Likewise.
580 (sibcall_internal): Likewise.
581 (sibcall_value_internal): Likewise.
583 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
584 Chung-Ju Wu <jasonwucj@gmail.com>
586 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
589 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
590 Chung-Ju Wu <jasonwucj@gmail.com>
592 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
593 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
594 (nds32_rtx_costs_impl): Simplify.
595 (nds32_address_cost_impl): Simplify.
596 (nds32_init_rtx_costs): New function.
597 (nds32_rtx_costs_speed_prefer): Likewise.
598 (nds32_rtx_costs_size_prefer): Likewise.
599 (nds32_address_cost_speed_prefer): Likewise.
600 (nds32_address_cost_speed_fwprop): Likewise.
601 (nds32_address_cost_size_prefer): Likewise.
602 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
603 * config/nds32/nds32.c (nds32_option_override): Use
604 nds32_init_rtx_costs function.
606 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
608 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
609 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
610 (TARGET_PIPELINE_N8): Likewise.
611 (TARGET_PIPELINE_N10): Likewise.
612 (TARGET_PIPELINE_N13): Likewise.
613 (TARGET_PIPELINE_GRAYWOLF): Likewise.
615 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
617 * config/nds32/nds32-fpu.md: Update copyright year.
619 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
621 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
623 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
625 * config/nds32/nds32.c
626 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
627 * config/nds32/nds32.opt (minline-asm-r15): New option.
629 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
631 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
633 * config/nds32/nds32.md (abssi2): New pattern.
635 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
637 * config/i386/i386.md (rex64namesuffix): New mode attribute.
638 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
639 Merge insn pattern from sse_cvtsi2ss<round_name> and
640 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
641 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
642 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
643 using SWI48 mode iterator.
644 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
645 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
646 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
647 pattern from sse_cvttss2si<round_saeonly_name>
648 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
649 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
650 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
651 using SWI48 mode iterator.
652 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
653 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
654 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
655 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
656 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
657 using SWI48 mode iterator.
658 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
659 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
660 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
661 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
662 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
664 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
665 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
666 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
667 pattern from sse_cvttsd2si<round_saeonly_name>
668 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
670 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
672 * config/nds32/nds32-md-auxiliary.c
673 (nds32_valid_smw_lwm_base_p): Refine.
674 (nds32_output_smw_single_word): Refine.
675 (nds32_output_smw_double_word): New.
676 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
678 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
680 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
681 (nds32_output_stack_pop): Refine.
682 (nds32_expand_unaligned_load): Refine.
683 (nds32_expand_unaligned_store): Refine.
685 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
686 Chung-Ju Wu <jasonwucj@gmail.com>
688 * config/nds32/constants.md: Add TP_REGNUM constant.
689 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
690 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
692 * config/nds32/nds32-doubleword.md: Consider flag_pic.
693 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
694 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
695 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
696 and PIC code generation.
697 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
699 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
701 * config/nds32/nds32.md: Support TLS and PIC.
702 * config/nds32/nds32.c: Support TLS and PIC.
703 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
704 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
707 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
709 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
712 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
713 Chung-Ju Wu <jasonwucj@gmail.com>
715 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
716 * config/nds32/nds32-md-auxiliary.c
717 (symbolic_reference_mentioned_p): New.
718 (nds32_legitimize_ict_address): New.
719 (nds32_expand_ict_move): New.
720 (nds32_indirect_call_referenced_p): New.
721 (nds32_symbol_binds_local_p): Delete.
722 (nds32_long_call_p): Modify.
723 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
724 * config/nds32/nds32-protos.h
725 (symbolic_reference_mentioned_p): Declare.
726 (nds32_legitimize_ict_address): Declare.
727 (nds32_expand_ict_move): Declare.
728 (nds32_indirect_call_referenced_p): Declare.
729 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
730 (nds32_relax_group): Use nds32_ict_const_p as condition.
731 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
732 (nds32_asm_file_start): Output ict_model directive in asm code.
733 (nds32_legitimate_address_p): Consider indirect call.
734 (nds32_print_operand): Consider indirect call.
735 (nds32_print_operand_address): Consider indirect call.
736 (nds32_insert_attributes): Handle "indirect_call" attribute.
737 (TARGET_LEGITIMATE_ADDRESS_P): Define.
738 (TARGET_LEGITIMATE_CONSTANT_P): Define.
739 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
740 (TARGET_DELEGITIMIZE_ADDRESS): Define.
741 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
742 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
743 (TARGET_ICT_MODEL_SMALL): Define.
744 (TARGET_ICT_MODEL_LARGE): Define.
745 * config/nds32/nds32.md (movsi): Consider ict model.
746 (call, call_value): Consider ict model.
747 (sibcall, sibcall_value): Consider ict model.
748 * config/nds32/nds32.opt (mict-model): New option.
749 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
752 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
753 Monk Chiang <sh.chiang04@gmail.com>
754 Jim Wilson <jimw@sifive.com>
756 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
757 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
758 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
759 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
760 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
761 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
762 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
763 compute save_libcall_adjustment properly.
764 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
765 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
766 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
767 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
768 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
769 (ABI_SPEC): Handle mabi=ilp32e.
770 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
772 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
773 <-march>: Add rv32e as an example.
775 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
778 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
779 (intra_create_variable_infos): Handle C++ constructors.
781 2018-05-18 Martin Liska <mliska@suse.cz>
783 * passes.def: Remove a redundant pass.
785 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
788 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
790 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
792 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
794 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
796 (armv2, armv2a, armv3, armv3m): Delete architectures.
797 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
798 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
799 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
801 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
802 (*mulsidi3adddi): Likewise.
803 (mulsidi3): Likewise.
804 (*mulsidi3_nov6): Likewise.
805 (umulsidi3): Likewise.
806 (umulsidi3_nov6): Likewise.
807 (umaddsidi4): Likewise.
808 (*umulsidi3adddi): Likewise.
809 (smulsi3_highpart): Likewise.
810 (*smulsi3_highpart_nov6): Likewise.
811 (umulsi3_highpart): Likewise.
812 (*umulsi3_highpart_nov6): Likewise.
813 * config/arm/arm.h (arm_arch3m): Delete.
814 * config/arm/arm.c (arm_arch3m): Delete.
815 (arm_option_override_internal): Update armv3-related comment.
816 (arm_configure_build_target): Delete use of isa_bit_mode32.
817 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
818 (arm_rtx_costs_internal): Delete check of arm_arch3m.
819 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
822 * config/arm/arm-protos.h (arm_arch3m): Delete.
823 * config/arm/arm-tables.opt: Regenerate.
824 * config/arm/arm-tune.md: Likewise.
825 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
826 deleted architectures.
828 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
830 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
831 (armv5t, armv5te): New features.
832 (ARMv5, ARMv5e): Delete fgroups.
833 (ARMv5t, ARMv5te): Adjust for above changes.
835 (armv5, armv5e): Delete arches.
836 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
838 (*call_reg_arm): Likewise.
839 (*call_value_reg_armv5): Likewise.
840 (*call_value_reg_arm): Likewise.
841 (*call_symbol): Likewise.
842 (*call_value_symbol): Likewise.
843 (*sibcall_insn): Likewise.
844 (*sibcall_value_insn): Likewise.
846 (prefetch): Likewise.
847 (define_split and define_peephole2 dependent on arm_arch5):
849 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
851 (TARGET_ARM_QBIT): Likewise.
852 (TARGET_DSP_MULTIPLY): Likewise.
853 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
854 (arm_arch5, arm_arch5e): Delete.
855 (arm_arch5t, arm_arch5te): Declare.
856 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
857 (arm_arch5t): Declare.
858 (arm_option_reconfigure_globals): Update for the above.
859 (arm_options_perform_arch_sanity_checks): Update comment, replace
860 use of arm_arch5 with arm_arch5t.
861 (use_return_insn): Likewise.
862 (arm_emit_call_insn): Likewise.
863 (output_return_instruction): Likewise.
864 (arm_final_prescan_insn): Likewise.
865 (arm_coproc_builtin_available): Likewise.
866 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
867 arm_arch5e with arm_arch5t and arm_arch5te.
868 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
869 (arm_arch5t, arm_arch5te): Declare.
870 * config/arm/arm-tables.opt: Regenerate.
871 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
872 * config/arm/t-multilib: Likewise.
873 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
874 instead of arm_arch5.
875 (*call_reg_thumb1): Likewise.
876 (*call_value_reg_thumb1_v5): Likewise.
877 (*call_value_reg_thumb1): Likewise.
878 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
880 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
882 2018-05-18 Martin Liska <mliska@suse.cz>
884 PR gcov-profile/84846
885 * doc/gcov.texi: Document -t option of gcov tool.
887 2018-05-18 Martin Liska <mliska@suse.cz>
889 PR gcov-profile/84846
890 * gcov.c (print_usage): Add new -t option.
891 (process_args): Handle the option.
892 (generate_results): Use stdout as output when requested by
895 2018-05-18 Martin Liska <mliska@suse.cz>
897 PR gcov-profile/84846
898 * coverage.c (coverage_init): Write PWD to .gcno file.
899 * doc/gcov.texi: Document how working directory is printed.
900 * gcov-dump.c (dump_gcov_file): Print PWD.
901 * gcov.c (output_intermediate_file): Likewise.
902 (read_graph_file): Read PWD string.
903 (output_lines): Print PWD.
905 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
908 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
909 for retval and return false if all args to phi are zero.
911 2018-05-18 Richard Biener <rguenther@suse.de>
913 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
915 (evrp_dom_walker::before_dom_children): Call it.
917 2018-05-18 Richard Biener <rguenther@suse.de>
919 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
920 results when processing array refs with variable index.
922 2018-05-18 Toon Moene <toon@moene.org>
924 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
925 directly after that of -floop-interchange. Indicate that both
926 options are enabled by default when specifying -O3.
928 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
930 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
931 iterator. Delete separate integer-mode vec_set<mode> expander.
932 (aarch64_simd_vec_setv2di): Delete.
933 (vec_setv2di): Delete.
934 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
935 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
936 the "w, r" alternative.
938 2018-05-18 Martin Liska <mliska@suse.cz>
940 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
941 * tree-pass.h (make_pass_lower_switch_O0): New function.
942 * tree-switch-conversion.c (node_has_low_bound): Remove.
943 (node_has_high_bound): Likewise.
944 (node_is_bounded): Likewise.
945 (class pass_lower_switch): Make it a template type and create
947 (pass_lower_switch::execute): Add template argument.
948 (make_pass_lower_switch): New function.
949 (make_pass_lower_switch_O0): New function.
950 (do_jump_if_equal): Remove.
951 (emit_case_nodes): Simplify to just handle all 3 cases and leave
952 all the hard work to tree optimization passes.
954 2018-05-18 Martin Liska <mliska@suse.cz>
956 * dbgcnt.c (limit_low): Renamed from limit.
957 (limit_high): New variable.
958 (dbg_cnt_is_enabled): Check for upper limit.
959 (dbg_cnt): Adjust dumping.
960 (dbg_cnt_set_limit_by_index): Add new argument for high
962 (dbg_cnt_set_limit_by_name): Likewise.
963 (dbg_cnt_process_single_pair): Parse new format.
964 (dbg_cnt_process_opt): Use strtok.
965 (dbg_cnt_list_all_counters): Remove 'value' and add
967 * doc/invoke.texi: Document changes.
969 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
971 * doc/sourcebuild.texi (scalar_all_fma): Document.
972 * tree.def (FMA_EXPR): Delete.
973 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
974 * internal-fn.c (ternary_direct): New macro.
975 (expand_ternary_optab_fn): Likewise.
976 (direct_ternary_optab_supported_p): Likewise.
977 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
978 * builtins.c (fold_builtin_fma): Delete.
979 (fold_builtin_3): Don't call it.
980 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
981 * expr.c (expand_expr_real_2): Likewise.
982 * fold-const.c (operand_equal_p): Likewise.
983 (fold_ternary_loc): Likewise.
984 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
985 * gimple.c (DEFTREECODE): Likewise.
986 * gimplify.c (gimplify_expr): Likewise.
987 * optabs-tree.c (optab_for_tree_code): Likewise.
988 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
989 * tree-eh.c (operation_could_trap_p): Likewise.
990 (stmt_could_throw_1_p): Likewise.
991 * tree-inline.c (estimate_operator_cost): Likewise.
992 * tree-pretty-print.c (dump_generic_node): Likewise.
993 (op_code_prio): Likewise.
994 * tree-ssa-loop-im.c (stmt_cost): Likewise.
995 * tree-ssa-operands.c (get_expr_operands): Likewise.
996 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
997 * fold-const-call.h (fold_fma): Delete.
998 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
999 CFN_FNMA and CFN_FNMS.
1001 * genmatch.c (combined_fn): New enum.
1002 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1003 (commutative_op): New function.
1004 (commutate): Use it. Handle more than 2 operands.
1005 (dt_operand::gen_gimple_expr): Use commutative_op.
1006 (parser::parse_expr): Allow :c to be used with non-binary
1007 operators if the commutative operand is known.
1008 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1009 CFN_FMS, CFN_FNMA and CFN_FNMS.
1010 (backprop::process_assign_use): Remove FMA_EXPR handling.
1011 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1012 (gen_hsa_fma): New function.
1013 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1014 IFN_FNMA and IFN_FNMS.
1015 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1016 * gimple-fold.h (follow_all_ssa_edges): Declare.
1017 * gimple-fold.c (follow_all_ssa_edges): New function.
1018 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1019 gimple_build interface and use follow_all_ssa_edges to fold the result.
1020 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1021 instead of checking for optabs directly.
1022 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1023 rather than FMA_EXPRs.
1024 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1025 call to IFN_FMA instead of an FMA_EXPR.
1027 2018-05-17 Jim Wilson <jimw@sifive.com>
1029 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1030 sign extended promoted subreg, and the range does not have the sign bit
1031 set, then do a sign extend.
1033 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1034 test, check for sign extended subreg and/or constant operands, and
1035 do a sign extend in that case.
1037 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1039 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1040 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1042 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1043 Change logics_shift_reg to logics_shift_imm.
1044 (thunderx2t99_fp_loadpair_basic): Delete.
1045 (thunderx2t99_fp_storepair_basic): Delete.
1046 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1047 (thunderx2t99_asimd_polynomial): Delete.
1048 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1049 and neon_fp_mul_d_scalar_q.
1050 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1051 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1052 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1053 (thunderx2t99_asimd_lut): Add missing tbl types.
1054 (thunderx2t99_asimd_ext): Delete.
1055 (thunderx2t99_asimd_load1_1_mult): Delete.
1056 (thunderx2t99_asimd_load1_2_mult): Delete.
1057 (thunderx2t99_asimd_load1_ldp): New.
1058 (thunderx2t99_asimd_load1): New.
1059 (thunderx2t99_asimd_load2): Add missing *load2* types.
1060 (thunderx2t99_asimd_load3): New.
1061 (thunderx2t99_asimd_load4): New.
1062 (thunderx2t99_asimd_store1_1_mult): Delete.
1063 (thunderx2t99_asimd_store1_2_mult): Delete.
1064 (thunderx2t99_asimd_store2_mult): Delete.
1065 (thunderx2t99_asimd_store2_onelane): Delete.
1066 (thunderx2t99_asimd_store_stp): New.
1067 (thunderx2t99_asimd_store1): New.
1068 (thunderx2t99_asimd_store2): New.
1069 (thunderx2t99_asimd_store3): New.
1070 (thunderx2t99_asimd_store4): New.
1072 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1074 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1075 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1077 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1078 Segher Boessenkool <segher@kernel.crashing.org>
1081 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1084 2018-05-17 Richard Biener <rguenther@suse.de>
1086 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1087 for pruning loop and prune defs feeding only already visited PHIs.
1089 2018-05-17 Richard Biener <rguenther@suse.de>
1091 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1093 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1094 Richard Biener <rguenther@suse.de>
1096 PR tree-optimization/85793
1097 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1098 for VMAT_ELEMENTWISE.
1100 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1102 * internal-fn.h (lookup_internal_fn): Declare
1103 * internal-fn.c (lookup_internal_fn): New function.
1104 * gimple.c (gimple_build_call_from_tree): Handle calls to
1106 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1107 internal function names.
1108 * tree-pretty-print.c (dump_generic_node): Likewise.
1109 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1111 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1113 * gimple-fold.h (gimple_build): Make the function forms take
1114 combined_fn rather than built_in_function.
1115 (gimple_simplify): Likewise.
1116 * gimple-match-head.c (gimple_simplify): Likewise.
1117 * gimple-fold.c (gimple_build): Likewise.
1118 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1119 rather than gimple_build_call_internal.
1120 (get_initial_defs_for_reduction): Likewise.
1121 (vect_create_epilog_for_reduction): Likewise.
1122 (vectorizable_live_operation): Likewise.
1124 2018-05-17 Martin Liska <mliska@suse.cz>
1126 * gimple-ssa-sprintf.c (format_directive): Do not use
1127 space in between 'G_' and '('.
1129 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1132 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1133 even if the mask is not all ones.
1136 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1138 (ix86_gimple_fold_builtin): Likewise.
1141 * config/i386/i386.c: Include tree-vector-builder.h.
1142 (ix86_vector_shift_count): New function.
1143 (ix86_fold_builtin): Fold shift builtins by scalar count.
1144 (ix86_gimple_fold_builtin): Likewise.
1146 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1147 _mm512_setzero): New intrinsics.
1149 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1150 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1152 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1153 code generation for cases where splatting a value is not useful.
1154 * simplify-rtx.c (simplify_ternary_operation): Simplify
1155 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1156 mode to a vec_concat.
1158 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1160 * config.gcc: Support "goldmont-plus".
1161 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1163 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1164 PROCESSOR_GOLDMONT_PLUS.
1165 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1166 (processor_target_table): Add "goldmont-plus".
1167 (PTA_GOLDMONT_PLUS): Define.
1168 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1169 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1170 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1171 (fold_builtin_cpu): Add "goldmont-plus".
1172 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1173 (ix86_option_override_internal): Add "goldmont-plus".
1174 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1175 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1176 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1177 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1179 2018-05-17 Richard Biener <rguenther@suse.de>
1181 PR tree-optimization/85757
1182 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1183 remove defs that only feed that PHI from further processing.
1185 2018-05-16 Jim Wilson <jimw@sifive.com>
1187 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1189 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1191 2018-05-16 Mark Wielaard <mark@klomp.org>
1193 * dwarf2out.c (count_index_strings): New function.
1194 (output_indirect_strings): Call count_index_strings and generate
1195 header for dwarf_version >= 5.
1197 2018-05-16 Mark Wielaard <mark@klomp.org>
1199 * dwarf2out.c (dwarf_FORM): New function.
1200 (set_indirect_string): Use dwarf_FORM.
1201 (reset_indirect_string): Likewise.
1202 (size_of_die): Likewise.
1203 (value_format): Likewise.
1204 (output_die): Likewise.
1205 (add_skeleton_AT_string): Likewise.
1206 (output_macinfo_op): Likewise.
1207 (index_string): Likewise.
1208 (output_index_string_offset): Likewise.
1209 (output_index_string): Likewise.
1210 (count_index_strings): Likewise.
1212 2018-05-16 Carl Love <cel@us.ibm.com>
1214 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1215 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1217 2018-05-16 Martin Jambor <mjambor@suse.cz>
1219 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1220 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1222 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1224 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1225 (fnma<mode>4): Likewise.
1226 (fms<mode>4): Likewise.
1227 (fnms<mode>4): Likewise.
1228 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1229 (aarch64_fnma<mode>4): Likewise.
1230 (aarch64_fms<mode>4): Likewise.
1231 (aarch64_fnms<mode>4): Likewise.
1232 (aarch64_fnmadd<mode>4): Likewise.
1234 2018-05-16 Jason Merrill <jason@redhat.com>
1236 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1238 2018-05-16 Richard Biener <rguenther@suse.de>
1240 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1241 (dump_stmt_cost): Declare.
1242 (add_stmt_cost): Dump cost we add.
1243 (add_stmt_costs): New function.
1244 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1246 (vect_analyze_stmt): Adjust prototype.
1247 (vectorizable_condition): Likewise.
1248 (vectorizable_live_operation): Likewise.
1249 (vectorizable_reduction): Likewise.
1250 (vectorizable_induction): Likewise.
1251 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1252 cost vector to pass to vectorizable_ and record afterwards.
1253 (vect_model_reduction_cost): Take cost vector argument and adjust.
1254 (vect_model_induction_cost): Likewise.
1255 (vectorizable_reduction): Likewise.
1256 (vectorizable_induction): Likewise.
1257 (vectorizable_live_operation): Likewise.
1258 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1259 SLP_TREE_NUMBER_OF_VEC_STMTS.
1260 (vect_analyze_slp_cost_1): Remove.
1261 (vect_analyze_slp_cost): Likewise.
1262 (vect_slp_analyze_node_operations): Take visited args and
1263 a target cost vector. Avoid processing already visited stmt sets.
1264 (vect_slp_analyze_operations): Use a local cost vector to gather
1265 costs and register those of non-discarded instances.
1266 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1267 (vect_schedule_slp_instance): Remove copying of
1268 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1270 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1271 adding cost. Record cost entry location.
1272 (vect_prologue_cost_for_slp_op): Function to compute cost of
1273 a constant or invariant generated for SLP vect in the prologue,
1274 split out from vect_analyze_slp_cost_1.
1275 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1276 (vect_model_promotion_demotion_cost): Likewise.
1277 (vect_model_store_cost): Likewise, make static.
1278 (vect_model_load_cost): Likewise.
1279 (vectorizable_bswap): Add cost vector arg and adjust.
1280 (vectorizable_call): Likewise.
1281 (vectorizable_simd_clone_call): Likewise.
1282 (vectorizable_conversion): Likewise.
1283 (vectorizable_assignment): Likewise.
1284 (vectorizable_shift): Likewise.
1285 (vectorizable_operation): Likewise.
1286 (vectorizable_store): Likewise.
1287 (vectorizable_load): Likewise.
1288 (vectorizable_condition): Likewise.
1289 (vectorizable_comparison): Likewise.
1290 (can_vectorize_live_stmts): Likewise.
1291 (vect_analyze_stmt): Likewise.
1292 (vect_transform_stmt): Adjust calls to vectorizable_*.
1293 * tree-vectorizer.c: Include gimple-pretty-print.h.
1294 (dump_stmt_cost): New function.
1296 2018-05-16 Richard Biener <rguenther@suse.de>
1298 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1299 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1300 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1301 (check_name): New callback.
1302 (dse_classify_store): Track cycles via a visited bitmap of PHI
1303 defs and simplify handling of in-loop and across loop dead stores
1304 and properly fail for loop-variant refs. Handle byte-tracking with
1305 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1308 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1310 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1311 (vect_get_mask_type_for_stmt): Likewise.
1312 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1314 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1315 to determine the statement's vector type and the vector type that
1316 should be used for calculating nunits. Deal with cases in which
1317 the type has to be deferred.
1318 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1319 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1320 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1321 (vect_determine_vf_for_stmt): New functions, split out from...
1322 (vect_determine_vectorization_factor): ...here.
1323 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1324 (vect_get_mask_type_for_stmt): New functions, split out from
1325 vect_determine_vectorization_factor.
1327 2018-05-16 Richard Biener <rguenther@suse.de>
1329 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1330 verify the [VEC_]COND_EXPR embedded comparison.
1332 2018-05-15 Martin Sebor <msebor@redhat.com>
1334 PR tree-optimization/85753
1335 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1336 RECORD_TYPE in addition to ARRAY_TYPE.
1338 2018-05-15 Martin Sebor <msebor@redhat.com>
1341 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1343 2018-05-15 Richard Biener <rguenther@suse.de>
1345 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1346 add by_clobber_p one. Change algorithm to collect all defs
1347 representing uses we need to walk and try reducing them to
1348 a single one before failing.
1349 (dse_dom_walker::dse_optimize_stmt): Adjust.
1351 2018-05-13 Mark Wielaard <mark@klomp.org>
1353 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1354 (size_of_loc_descr): Likewise.
1355 (output_loc_operands): Likewise.
1356 (output_loc_operands_raw): Likewise.
1357 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1358 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1359 (hash_loc_operands): Likewise.
1360 (compare_loc_operands): Likewise.
1362 2018-05-14 Mark Wielaard <mark@klomp.org>
1364 * dwarf2out.c (count_index_addrs): New function.
1365 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1367 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1369 PR tree-optimization/83648
1370 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1371 return value as malloc candidate.
1373 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1376 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1377 as true in call to suggest_attribute.
1379 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1381 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1382 -mreadonly-in-sdata.
1384 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1386 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1388 (aarch64_crypto_aesd_fused): Likewise.
1390 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1392 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1393 (movsi_aarch64): Likewise.
1394 (load_pairsi): Likewise.
1395 (load_pairdi): Likewise.
1396 (store_pairsi): Likewise.
1397 (store_pairdi): Likewise.
1398 (load_pairsf): Likewise.
1399 (load_pairdf): Likewise.
1400 (store_pairsf): Likewise.
1401 (store_pairdf): Likewise.
1402 (zero_extend): Likewise.
1403 (trunc): Swap alternatives.
1404 (fcvt_target): Add '?' to prefer w over r.
1406 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1409 * config/i386/i386.md: Disallow non-commutative arithmetics in
1410 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1411 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1412 in the peephole2 before it.
1414 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1416 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1417 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1418 (ix86_handle_option): Handle -mcldemote.
1419 * config.gcc: New header.
1420 * config/i386/cldemoteintrin.h: New file.
1421 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1422 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1424 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1425 OPTION_MASK_ISA_CLDEMOTE.
1426 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1427 (ix86_valid_target_attribute_inner_p): Ditto.
1428 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1429 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1430 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1431 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1432 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1434 * config/i386/i386.opt: Add -mcldemote.
1435 * config/i386/x86intrin.h: New header.
1436 * doc/invoke.texi: Add -mcldemote.
1438 2018-05-14 Richard Biener <rguenther@suse.de>
1440 * doc/match-and-simplify.texi: Adjust :s documentation.
1442 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1444 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1445 intended memcpy size.
1446 (REORDER_45): Likewise.
1448 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1450 * sort.cc: New file.
1451 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1452 * vec.c (qsort_chk): Use gcc_qsort.
1453 * Makefile.in (OBJS-libcommon): Add sort.o.
1454 (build/sort.o): New target. Use it...
1455 (BUILD_RTL): ... here, and...
1456 (build/gencfn-macros): ... here, and...
1457 (build/genmatch): ... here.
1459 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1460 Chung-Ju Wu <jasonwucj@gmail.com>
1462 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1463 * config/nds32/nds32-graywolf.md: New file.
1464 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1465 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1467 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1468 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1469 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1470 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1471 * config/nds32/pipelines.md: Include n15 settings.
1473 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1474 Chung-Ju Wu <jasonwucj@gmail.com>
1476 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1477 * config/nds32/nds32-n13.md: New file.
1478 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1479 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1481 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1482 * config/nds32/nds32.md (pipeline_model): Add n13.
1483 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1484 * config/nds32/pipelines.md: Include n13 settings.
1486 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1487 Chung-Ju Wu <jasonwucj@gmail.com>
1489 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1490 * config/nds32/nds32-n10.md: New file.
1491 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1492 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1494 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1495 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1496 * config/nds32/nds32.md (pipeline_model): Add n10.
1497 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1498 * config/nds32/pipelines.md: Include n10 settings.
1500 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1501 Kito Cheng <kito.cheng@gmail.com>
1502 Chung-Ju Wu <jasonwucj@gmail.com>
1504 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1505 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1506 Add enum values for DSP extension instructions.
1507 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1509 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1510 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1512 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1513 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1514 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1515 * config/nds32/nds32-intrinsic.md: Likewise.
1516 * config/nds32/nds32_intrinsic.h: Likewise.
1517 * config/nds32/nds32-md-auxiliary.c: Likewise.
1518 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1519 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1520 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1521 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1522 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1523 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1524 TYPE_DMAC in switch statement.
1525 * config/nds32/nds32.c: New checking and implementation for DSP
1526 extension instructions.
1527 * config/nds32/nds32.h: Likewise.
1528 * config/nds32/nds32.md: Likewise.
1529 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1530 * config/nds32/predicates.md: Implement new predicates for DSP
1533 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1535 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1536 Reformat alternatives and attributes so it is easier to identify
1537 which constraints/attributes go with which instruction.
1538 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1539 (mov<mode>_softfloat32, FMOVE64): Likewise.
1540 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1541 (mov<mode>_softfloat64, FMOVE64): Likewise.
1543 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1545 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1547 (Basic PowerPC Built-in Functions): The new name of the
1548 subsection previously known as "PowerPC Built-in Functions".
1549 (Basic PowerPC Built-in Functions Available on all Configurations):
1551 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1552 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1553 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1554 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1556 2018-05-11 Martin Jambor <mjambor@suse.cz>
1559 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1562 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1565 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1567 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1569 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1570 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1571 (ix86_handle_option): Handle -mwaitpkg.
1572 * config.gcc: New header.
1573 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1574 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1575 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1577 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1578 OPTION_MASK_ISA_WAITPKG.
1579 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1580 (ix86_option_override_internal): Add PTA_WAITPKG.
1581 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1582 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1583 IX86_BUILTIN_TPAUSE.
1584 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1585 __builtin_ia32_umwait and __builtin_ia32_tpause.
1586 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1587 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1588 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1589 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1590 UNSPECV_TPAUSE): New.
1591 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1592 * config/i386/i386.opt: Add -mwaitpkg.
1593 * config/i386/waitpkgintrin.h: New file.
1594 * config/i386/x86intrin.h: New header.
1595 * doc/invoke.texi: Add -mwaitpkg.
1597 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1600 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1602 (cortex-m0): Use armv6s-m isa.
1603 (cortex-m0plus): Likewise.
1604 (cortex-m1): Likewise.
1605 (cortex-m0.small-multiply): Likewise.
1606 (cortex-m0plus.small-multiply): Likewise.
1607 (cortex-m1.small-multiply): Likewise.
1609 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1610 Jakub Jelinek <jakub@redhat.com>
1612 PR tree-optimization/85692
1613 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1614 source permute as well.
1616 2018-05-11 Martin Liska <mliska@suse.cz>
1619 * doc/extend.texi: Document LLVM style format for no_sanitize
1622 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1624 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1625 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1626 (mode_supports_vsx_dform_quad): Likewise.
1627 (mode_supports_vmx_dform): Move these functions to be next to the
1628 other mode_supports functions.
1629 (mode_supports_dq_form): Likewise.
1630 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1631 mode_supports_dq_form.
1632 (reg_offset_addressing_ok_p): Likewise.
1633 (offsettable_ok_by_alignment): Likewise.
1634 (rs6000_legitimate_offset_address_p): Likewise.
1635 (legitimate_lo_sum_address_p): Likewise.
1636 (rs6000_legitimize_address): Likewise.
1637 (rs6000_legitimize_reload_address): Likewise.
1638 (rs6000_secondary_reload_inner): Likewise.
1639 (rs6000_preferred_reload_class): Likewise.
1640 (rs6000_output_move_128bit): Likewise.
1642 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1644 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1645 Generate SImode target register for null target.
1646 <case IX86_BUILTIN_XGETBV>: Ditto.
1647 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1648 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1650 2018-05-10 Carl Love <cel@us.ibm.com>
1652 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1653 dcbtt and dcbtstt if operands[2] is 0.
1655 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1658 * config/i386/sse.md (usadv64qi): New expander.
1660 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1662 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1663 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1664 -maltivec=be support.
1665 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1666 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1667 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1668 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1669 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1670 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1671 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1672 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1673 altivec_vsumsws): Adjust.
1674 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1675 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1676 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1678 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1679 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1680 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1681 (altivec_lve<VI_char>x): Delete expand.
1682 (*altivec_lve<VI_char>x_internal): Rename to...
1683 (altivec_lve<VI_char>x): ... this.
1684 (altivec_lvxl_<mode>): Delete expand.
1685 (*altivec_lvxl_<mode>_internal): Rename to ...
1686 (altivec_lvxl_<mode>): ... this.
1687 (altivec_stvxl_<mode>): Delete expand.
1688 (*altivec_stvxl_<mode>_internal): Rename to ...
1689 (altivec_stvxl_<mode>): ... this.
1690 (altivec_stve<VI_char>x): Delete expand.
1691 (*altivec_stve<VI_char>x_internal): Rename to ...
1692 (altivec_stve<VI_char>x): ... this.
1693 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1694 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1695 reduc_plus_scal_<mode>): Adjust.
1696 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1698 (rs6000_cpu_cpp_builtins): Adjust.
1699 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1700 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1701 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1702 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1703 -maltivec=be support.
1704 (rs6000_split_vec_extract_var): Adjust.
1705 (rs6000_split_v4si_init): Adjust.
1706 (swap_selector_for_mode): Delete.
1707 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1708 altivec_expand_stvex_be): Delete.
1709 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1710 -maltivec=be support.
1711 (rs6000_gimple_fold_builtin): Ditto.
1712 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1714 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1715 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1716 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1717 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1718 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1719 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1720 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1721 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1722 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1723 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1724 anonymous split): Adjust.
1725 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1726 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1728 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1730 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1731 when --with-gxx-include-dir is also specified.
1732 * configure: Regenerate.
1734 2018-05-09 Jim Wilson <jimw@sifive.com>
1737 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1738 * config/riscv/t-withmultilib: New.
1739 * config/riscv/withmultilib.h: New.
1740 * doc/install.texi: Document RISC-V --with-multilib-list support.
1742 2018-05-09 Richard Biener <rguenther@suse.de>
1744 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1746 (vect_bb_vectorization_profitable_p): Adjust. Compute
1747 actual scalar cost using the cost vector and the add_stmt_cost
1750 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1752 PR rtl-optimization/85645
1753 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1754 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1756 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1758 PR rtl-optimization/85645
1759 * shrink-wrap.c (spread_components): Return a boolean saying if
1760 anything was changed.
1761 (try_shrink_wrapping_separate): Iterate spread_components until
1762 nothing changes anymore.
1764 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1766 PR rtl-optimization/85645
1767 * regrename.c (build_def_use): Also kill the chains that include the
1768 destination of a REG_CFA_REGISTER note.
1770 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1772 PR rtl-optimization/85645
1773 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1774 insn that has a REG_CFA_REGISTER note.
1776 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1778 * cfgexpand.c (expand_clobber): New function.
1779 (expand_gimple_stmt_1): Use it.
1780 * tree-vect-stmts.c (vect_clobber_variable): New function,
1782 (vectorizable_simd_clone_call): ...here.
1783 (vectorizable_store): Emit a clobber either side of an
1784 IFN_STORE_LANES sequence.
1785 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1787 2018-05-09 Tom de Vries <tom@codesourcery.com>
1790 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1791 (define_insn "trap_if_false"): Add exit after trap.
1793 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1795 PR rtl-optimization/85638
1796 * bb-reorder.c: Include common/common-target.h.
1797 (create_forwarder_block): New function extracted from...
1798 (fix_up_crossing_landing_pad): ...here. Rename into...
1799 (dw2_fix_up_crossing_landing_pad): ...this.
1800 (sjlj_fix_up_crossing_landing_pad): New function.
1801 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1802 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1803 from both partitions and exit the loop after one iteration.
1805 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1808 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1810 (Basic PowerPC Built-in Functions): The new name of the
1811 subsection previously known as "PowerPC Built-in Functions".
1812 (Basic PowerPC Built-in Functions Available on all Configurations):
1814 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1816 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1818 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1820 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1823 2018-05-08 Jim Wilson <jimw@sifive.com>
1825 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1826 (LD_EMUL_SUFFIX): New.
1827 (LINK_SPEC): Use it.
1829 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1831 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1833 (Basic PowerPC Built-in Functions): The new name of the
1834 subsection previously known as "PowerPC Built-in Functions".
1835 (Basic PowerPC Built-in Functions Available on all Configurations):
1837 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1839 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1841 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1843 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1846 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1849 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1850 after cmpelim optimization.
1852 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1854 * config.gcc: Support "goldmont".
1855 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1856 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1858 * config/i386/i386.c (m_GOLDMONT): Define.
1859 (processor_target_table): Add "goldmont".
1860 (PTA_GOLDMONT): Define.
1861 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1862 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1863 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1864 (fold_builtin_cpu): Add "goldmont".
1865 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1866 (ix86_option_override_internal): Add "goldmont".
1867 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1868 (processor_type): Add PROCESSOR_GOLDMONT.
1869 * config/i386/i386.md: Add CPU "glm".
1870 * config/i386/glm.md: New file.
1871 * config/i386/x86-tune.def: Add m_GOLDMONT.
1872 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1874 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1877 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1879 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1880 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1881 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1884 * config/i386/i386.c (ix86_fold_builtin): Handle
1885 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1888 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1889 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1891 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1894 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1895 (check_arch): Likewise.
1896 (check_fpu): Return the result rather than printing it.
1897 (end arch): Fix operator precedence.
1898 (end cpu): Likewise.
1899 (END): Print the result from check_fpu.
1901 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1902 Alan Hayward <alan.hayward@arm.com>
1903 David Sherwood <david.sherwood@arm.com>
1905 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1906 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1907 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1908 (*fcmuo<mode>_and): New patterns.
1910 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1912 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1913 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1914 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1915 (cmp_op, sve_imm_con): New code attributes.
1916 (SVE_COND_INT_CMP, imm_con): Delete.
1917 (cmp_op): Remove above unspecs from int attribute.
1918 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1920 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1921 comparison-specific unspecs.
1922 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1923 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1924 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1925 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1926 (*vec_fcm<cmp_op><mode>): Rename to...
1927 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1928 (*vec_fcmuo<mode>): Rename to...
1929 (*fcmuo<mode>): ...this and adjust likewise.
1930 (*pred_fcm<cmp_op><mode>): New pattern.
1931 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1932 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1934 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1936 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1937 (aarch64_emit_sve_predicated_cond): New function.
1938 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1939 (aarch64_emit_unspec_cond_or): Replace with...
1940 (aarch64_emit_sve_or_conds): ...this new function. Use
1941 aarch64_emit_sve_ptrue_op for the individual comparisons and
1942 aarch64_emit_binop to OR them together.
1943 (aarch64_emit_inverted_unspec_cond): Replace with...
1944 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1945 aarch64_emit_sve_ptrue_op for the comparison and
1946 aarch64_emit_unop to invert the result.
1947 (aarch64_expand_sve_vec_cmp_float): Update after the above
1948 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1950 2018-05-07 Nathan Sidwell <nathan@acm.org>
1952 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1953 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1954 (Backwards Compatibility): Likewise.
1956 2018-05-07 Luis Machado <luis.machado@linaro.org>
1960 2018-05-07 Luis Machado <luis.machado@linaro.org>
1962 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1963 <prefetch_dynamic_strides>: New const bool field.
1964 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1965 prefetch_dynamic_strides.
1966 (exynosm1_prefetch_tune): Likewise.
1967 (thunderxt88_prefetch_tune): Likewise.
1968 (thunderx_prefetch_tune): Likewise.
1969 (thunderx2t99_prefetch_tune): Likewise.
1970 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1971 (aarch64_override_options_internal): Update to set
1972 PARAM_PREFETCH_DYNAMIC_STRIDES.
1973 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1974 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1975 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1976 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1977 prefetch-dynamic-strides setting.
1979 2018-05-07 Luis Machado <luis.machado@linaro.org>
1981 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1982 <minimum_stride>: New const int field.
1983 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1984 minimum_stride field.
1985 (exynosm1_prefetch_tune): Likewise.
1986 (thunderxt88_prefetch_tune): Likewise.
1987 (thunderx_prefetch_tune): Likewise.
1988 (thunderx2t99_prefetch_tune): Likewise.
1989 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1990 (aarch64_override_options_internal): Update to set
1991 PARAM_PREFETCH_MINIMUM_STRIDE.
1992 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1993 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1994 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1995 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1996 stride is constant and is below the minimum stride threshold.
1998 2018-05-07 Luis Machado <luis.machado@linaro.org>
2000 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2003 2018-05-07 Luis Machado <luis.machado@linaro.org>
2005 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2006 <prefetch_dynamic_strides>: New const bool field.
2007 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2008 prefetch_dynamic_strides.
2009 (exynosm1_prefetch_tune): Likewise.
2010 (thunderxt88_prefetch_tune): Likewise.
2011 (thunderx_prefetch_tune): Likewise.
2012 (thunderx2t99_prefetch_tune): Likewise.
2013 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2014 (aarch64_override_options_internal): Update to set
2015 PARAM_PREFETCH_DYNAMIC_STRIDES.
2016 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2017 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2018 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2019 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2020 prefetch-dynamic-strides setting.
2022 2018-05-07 Luis Machado <luis.machado@linaro.org>
2024 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2025 <minimum_stride>: New const int field.
2026 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2027 minimum_stride field.
2028 (exynosm1_prefetch_tune): Likewise.
2029 (thunderxt88_prefetch_tune): Likewise.
2030 (thunderx_prefetch_tune): Likewise.
2031 (thunderx2t99_prefetch_tune): Likewise.
2032 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2033 (aarch64_override_options_internal): Update to set
2034 PARAM_PREFETCH_MINIMUM_STRIDE.
2035 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2036 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2037 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2038 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2039 stride is constant and is below the minimum stride threshold.
2041 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2044 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2045 the type is addressable. Don't force op into register if it has
2048 2018-05-05 Roland McGrath <mcgrathr@google.com>
2051 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2052 any section for which we don't know a specific type it should have,
2053 regardless of name. Previously this was done only for the exact
2054 names ".init_array", ".fini_array", and ".preinit_array".
2055 (default_elf_asm_named_section): Add comment about
2056 relationship with default_section_type_flags and SECTION_NOTYPE.
2057 (get_section): Don't consider it a type conflict if one side has
2058 SECTION_NOTYPE and the other doesn't, as long as neither has the
2059 SECTION_BSS et al used in the default_section_type_flags logic.
2061 2018-05-05 Tom de Vries <tom@codesourcery.com>
2064 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2065 (workaround_barsyncs): New function.
2066 (nvptx_reorg): Use workaround_barsyncs.
2067 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2068 (define_expand "nvptx_membar_cta"): New define_expand.
2069 (define_insn "*nvptx_membar_cta"): New insn.
2071 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2073 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2074 To improve optimization opportunities.
2075 * builtin-types.def: The new needed builtin types for the above.
2077 2018-05-04 Richard Biener <rguenther@suse.de>
2079 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2080 * gimple-ssa-store-merging.c
2081 (imm_store_chain_info::output_merged_store): Remove redundant create,
2082 release split_store vector contents on failure.
2083 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2084 scalar stmt vector on cache hit.
2086 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2088 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2090 * config.gcc (powerpc-xilinx-eabi*): Remove.
2091 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2093 (fusion_addis_mem_combo_load): Ditto.
2094 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2096 (rs6000_cpu_cpp_builtins): Ditto.
2097 * config/rs6000/rs6000-linux.c
2098 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2099 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2100 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2102 (rs6000_setup_reg_addr_masks): Ditto.
2103 (rs6000_init_hard_regno_mode_ok): Ditto.
2104 (rs6000_option_override_internal): Ditto.
2105 (legitimate_lo_sum_address_p): Ditto.
2106 (rs6000_legitimize_address): Ditto.
2107 (rs6000_legitimize_reload_address): Ditto.
2108 (rs6000_legitimate_address_p): Ditto.
2109 (abi_v4_pass_in_fpr): Ditto.
2110 (setup_incoming_varargs): Ditto.
2111 (rs6000_gimplify_va_arg): Ditto.
2112 (rs6000_split_multireg_move): Ditto.
2113 (rs6000_savres_strategy): Ditto.
2114 (rs6000_emit_prologue_components): Ditto.
2115 (rs6000_emit_epilogue_components): Ditto.
2116 (rs6000_emit_prologue): Ditto.
2117 (rs6000_emit_epilogue): Ditto.
2118 (rs6000_elf_file_end): Ditto.
2119 (rs6000_function_value): Ditto.
2120 (rs6000_libcall_value): Ditto.
2121 * config/rs6000/rs6000.h: Ditto.
2122 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2123 (TARGET_MINMAX): ... this. New.
2124 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2125 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2126 (*movsi_internal1_single): Delete.
2127 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2128 mfpu=, mxilinx-fpu): Delete.
2129 * config/rs6000/singlefp.h: Delete.
2130 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2131 * config/rs6000/t-rs6000: Ditto.
2132 * config/rs6000/t-xilinx: Delete.
2133 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2134 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2135 (VStype_simple): Delete.
2136 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2137 * config/rs6000/xfpu.h: Delete.
2138 * config/rs6000/xfpu.md: Delete.
2139 * config/rs6000/xilinx.h: Delete.
2140 * config/rs6000/xilinx.opt: Delete.
2141 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2142 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2144 2018-05-04 Tom de Vries <tom@codesourcery.com>
2147 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2150 2018-05-04 Richard Biener <rguenther@suse.de>
2153 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2154 (expand_complex_div_wide): Likewise.
2155 (expand_complex_operations_1): Likewise.
2156 (expand_complex_libcall): Preserve EH info of the original stmt.
2157 (tree_lower_complex): Handle removed blocks.
2158 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2159 on complex multiplication and division libcall builtins.
2161 2018-05-04 Richard Biener <rguenther@suse.de>
2164 * fold-const.c (negate_expr_p): Restrict negation of operand
2165 zero of a division to when we know that can happen without
2167 (fold_negate_expr_1): Likewise.
2169 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2172 * real.h (real_nextafter): Declare.
2173 * real.c (real_nextafter): New function.
2174 * fold-const-call.c (fold_const_nextafter): New function.
2175 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2176 CASE_CFN_NEXTTOWARD.
2177 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2178 even when arg1_mode is different from arg0_mode.
2180 2018-05-03 Nathan Sidwell <nathan@acm.org>
2182 * doc/extend.texi (Deprecated Features): Remove
2184 (Backwards Compatibility): Likewise.
2185 * doc/invoke.texi (C++ Language Options): Likewise.
2186 (C++ Dialect Options): Likewise.
2188 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2191 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2192 _mm512_mask_mullox_epi64): New intrinsics.
2194 2018-05-03 Tom de Vries <tom@codesourcery.com>
2197 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2198 dump files): Add offload-tree.
2200 2018-05-03 Richard Biener <rguenther@suse.de>
2202 PR tree-optimization/85615
2203 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2204 to loops not nested in BBs loop father to avoid creating multi-entry
2207 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2209 PR tree-optimization/70291
2210 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2211 arguments. Change return type to tree. Emit libcall as a new
2212 statement rather than replacing existing one when inplace_p is true.
2213 (expand_complex_multiplication_components): New function.
2214 (expand_complex_multiplication): Expand floating-point complex
2215 multiplication using the above.
2216 (expand_complex_division): Rename inner_type parameter to type.
2217 Update expand_complex_libcall call-site.
2218 (expand_complex_operations_1): Update expand_complex_multiplication
2219 and expand_complex_division call-sites.
2221 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2224 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2225 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2226 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2227 the highest significant bit of the shift count mask is clear. In
2228 check whether and[sq]i3 is needed verify that all significant bits
2229 of the shift count other than the highest are set.
2231 2018-05-02 Tom de Vries <tom@codesourcery.com>
2234 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2235 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2236 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2237 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2238 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2239 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2240 __builtin_goacc_parlevel_size.
2242 2018-05-02 Richard Biener <rguenther@suse.de>
2244 PR tree-optimization/85597
2245 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2246 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2249 2018-05-02 Tom de Vries <tom@codesourcery.com>
2252 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2253 dump files): Add ltrans-tree.
2255 2018-05-02 Tom de Vries <tom@codesourcery.com>
2258 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2259 dump files): Add wpa-ipa.
2261 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2263 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2264 powerpc*-*-linux*paired* target.
2265 * config/rs6000/750cl.h: Delete.
2266 * config/rs6000/paired.h: Delete.
2267 * config/rs6000/paired.md: Delete.
2268 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2270 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2271 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2272 comment. Remove paired float support.
2273 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2274 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2276 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2277 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2279 * config/rs6000/rs6000.c: Remove paired float support.
2280 (paired_expand_vector_init, paired_expand_vector_move,
2281 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2282 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2283 paired_expand_builtin, paired_expand_predicate_builtin,
2284 paired_init_builtins): Delete.
2285 * config/rs6000/rs6000.h: Remove paired float support.
2286 * config/rs6000/rs6000.md: Remove paired float support.
2287 (move_from_CR_ov_bit): Delete.
2288 * config/rs6000/rs6000.opt (mpaired): Delete.
2289 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2290 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2292 2018-05-02 Richard Biener <rguenther@suse.de>
2295 * gimplify.c (gimplify_save_expr): When in SSA form allow
2296 SAVE_EXPRs to compute to SSA vars.
2298 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2301 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2302 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2303 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2304 clobber operands[2], instead use a new pseudo. Formatting fixes.
2306 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2308 PR tree-optimization/85586
2309 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2310 exit early for statements in the same group if the accesses are
2313 2018-05-02 Tom de Vries <tom@codesourcery.com>
2316 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2319 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2321 PR tree-optimization/85143
2322 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2324 2018-05-01 Tom de Vries <tom@codesourcery.com>
2327 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2328 not found" error message.
2330 2018-05-01 Tom de Vries <tom@codesourcery.com>
2333 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2334 * vec.c (test_ordered_remove_if): New function.
2335 (vec_c_tests): Call test_ordered_remove_if.
2336 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2337 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2338 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2339 VEC_ORDERED_REMOVE_IF.
2341 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2343 PR tree-optimization/82665
2344 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2345 pointer subtraction where arguments come from a memchr call.
2347 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2349 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2350 --push-state --as-needed and --pop-state instead of --as-needed and
2351 --no-as-needed if ld supports it.
2352 * configure: Regenerated.
2355 * doc/install.texi2html: Replace _002d with - and _002a with * in
2356 generated html files using sed.
2358 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2361 * gcc-rich-location.c (blank_line_before_p): New function.
2362 (use_new_line): New function.
2363 (gcc_rich_location::add_fixit_insert_formatted): New function.
2364 * gcc-rich-location.h
2365 (gcc_rich_location::add_fixit_insert_formatted): New function.
2367 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2369 * selftest.c (assert_streq): Rename "expected" and "actual" to
2370 "val1" and "val2". Extend NULL-handling to cover both inputs
2371 symmetrically, while still requiring both to be non-NULL for a pass.
2372 * selftest.h (assert_streq): Rename "expected" and "actual" to
2374 (ASSERT_EQ): Likewise.
2375 (ASSERT_EQ_AT): Likewise.
2376 (ASSERT_KNOWN_EQ): Likewise.
2377 (ASSERT_KNOWN_EQ_AT): Likewise.
2378 (ASSERT_NE): Likewise.
2379 (ASSERT_MAYBE_NE): Likewise.
2380 (ASSERT_MAYBE_NE_AT): Likewise.
2381 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2382 the assertion to pass.
2383 (ASSERT_STREQ_AT): Likewise.
2385 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2387 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2388 interaction with -pie.
2390 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2392 * selftest.h: Fix alphabetization of per-source-file selftest
2395 2018-04-30 Jason Merrill <jason@redhat.com>
2397 PR c++/61982 - dead stores to destroyed objects.
2398 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2401 2018-04-30 Jason Merrill <jason@redhat.com>
2403 * tree.c (build_clobber): New.
2404 * tree.h: Declare it.
2405 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2407 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2409 * diagnostic-show-locus.c (layout::layout): Update for
2410 location_get_source_line returning a char_span.
2411 (struct char_span): Move to input.h.
2412 (struct correction): Update for fields in char_span becoming
2414 (struct source_line): Update for location_get_source_line
2415 returning a char_span.
2416 (layout::print_line): Likewise.
2417 * edit-context.c (edited_file::print_content): Likewise.
2418 (edited_file::print_diff_hunk): Likewise.
2419 (edited_file::print_run_of_changed_lines): Likewise.
2420 (edited_file::get_num_lines): Likewise.
2421 (edited_line::edited_line): Likewise.
2422 * final.c (asm_show_source): Likewise.
2423 * input.c (location_get_source_line): Convert return type
2424 from const char * to char_span, losing the final "line_len"
2426 (dump_location_info): Update for the above.
2427 (get_substring_ranges_for_loc): Likewise. Use a char_span
2428 when handling the literal within the line.
2429 (test_reading_source_line): Update for location_get_source_line
2430 returning a char_span.
2431 * input.h (class char_span): Move here from
2432 diagnostic-show-locus.c, converting from a struct to a class.
2433 Make data members private.
2434 (char_span::operator bool): New.
2435 (char_span::length): New.
2436 (char_span::get_buffer): New.
2437 (char_span::operator[]): New.
2438 (char_span::subspan): Make const.
2439 (char_span::xstrdup): New.
2440 (location_get_source_line): Convert return type from const char *
2441 to char_span, losing the final "line_size" param.
2443 2018-04-30 Jan Hubicka <jh@suse.cz>
2445 * lto-wrapper.c (ltrans_priorities): New static var.
2446 (cmp_priority): New.
2447 (run_gcc): Read priorities and if doing parallel build order
2448 the Makefile by them.
2450 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2452 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2454 2018-04-30 Richard Biener <rguenther@suse.de>
2456 * tree-cfg.c (verify_address): Remove base argument, add
2457 flag whether to check TREE_ADDRESSABLE and do that.
2458 (verify_expr): Remove.
2459 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2460 (verify_gimple_assign_single): Likewise.
2461 (verify_gimple_switch): Likewise.
2462 (verify_expr_location_1): Dereference tp once. Add (disabled)
2463 piece from verify_expr.
2464 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2466 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2468 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2470 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2472 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2473 (small_data_pattern): Likewise.
2474 (arc_rewrite_small_data): Likewise.
2475 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2476 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2477 (get_symbol_alignment): New function.
2478 (legitimate_small_data_address_p): Likewise.
2479 (legitimate_scaled_address): Update, call
2480 legitimate_small_data_address_p.
2481 (output_sdata): New static variable.
2482 (arc_print_operand): Update how we handle small data operands.
2483 (arc_print_operand_address): Likewise.
2484 (arc_legitimate_address_p): Update, use
2485 legitimate_small_data_address_p.
2486 (arc_rewrite_small_data_p): Remove.
2487 (arc_rewrite_small_data_1): Likewise.
2488 (arc_rewrite_small_data): Likewise.
2489 (small_data_pattern): Likewise.
2490 (compact_sda_memory_operand): Update to use
2491 legitimate_small_data_address_p and get_symbol_alignment.
2492 (prepare_move_operands): Don't rewite sdata pattern.
2493 (prepare_extend_operands): Remove.
2494 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2496 (zero_extendqisi2): Likewise.
2497 (zero_extendhisi2): Likewise.
2498 (extendqihi2): Likewise.
2499 (extendqisi2): Likewise.
2500 (extendhisi2): Likewise.
2504 * config/arc/constraints.md (Usd): Change it to memory constraint.
2506 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2508 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2509 as source of std instructions.
2510 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2511 allow 6-bit constants as source for store instructions.
2512 (movdi_insn): Update instruction pattern to allow 6-bit constants
2513 as source for store instructions.
2515 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2517 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2519 2018-04-30 Nathan Sidwell <nathan@acm.org>
2520 Sandra Loosemore <sandra@codesourcery.com>
2522 * dumpfile.c (dump_open): Allow '-' for stdout.
2523 * doc/invoke.texi (Developer Options): Document dump filename
2524 determination early. Document stdin/stdout selection.
2526 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2528 Microblaze Target: PIC data text relative
2530 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2531 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2533 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2534 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2535 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2536 New addressing mode for data-text relative position indepenedent code.
2537 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2538 'ADDRESS_SYMBOLIC_TXT_REL'.
2539 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2540 (microblaze_legitimate_pic_operand): Exclude function calls from
2541 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2542 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2544 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2545 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2546 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2547 for 'address + offset'.
2548 (microblaze_expand_prologue): Add new function prologue call for
2550 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2551 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2552 table in case of TARGET_PIC_DATA_TEXT_REL.
2553 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2554 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2555 Add new macros 'UNSPEC_TEXT',
2556 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2557 + exclude function calls from 'UNSPEC_PLT' in case of data text
2559 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2560 new target hook for generating address diff vector tables in case of
2562 * doc/tm.texi : Regenerate.
2563 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2564 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2565 of addr diff vector generation.
2566 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2567 target hook definition.
2568 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2569 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2570 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2571 option for data text relative.
2573 2018-04-30 Richard Biener <rguenther@suse.de>
2575 * tree-chrec.h (evolution_function_is_constant_p): Remove
2577 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2579 2018-04-30 Richard Biener <rguenther@suse.de>
2582 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2584 2018-04-30 Richard Biener <rguenther@suse.de>
2586 PR tree-optimization/28364
2587 PR tree-optimization/85275
2588 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2589 copying first exit test.
2591 2018-04-28 Mark Wielaard <mark@klomp.org>
2593 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2595 (dwarf_AT): Handle DW_AT_addr_base.
2596 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2598 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2601 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2602 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2603 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2604 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2606 2018-04-28 Richard Biener <rguenther@suse.de>
2608 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2609 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2610 to reflect use. Only add interesting stmts.
2612 2018-04-27 Martin Jambor <mjambor@suse.cz>
2615 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2616 the jump function allows for passing through aggregate values.
2618 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2620 * input.h (in_system_header_at): Convert from macro to inline
2622 (from_macro_expansion_at): Likewise.
2623 (from_macro_definition_at): Likewise.
2625 2018-04-27 Jeff Law <law@redhat.com>
2627 * config.gcc: Mark tile* targets as deprecated/obsolete.
2629 2018-04-27 Richard Biener <rguenther@suse.de>
2631 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2634 2018-04-27 Richard Biener <rguenther@suse.de>
2636 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2638 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2640 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2641 with Yd constraint. Set "preferred_for_speed" attribute from
2642 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2644 (*movdi_internal): Ditto.
2645 (movti_interunit splitters): Remove
2646 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2647 (movdi_interunit splitters): Ditto.
2648 * config/i386/constraints.md (Ye): Remove.
2649 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2651 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2654 * config/aarch64/constraints.md (Usg): Limit to 31.
2657 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2659 PR tree-optimization/85529
2660 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2661 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2662 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2663 zero extension or masking of the MSB bit.
2664 (optimize_range_tests): Add FIRST_BB argument, pass it through
2665 to optimize_range_tests_var_bound.
2666 (maybe_optimize_range_tests, reassociate_bb): Adjust
2667 optimize_range_tests callers.
2669 2018-04-26 Richard Biener <rguenther@suse.de>
2670 Jakub Jelinek <jakub@redhat.com>
2672 * cgraph.h (symbol_table): Just declare debug method here.
2673 * symtab.c (symbol_table::debug): Define.
2675 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2677 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2679 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2681 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2682 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2683 (*movdi_internal): Substitute Yi and Yj constraint with x
2684 and Ym and Yn constraint with y constraint. Update "isa"
2685 attribute and set "preferred_for_speed" attribute from
2686 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2687 (*movsi_internal): Ditto.
2688 (*movdf_internal): Ditto.
2689 (*movsf_internal): Ditto.
2690 (*zero_extendsidi2): Ditto.
2691 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2692 (sse2_loadld): Ditto.
2693 (*vec_extract<ssevecmodelower>_0): Ditto.
2694 (*vec_extractv4si_0_zext_sse4): Ditto.
2695 (vec_concatv2di): Ditto.
2696 (*vec_dup<mode>): Ditto.
2697 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2698 * config/i386/constraints.md (Yi): Remove.
2703 2018-04-26 Nathan Sidwell <nathan@acm.org>
2705 * dumpfile.c (dump_open): New.
2706 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2707 (dump_finish): Detect stdio/stderr by value not name.
2709 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2711 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2713 2018-04-26 Tom de Vries <tom@codesourcery.com>
2716 * config/nvptx/nvptx.c (verify_neutering_jumps)
2717 (verify_neutering_labels): New function
2718 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2720 2018-04-26 Tom de Vries <tom@codesourcery.com>
2723 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2724 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2727 2018-04-26 Richard Biener <rguenther@suse.de>
2728 Tom de Vries <tom@codesourcery.com>
2731 * lto-streamer-out.c (output_function): Fixup loops if required to match
2732 discovery done in the reader.
2734 2018-04-26 Richard Biener <rguenther@suse.de>
2736 PR tree-optimization/85116
2737 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2738 have a loop exit from the single latch predecessor. Remove
2739 case of header with just condition.
2740 (ch_base::copy_headers): Exclude infinite loops from any
2742 (pass_ch::execute): Record exits.
2744 2018-04-26 Richard Biener <rguenther@suse.de>
2746 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2747 prologue cost vector and pass it to vect_get_load_cost.
2748 (vect_get_peeling_costs_all_drs): Likewise.
2749 (vect_peeling_hash_get_lowest_cost): Likewise.
2750 (vect_enhance_data_refs_alignment): Likewise.
2752 2018-04-26 Richard Biener <rguenther@suse.de>
2755 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2756 checking of integer<->pointer conversions.
2757 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2758 sign-/zero-extending pointer types.
2759 (expand_omp_for_static_chunk): Likewise.
2761 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2762 Jean Lee <xiaoyur347@gmail.com>
2764 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2765 (TARGET_ASAN_SHADOW_OFFSET): Define.
2766 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2767 true for -fsanitize=address.
2769 2018-04-25 Mark Wielaard <mark@klomp.org>
2771 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2774 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2776 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2777 than "alu", remove explicit "memory" and "imm_disp" attributes.
2778 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2781 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2782 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2783 gen_lowpart_no_emit.
2785 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2788 * config/i386/i386.c (ix86_expand_builtin): Change memory
2789 operand to XI, extend p0 to Pmode.
2790 * config/i386/i386.md: Change unspec volatile and operand
2791 1 mode to XI, change operand 0 mode to P.
2793 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2795 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2796 GET_MODE_MASK before any checking.
2797 (nds32_can_use_bset_p): Likewise.
2798 (nds32_can_use_btgl_p): Likewise.
2800 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2802 * config/nds32/nds32-doubleword.md: New define_split pattern for
2803 illegal register number.
2805 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2807 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2809 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2811 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2813 2018-04-25 Richard Biener <rguenther@suse.de>
2815 * lto-streamer.h (LTO_major_version): Bump to 8.
2817 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2819 * BASE-VER: Set to 9.0.0.
2821 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2823 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2824 in __abskf2 and __powikf2.
2826 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2829 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2830 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2831 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2832 Use the above on operand 2. Reindent.
2833 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2835 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2838 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2840 * config/i386/i386.opt (mcet): Removed.
2841 * doc/install.texi: Remove -mcet documentation.
2842 * doc/invoke.texi: Likewise.
2844 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2847 * doc/install.texi: Remove -mcet from bootstrap-cet.
2849 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2852 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2853 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2857 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2858 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2859 containing a CONST_VECTOR.
2861 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2863 * doc/install.texi: Update newlib dependency for nvptx.
2865 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2868 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2869 instead of INTVAL when shifting x left.
2871 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2873 PR tree-optimization/85478
2874 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2875 vect_grouped_store_supported for single element vectors.
2877 2018-04-24 Richard Biener <rguenther@suse.de>
2880 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2881 load cost increase to the case of non-constant step.
2883 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2886 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2887 destination if any_malformed_asm.
2889 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2892 * expr.c (store_field): In the bitfield case, if the value comes from
2893 a function call and is returned in registers by means of a PARALLEL,
2894 do not change the mode of the temporary unless BLKmode and VOIDmode.
2896 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2898 PR rtl-optimization/85423
2899 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2900 dependencies to debug insns when the previous insn is non-debug.
2902 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2904 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2905 enums into a single definition.
2906 (fls): Fix predicates and printing.
2909 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2911 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2912 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2913 and short u6 immediate.
2914 (check_if_valid_sleep_operand): Remove.
2915 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2917 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2919 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2920 flag_always_save_lp condition.
2921 * config/nds32/nds32.opt (malways-save-lp): New option.
2923 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2925 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2926 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2927 * config/nds32/nds32.h
2928 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2929 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2931 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2933 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2934 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2936 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2937 Chung-Ju Wu <jasonwucj@gmail.com>
2939 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2941 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2942 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2944 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2946 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2948 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2950 * config/nds32/nds32-protos.h (nds32_data_alignment,
2951 nds32_local_alignment): Declare.
2952 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2953 nds32_local_alignment): New functions.
2954 (TARGET_CONSTANT_ALIGNMENT): Define.
2955 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2957 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2959 * config/nds32/nds32.c
2960 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2961 (TARGET_MODES_TIEABLE_P): Likewise.
2963 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2965 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2968 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2969 Chung-Ju Wu <jasonwucj@gmail.com>
2971 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2972 for unaligned access.
2973 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2975 * config/nds32/nds32-intrinsic.md: Likewise.
2976 * config/nds32/nds32_intrinsic.h: Likewise.
2977 * config/nds32/nds32.h (nds32_builtins): Likewise.
2978 * config/nds32/nds32.opt (munaligned-access): New option.
2979 * config/nds32/nds32.c (nds32_asm_file_start): Display
2980 flag_unaligned_access status.
2982 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2984 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2985 -mno-relax is present.
2986 * config/riscv/linux.h (LINK_SPEC): Ditto.
2988 2018-04-20 Martin Sebor <msebor@redhat.com>
2991 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2993 (gimple_fold_builtin_stxcpy_chk): Same.
2994 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2996 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2999 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3000 __powikf2 when long double is IEEE 128-bit.
3002 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3004 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3005 step to make sure stack always aligned.
3007 2018-04-20 Carl Love <cel@us.ibm.com>
3010 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3011 size check for arg0.
3013 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3014 Tom de Vries <tom@codesourcery.com>
3017 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3018 Emit insns for calls too.
3019 (nvptx_find_par): Always look for worker-level predecessor insn.
3020 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3022 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3023 (nvptx_process_pars): Propagate frames for calls.
3025 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3028 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3030 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3031 (ix86_handle_option): Don't handle OPT_mibt.
3032 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3034 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3036 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3037 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3038 (ix86_target_macros): Define __CET__ with flag_cf_protection
3039 for -fcf-protection.
3040 * config/i386/i386.c (isa2_opts): Remove -mibt.
3041 * config/i386/i386.h (TARGET_IBT): Removed.
3042 (TARGET_IBT_P): Likewise.
3043 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3044 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3045 * config/i386/i386.opt (mcet): Update help message.
3048 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3049 -mcet as an alias for -mshstk.
3051 2018-04-20 Richard Biener <rguenther@suse.de>
3054 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3055 complexity by forcing a single use of the multiply operand.
3057 2018-04-20 Martin Jambor <mjambor@suse.cz>
3060 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3061 recursion dependency to only apply to non-clones.
3063 2018-04-20 Martin Jambor <mjambor@suse.cz>
3066 * ipa-cp.c (create_specialized_node): Check that clones of
3067 self-recursive edges exist during IPA-CP.
3069 2018-04-19 Toon Moene <toon@moene.org>
3071 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3074 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3076 PR tree-optimization/85467
3077 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3078 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3079 VECTOR_CST element to type.
3081 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3084 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3085 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3086 (builtin_longjmp): Likewise.
3087 (save_stack_nonlocal): New pattern.
3088 (restore_stack_nonlocal): Likewise.
3090 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3093 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3094 Replace ASM_OUTPUT_LABEL with fprintf.
3096 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3099 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3100 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3101 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3102 define __IBT__ and __SHSTK__ for -fcf-protection.
3103 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3105 (ix86_trampoline_init): Likewise.
3106 (x86_output_mi_thunk): Likewise.
3107 (ix86_notrack_prefixed_insn_p): Likewise.
3108 (ix86_option_override_internal): Don't disallow -fcf-protection.
3109 * config/i386/i386.md (rdssp<mode>): Also enable for
3111 (incssp<mode>): Likewise.
3112 (nop_endbr): Likewise.
3113 * config/i386/i386.opt (mcet): Change help message to built-in
3117 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3118 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3119 enable CET built-in functions.
3121 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3123 * common/config/i386/i386-common.c
3124 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3125 OPTION_MASK_ISA_MOVDIRI_UNSET,
3126 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3127 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3128 * config.gcc (movdirintrin.h): New header.
3129 * config/i386/cpuid.h (bit_MOVDIRI,
3130 bit_MOVDIR64B): New bits.
3131 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3133 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3134 (VOID, PVOID, PCVOID)): New function types.
3135 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3136 __builtin_ia32_directstoreu_u64,
3137 __builtin_ia32_movdir64b): New builtins.
3138 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3139 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3141 (ix86_valid_target_attribute_inner_p): Ditto.
3142 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3143 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3144 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3145 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3146 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3147 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3148 (movdiri<mode>, movdir64b_<mode>): New.
3149 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3150 * config/i386/immintrin.h: Include movdirintrin.h.
3151 * config/i386/movdirintrin.h: New file.
3152 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3154 2018-04-19 Richard Biener <rguenther@suse.de>
3157 * cfg.c (clear_bb_flags): When loop state says we have
3158 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3160 2018-04-19 Richard Biener <rguenther@suse.de>
3162 PR tree-optimization/84737
3163 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3164 copying restrict info.
3165 (vect_setup_realignment): Use it.
3166 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3167 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3168 the first DR to all generated stores.
3169 (vectorizable_load): Likewise for loads.
3171 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3173 PR tree-optimization/85446
3174 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3175 the integral and pointer types to have the same precision.
3177 * doc/install.texi: Document --disable-cet being the default and
3180 2018-04-18 Martin Liska <mliska@suse.cz>
3182 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3185 2018-04-18 Martin Liska <mliska@suse.cz>
3188 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3191 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3192 arguments if they are comparable.
3194 2018-04-18 Martin Liska <mliska@suse.cz>
3197 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3200 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3203 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3206 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3207 ENDBR after calling __morestack.
3209 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3212 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3213 by using gcc_base_ver to generate a gcc_driver_version, and use
3214 it when generating GCC_DRIVER_NAME.
3215 * configure: Regenerate.
3217 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3220 * config.gcc: Obsolete powerpc*-*-*spe*.
3222 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3225 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3226 (stabstr_D): Change type of unum from unsigned int to
3227 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3230 2018-04-17 Jim Wilson <jimw@sifive.com>
3233 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3234 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3235 Set arg_pointer_offset after using pretend_args_size.
3237 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3239 PR rtl-optimization/85431
3240 * dse.c (record_store): Ignore zero width stores.
3243 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3244 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3245 __builtin_stack_restore rather than after it.
3246 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3247 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3248 argument instead of virtual_dynamic_stack_rtx.
3250 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3252 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3254 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3255 Add note to error message to explain internal mapping of overloaded
3256 built-in function name to non-overloaded built-in function name.
3257 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3260 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3263 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3264 where the inputs overlap with the output.
3266 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3268 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3269 (=v, v) alternative and explicit "memory" attribute.
3270 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3271 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3273 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3274 "sselog1" type instead of "sselog".
3275 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3276 "sselog". Remove explicit "memory" attribute.
3277 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3278 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3280 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3281 "sselog1" type instead of "sselog". Remove explicit "memory"
3283 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3284 use "sselog1" type instead of "sselog". Remove explicit "memory"
3286 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3287 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3289 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3290 "sselog1" type instead of "sselog". Remove explicit "memory"
3292 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3293 use "sselog1" type instead of "sselog". Remove explicit "memory"
3297 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3300 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3303 2018-04-17 Martin Jambor <mjambor@suse.cz>
3306 * ipa-cp.c (create_specialized_node): Call
3307 expand_all_artificial_thunks if necessary.
3309 2018-04-17 Martin Liska <mliska@suse.cz>
3312 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3313 in message, remote space in between '_G' and '('.
3315 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3318 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3319 avx512f_vmcmp<mode>3<round_saeonly_name>,
3320 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3321 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3322 avx512f_rndscale<mode><round_saeonly_name>,
3323 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3324 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3325 Use %<iptr>2 instead of %2 for -masm=intel.
3326 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3327 avx512f_vcvttss2usi<round_saeonly_name>,
3328 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3330 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3331 avx512f_vcvttsd2usi<round_saeonly_name>,
3332 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3333 Use %q1 instead of %1 for -masm=intel.
3334 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3335 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3336 of %3 for -masm=intel.
3337 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3339 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3341 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3343 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3344 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3346 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3348 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3349 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3350 %g1 and one with %0 and %1.
3351 (avx512er_vmrcp28<mode><round_saeonly_name>,
3352 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3354 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3355 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3356 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3357 of %0 and %{%4%} for -masm=intel.
3358 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3359 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3360 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3361 order of %0 and %{%5%}%{z%} for -masm=intel.
3363 2018-04-17 Jan Hubicka <jh@suse.cz>
3366 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3368 2018-04-17 Martin Liska <mliska@suse.cz>
3371 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3372 for target_clone error message. Make default implementation
3373 clone to be a local declaration.
3374 (separate_attrs): Add new argument and check for an empty
3376 (expand_target_clones): Handle it.
3377 (ipa_target_clone): Make redirection just for target_clones
3380 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3381 Tom de Vries <tom@codesourcery.com>
3384 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3385 tiled basic blocks without omp continue statements.
3387 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3390 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3391 vec_extract expression as having side effects to make sure it gets
3394 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3397 * config/i386/i386.c (get_builtin_code_for_version): Check
3400 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3403 * gcc/config.gcc: Support "skylake".
3404 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3406 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3407 (processor_target_table): Add "skylake".
3408 (ix86_option_override_internal): Add "skylake".
3409 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3410 PROCESSOR_CANNONLAKE.
3411 (get_builtin_code_for_version): Fix priority for
3412 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3413 PROCESSOR_SKYLAKE-AVX512.
3414 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3415 (processor_type): Add PROCESSOR_SKYLAKE.
3417 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3418 Jason Merrill <jason@redhat.com>
3421 * convert.c (convert_to_integer_1): Use direct recursion for
3422 enumeral types and types with a precision less than the number
3423 of bits in their mode.
3425 2018-04-16 Julia Koval <julia.koval@intel.com>
3428 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3429 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3431 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3434 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3435 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3436 and -mno-direct-move.
3438 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3441 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3442 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3443 Ensure negative shifts result in {0}.
3445 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3447 PR rtl-optimization/79916
3448 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3449 regs (if any) to define how to gnerate SD moves when LRA is in
3452 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3454 PR rtl-optimization/85393
3455 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3456 * except.c (expand_dw2_landing_pad_for_region): Make static.
3457 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3458 a label and unconditional jump to old_bb, rather than
3459 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3462 PR rtl-optimization/85376
3463 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3464 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3465 instead of a specific value.
3467 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3468 Bin Cheng <bin.cheng@arm.com>
3470 PR tree-optimization/82965
3471 PR tree-optimization/83991
3472 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3473 by_profile_only parameter.
3474 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3475 information if the loop was predicted to iterate too many times.
3476 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3478 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3481 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3484 2018-04-13 Martin Liska <mliska@suse.cz>
3485 Jakub Jelinek <jakub@redhat.com>
3488 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3489 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3490 * builtins.c (expand_builtin_memory_copy_args): Use
3491 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3492 handle dest_addr == pc_rtx.
3494 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3497 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3498 asked to not generate direct moves.
3499 (fix_trunc<mode>si2_stfiwx): Similar.
3500 (fix_trunc<mode>si2_internal): Similar.
3502 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3505 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3506 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3507 lookup if dest in some wider mode is known to be const0_rtx and
3508 if so, record permanent equivalence for it to be ZERO_EXTEND of
3509 the narrower mode destination.
3511 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3513 * lto-streamer-out.c (output_function): Revert 259346.
3514 * omp-expand.c (expand_oacc_for): Likewise.
3516 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3518 PR rtl-optimization/85354
3519 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3520 * sel-sched.c (sel_global_init): ... here.
3522 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3525 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3526 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3527 mode for PE-COFF targets.
3528 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3529 (i386_pe_asm_lto_end): Likewise.
3530 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3531 (TARGET_ASM_LTO_END): Likewise.
3532 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3533 (i386_pe_asm_lto_start): New function.
3534 (i386_pe_asm_lto_end): Likewise.
3536 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3537 Richard Biener <rguenther@suse.de>
3540 * lto-streamer-out.c (output_function): Fix CFG loop state before
3542 * omp-expand.c (expand_oacc_for): Handle calls to internal
3543 functions like regular functions.
3545 2018-04-12 Richard Biener <rguenther@suse.de>
3548 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3549 for the early LTO debug to properly generate references to it
3550 during DIE emission. Do not re-use that for the skeleton for
3552 (dwarf2out_early_finish): Likewise.
3554 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3557 * config/i386/sse.md
3558 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3559 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3560 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3561 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3562 and output is a reg, avoid creating invalid lowpart subreg, but
3563 instead split into a 512-bit move. Don't split if not AVX512VL,
3564 input is xmm16+ reg and output is a mem.
3565 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3566 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3567 xmm16+ reg and output is a mem.
3569 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3571 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3572 also for flag_dwarf2_cfi_asm.
3574 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3576 PR rtl-optimization/85342
3577 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3578 a bool scalar var inside of the loop instead. Don't try to update
3579 recog_data.operand after failed apply_change_group.
3581 2018-04-12 Tom de Vries <tom@codesourcery.com>
3584 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3585 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3586 array with flexible array member as array without given dimension.
3587 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3588 argument for undefined param to true.
3590 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3593 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3594 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3595 from PowerPC section.
3596 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3597 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3598 help text that is too long.
3599 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3600 help text that is too long.
3601 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3602 help text that is too long.
3604 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3606 * config/alpha/alpha.md (stack_probe_internal): Rename
3607 from "probe_stack". Update all callers.
3609 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3611 PR rtl-optimization/84566
3612 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3613 sched_macro_fuse_insns.
3615 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3618 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3619 (compute_block_dependences): ... from here.
3621 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3623 PR tree-optimization/85331
3624 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3625 from int to HOST_WIDE_INT.
3627 2018-04-11 Martin Jambor <mjambor@suse.cz>
3630 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3631 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3632 not the same as the source val.
3633 (cgraph_edge_brings_value_p): New parameter.
3634 (gather_edges_for_value): Pass destination value to
3635 cgraph_edge_brings_value_p.
3636 (perhaps_add_new_callers): Likewise.
3637 (get_info_about_necessary_edges): Likewise and exclude values brought
3638 only by self-recursive edges.
3639 (create_specialized_node): Redirect only clones of self-calling edges.
3640 (+self_recursive_pass_through_p): New function.
3641 (find_more_scalar_values_for_callers_subset): Use it.
3642 (find_aggregate_values_for_callers_subset): Likewise.
3643 (known_aggs_to_agg_replacement_list): Removed.
3644 (decide_whether_version_node): Re-calculate known constants for all
3645 remaining context clones.
3647 2018-04-11 Richard Biener <rguenther@suse.de>
3650 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3651 from early DWARF output.
3652 (dwarf2out_early_finish): Output line info unconditionally into
3653 early DWARF and add reference to it.
3655 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3658 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3659 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3660 other than V2DFmode using iptr mode attribute.
3661 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3663 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3665 PR rtl-optimization/84659
3666 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3668 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3671 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3673 (output_loc_list): Pass address of a dummy size variable even in the
3674 locview handling loop.
3675 (index_location_lists): Add comment on why skip_loc_list_entry can't
3678 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3681 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3684 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3687 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3688 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3689 and -mstring-compare-inline-limit.
3691 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3694 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3695 for stack clash protection in a register whenever we need it to be in
3698 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3700 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3701 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3703 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3706 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3708 (mlong-double-): Ditto.
3709 * config/rs6000/sysv4.opt (msdata=): Ditto.
3710 (mtls-size=): Ditto.
3712 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3714 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3715 erroneous entries for
3716 "vector int vec_ldl (int, long int *)", and
3717 "vector unsigned int vec_ldl (int, unsigned long int *)".
3718 Add comments and entries for
3719 "vector bool char vec_ldl (int, bool char *)",
3720 "vector bool short vec_ldl (int, bool short *)",
3721 "vector bool int vec_ldl (int, bool int *)",
3722 "vector bool long long vec_ldl (int, bool long long *)",
3723 "vector pixel vec_ldl (int, pixel *)",
3724 "vector long long vec_ldl (int, long long *)",
3725 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3726 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3727 type tree bool_long_long_type_node and correct definition of
3728 bool_V2DI_type_node to make reference to this new type tree.
3729 (rs6000_mangle_type): Replace erroneous reference to
3730 bool_long_type_node with bool_long_long_type_node.
3731 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3732 comments to emphasize sign distinctions for char and int types and
3733 replace RS6000_BTI_bool_long constant with
3734 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3735 use of RS6000_BTI_pixel.
3736 (bool_long_type_node): Remove this macro definition.
3737 (bool_long_long_type_node): New macro definition
3739 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3741 PR rtl-optimization/85300
3742 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3743 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3744 simplify_unary_operation fails.
3746 2018-04-10 Martin Liska <mliska@suse.cz>
3748 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3749 cgraph_edge and ipa_ref.
3751 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3755 * config/i386/sse.md
3756 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3757 computation of the VEC_MERGE selector from mask.
3758 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3759 Fix decoding of the VEC_MERGE selector into mask.
3761 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3763 PR tree-optimization/85286
3764 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3766 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3768 * final.c (final_1): Set insn_last_address as well as
3769 insn_current_address.
3771 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3774 * explow.c (emit_stack_probe): Call validize_mem on memory location
3775 before passing it to gen_probe_stack. Create address operand and
3776 legitimize it for the probe_stack_address case.
3778 2018-04-09 Jan Hubicka <jh@suse.cz>
3781 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3782 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3783 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3785 (free_lang_data): Rebuild type inheritance graph.
3787 2018-04-09 Martin Sebor <msebor@redhat.com>
3789 * invoke.texi (-finline-small-functions): Mention other optimization
3791 (-findirect-inlining, -fpartial-inlining): Same.
3792 (-finline-functions-called-once): Same.
3793 (-freorder-blocks-and-partition): Same.
3795 2018-04-09 Jan Hubicka <jh@suse.cz>
3798 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3799 jumps; choose last target that matches the criteria (i.e.
3800 no partition changes for non-crossing jumps).
3801 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3802 support for redirecting crossing jumps to non-crossing.
3804 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3806 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3807 also for naked functions.
3809 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3811 * config/arc/arc.md (add_shift): New pattern.
3812 (add_shift2): Likewise.
3813 (sub_shift): Likewise.
3814 (sub_shift_cmp0_noout): Likewise.
3815 (compare_si_ashiftsi): Likewise.
3816 (xbfu_cmp0_noout): New combine pattern.
3817 (xbfu_cmp0"): Likewise.
3818 (movsi_set_cc_insn): Place the predicable variant first.
3819 (commutative_binary_cmp0_noout): Remove clobber.
3820 (commutative_binary_cmp0): New pattern.
3821 (noncommutative_binary_cmp0): Likewise.
3822 (noncommutative_binary_cmp0_noout): Likewise.
3823 (noncommutative_binary_comparison_result_used): Removed.
3824 (rsub_cmp0): New pattern.
3825 (rsub_cmp0_noout): Likewise.
3826 (extzvsi): Changed, keep only meaningful variants.
3827 (SQH, SEZ): New iterators.
3828 (SQH_postfix): New mode attribute.
3829 (SEZ_prefix): New code attribute.
3830 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3831 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3832 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3834 (noncommutative_operator): Check the availability of barrel
3837 2018-04-09 Richard Biener <rguenther@suse.de>
3839 PR tree-optimization/85284
3840 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3841 Only use the niter constraining form of simple_iv when the exit
3844 2018-04-09 Tom de Vries <tom@codesourcery.com>
3847 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3848 (define_expand "*memory_barrier"): New define_expand.
3849 (define_insn "memory_barrier"): New insn.
3851 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3853 PR rtl-optimization/80463
3854 PR rtl-optimization/83972
3855 PR rtl-optimization/83480
3857 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3858 correct producer for the insn.
3859 (tidy_control_flow): Fixup seqnos in case of debug insns.
3861 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3863 PR rtl-optimization/83913
3865 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3866 different sched-times when merging exprs.
3868 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3870 PR rtl-optimization/83962
3872 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3873 tidy_fallthru_edge and tidy_control_flow.
3875 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3877 PR rtl-optimization/83530
3879 * sel-sched.c (force_next_insn): New global variable.
3880 (remove_insn_for_debug): When force_next_insn is true, also leave only
3881 next insn in the ready list.
3882 (sel_sched_region): When the region wasn't scheduled, make another pass
3883 over it with force_next_insn set to 1.
3885 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3887 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3889 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3890 for interrupt control.
3891 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3892 functions for interrupt control.
3893 * config/nds32/nds32-intrinsic.md: Likewise.
3894 * config/nds32/nds32_intrinsic.h: Likewise.
3895 * config/nds32/nds32.h (nds32_builtins): Likewise.
3897 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3899 * config/nds32/nds32.c (nds32_init_machine_status,
3900 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3901 strict_aligned_p field.
3902 (nds32_expand_to_rtl_hook): New function.
3903 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3904 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3906 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3907 Chung-Ju Wu <jasonwucj@gmail.com>
3909 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3910 * config/nds32/nds32-n7.md: New file.
3911 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3912 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3914 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3915 * config/nds32/nds32.md (pipeline_model): Add n7.
3916 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3917 * config/nds32/pipelines.md: Include n7 settings.
3919 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3920 Chung-Ju Wu <jasonwucj@gmail.com>
3922 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3923 * config/nds32/nds32-e8.md: New file.
3924 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3925 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3927 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3928 * config/nds32/nds32.md (pipeline_model): Add e8.
3929 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3930 * config/nds32/pipelines.md: Include e8 settings.
3932 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3933 Chung-Ju Wu <jasonwucj@gmail.com>
3935 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3936 * config/nds32/nds32-n8.md: New file.
3937 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3938 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3940 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3941 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3942 * config/nds32/nds32.md (pipeline_model): Add n8.
3943 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3944 * config/nds32/pipelines.md: Include n8 settings.
3946 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3947 Chung-Ju Wu <jasonwucj@gmail.com>
3949 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3950 * config/nds32/nds32-n9-2r1w.md: New file.
3951 * config/nds32/nds32-n9-3r2w.md: New file.
3952 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3953 nds32_register_ports): New or modify for cpu n9.
3954 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3956 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3957 * config/nds32/nds32-utils.c: New file.
3958 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3959 TARGET_MUL_SLOW): Define.
3960 * config/nds32/nds32.md (pipeline_model): New attribute.
3961 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3962 New options that support cpu n9.
3963 * config/nds32/pipelines.md: Include n9 settings.
3964 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3966 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3968 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3969 information if necessary.
3970 (output_cond_branch_compare_zero): Likewise.
3971 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3972 (nds32_target_alignment): Refine for alignment.
3973 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3974 (FUNCTION_BOUNDARY): Modify.
3975 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3977 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3979 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3981 * config/nds32/constants.md (unspec_volatile_element): Add values for
3982 TLB operation and data prefetch.
3983 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3984 functions for TLB operation and data prefetch.
3985 * config/nds32/nds32-intrinsic.md: Likewise.
3986 * config/nds32/nds32_intrinsic.h: Likewise.
3987 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3988 (nds32_print_operand): Likewise.
3989 * config/nds32/nds32.h (nds32_builtins): Likewise.
3991 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3992 Andrew Pinski <pinsika@gcc.gnu.org>
3995 * match.pd: Use constant_boolean_node of correct type instead of
3996 boolean_true_node or boolean_false_node for simplifying
3997 pointer comparisons to zero.
3999 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4001 PR tree-optimization/80021
4002 * tree.c (verify_type_variant): Make error call in verify_variant_match
4003 translatable and remove final full stop.
4005 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4007 * config/nds32/constants.md (unspec_volatile_element): Add
4008 UNSPEC_VOLATILE_EH_RETURN.
4009 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4010 nds32_output_stack_pop): Support dwarf exception handling process.
4011 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4012 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4013 exception handling process.
4014 (nds32_compute_stack_frame): Likewise.
4015 (nds32_return_addr_rtx): Likewise.
4016 (nds32_initial_elimination_offset): Likewise.
4017 (nds32_expand_prologue): Likewise.
4018 (nds32_expand_epilogue): Likewise.
4019 (nds32_dynamic_chain_address): New function.
4020 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4022 (DYNAMIC_CHAIN_ADDRESS): Define.
4023 (EH_RETURN_DATA_REGNO): Define.
4024 (EH_RETURN_STACKADJ_RTX): Define.
4025 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4026 patterns for dwarf exception handling.
4028 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4030 * config/nds32/nds32.h: Clean up obsolete macros.
4032 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4034 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4035 Add enum values for particular instructions.
4036 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4037 particular intrinsic functions.
4038 * config/nds32/nds32-intrinsic.md: Likewise.
4039 * config/nds32/nds32_intrinsic.h: Likewise.
4040 * config/nds32/nds32.h (nds32_builtins): Likewise.
4041 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4042 (btst, ave): New patterns for particular instructions.
4044 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4046 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4047 Add enum values for atomic load/store and memory sync.
4048 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4050 * config/nds32/nds32-intrinsic.md: Likewise.
4051 * config/nds32/nds32_intrinsic.h: Likewise.
4052 * config/nds32/nds32.h (nds32_builtins): Likewise.
4054 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4056 PR tree-optimization/85257
4057 * fold-const.c (native_encode_vector): If not all elts could fit
4058 and off is -1, return 0 rather than offset.
4059 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4060 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4061 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4062 adjust buffer in native_interpret_expr call.
4064 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4066 * config/nds32/constants.md (unspec_volatile_element): Add cache
4067 control enum values.
4068 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4069 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4070 * config/nds32/nds32.c (nds32_cctl_names): New.
4071 (nds32_print_operand): Handle cache control register names.
4072 * config/nds32/nds32.h (nds32_builtins): New enum values.
4073 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4075 * config/nds32/nds32.md (type): Add mmu.
4076 * config/nds32/pipelines.md (simple_insn): Add mmu.
4078 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4080 * config/nds32/nds32.md (type): Remove call.
4081 * config/nds32/pipelines.md (simple_insn): Likewise.
4083 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4085 * config/nds32/constants.md (unspec_volatile_element): Add
4086 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4087 UNSPEC_VOLATILE_FMFCFG.
4088 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4089 description for fmfcfg and fmfcsr.
4090 (bdesc_1arg): Add fmtcsr.
4091 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4092 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4093 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4094 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4095 unspec_fmfcfg): New patterns.
4096 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4097 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4098 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4099 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4100 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4101 __nds32__fmfcfg): Define.
4103 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4105 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4106 intrinsic register names.
4107 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4108 intrinsic register enum values and macros.
4110 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4112 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4113 for load/store addressing form.
4114 (nds32_print_operand_address): Likewise.
4116 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4119 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4120 based on LABEL_REF. Remove useless assertion.
4121 (pic_address_needs_scratch): Fix formatting.
4122 (sparc_legitimize_pic_address): Minor tweaks.
4123 (sparc_delegitimize_address): Adjust assertion accordingly.
4124 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4125 into symbolic_operand.
4126 (movsi_high_pic_label_ref): Likewise.
4127 (movsi_lo_sum_pic_label_ref): Likewise.
4128 (movdi_pic_label_ref): Likewise.
4129 (movdi_high_pic_label_ref): Likewise.
4130 (movdi_lo_sum_pic_label_ref): Likewise.
4132 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4134 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4135 custom LIB_SPEC setup.
4137 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4138 Kito Cheng <kito.cheng@gmail.com>
4140 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4141 * config/riscv/freebsd.h: New.
4143 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4145 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4146 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4149 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4150 Kito Cheng <kito.cheng@gmail.com>
4152 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4153 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4154 * config/nds32/nds32-protos.h (nds32_output_call,
4155 nds32_output_return): Declare.
4156 * config/nds32/nds32.md: Refine all the call and return patterns.
4158 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4161 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4162 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4164 PR rtl-optimization/84872
4165 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4166 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4169 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4171 * expr.c (copy_blkmode_to_reg): Revert 254862.
4172 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4174 2018-04-06 Richard Biener <rguenther@suse.de>
4177 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4178 after seeing a component reference with an adjacent field. Treat
4179 refs to arrays at struct end of external decls similar to
4180 refs to unconstrained commons.
4182 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4185 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4186 look through SAVE_EXPRs with non-side-effects argument. Adjust
4188 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4191 2018-04-06 Richard Biener <rguenther@suse.de>
4194 * alias.c (find_base_term): New wrapper around find_base_term
4195 unwinding CSELIB_VAL_PTR changes.
4196 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4199 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4201 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4203 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4204 constant definitions.
4205 ("nop"): lr 0,0 -> nopr r0
4206 ("nop_lr0", "nop_lr1"): New insn definitions.
4208 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4210 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4211 NDS32_V3PUSH_AVAILABLE_P macro.
4213 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4214 Chung-Ju Wu <jasonwucj@gmail.com>
4216 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4217 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4218 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4219 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4220 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4221 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4222 * config/nds32/constraints.md: New constraints and checking for hard
4223 float configuration.
4224 * config/nds32/iterators.md: New mode iterator and attribute for hard
4225 float configuration.
4226 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4228 * config/nds32/nds32-fpu.md: New file.
4229 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4230 deal with hard float code generation.
4231 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4233 (abi_type, float_reg_number): New enum type.
4234 * config/nds32/nds32-predicates.c: New predicates for hard float.
4235 * config/nds32/nds32-protos.h: Declare functions for hard float.
4236 * config/nds32/nds32.c: Implementation for hard float configuration.
4237 * config/nds32/nds32.h: Definitions for hard float configuration.
4238 * config/nds32/nds32.md: Include hard float machine description and
4239 modify patterns for hard float configuration.
4240 * config/nds32/nds32.opt: New options for hard float configuration.
4241 * config/nds32/predicates.md: New predicates for hard float
4244 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4246 * common/config/nds32/nds32-common.c
4247 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4249 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4252 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4253 CONSTRUCTOR_ELT (ctor, ...)->value.
4255 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4258 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4260 2018-04-05 Tom de Vries <tom@codesourcery.com>
4263 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4266 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4267 Kito Cheng <kito.cheng@gmail.com>
4269 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4270 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4271 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4272 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4274 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4275 Kito Cheng <kito.cheng@gmail.com>
4277 * config/nds32/constraints.md (Ufe): New memory constraint.
4278 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4279 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4280 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4282 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4283 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4285 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4287 * config/nds32/nds32.md: Use optimize_size in the condition for
4288 alu-shift instructions.
4290 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4292 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4294 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4296 * config/nds32/nds32.md (negsi2): Refine pattern.
4298 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4299 Chung-Ju Wu <jasonwucj@gmail.com>
4301 * config/nds32/iterators.md (shift_rotate): New code iterator.
4302 (shift): New code attribute.
4303 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4304 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4305 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4306 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4307 bit-wise operations.
4308 (andsi3, *andsi3): Ditto.
4309 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4310 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4311 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4312 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4313 nds32_ior_operand, nds32_xor_operand): New predicates.
4315 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4317 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4318 (addsi3, subsi3): ... this.
4320 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4322 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4324 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4326 * config/nds32/nds32.md: Adjust indention.
4328 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4330 * config/nds32/nds32.md (feature): New attribute.
4332 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4334 * config/nds32/nds32.md (subtype): New attribute.
4336 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4339 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4340 expansion to perform a bitwise AND of the argument followed by a
4341 boolean negation of the result.
4343 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4345 PR rtl-optimization/84878
4346 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4347 the basic block. Assert the use reference is not artificial and that
4348 it has an associated insn.
4350 2018-04-04 Michael Matz <matz@suse.de>
4352 * builtins.c (compute_objsize): Pass correct operand
4353 to array_at_struct_end_p.
4355 2018-04-04 Richard Biener <rguenther@suse.de>
4358 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4359 from contexts for DINFO_LEVEL_TERSE and below.
4361 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4363 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4364 resiter_operand condition.
4365 * config/nds32/nds32.md (*move<mode>): Ditto.
4367 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4368 Monk Chiang <sh.chiang04@gmail.com>
4370 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4372 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4374 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4376 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4377 Kito Cheng <kito.cheng@gmail.com>
4379 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4380 nds32_cond_code_str, output_cond_branch,
4381 output_cond_branch_compare_zero, nds32_expand_cbranch,
4382 nds32_expand_cstore, nds32_expand_movcc,
4383 nds32_output_cbranchsi4_equality_zero,
4384 nds32_output_cbranchsi4_equality_reg,
4385 nds32_output_cbranchsi4_equality_reg_or_const_int,
4386 nds32_output_cbranchsi4_greater_less_zero: New functions.
4387 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4388 nds32_expand_cstore, nds32_expand_movcc,
4389 nds32_output_cbranchsi4_equality_zero,
4390 nds32_output_cbranchsi4_equality_reg,
4391 nds32_output_cbranchsi4_equality_reg_or_const_int,
4392 nds32_output_cbranchsi4_greater_less_zero): Declare.
4393 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4394 nds32_rimm11s_operand): New predicates.
4395 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4396 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4399 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4401 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4402 * config/nds32/nds32.md: Ditto.
4403 * config/nds32/pipelines.md: Ditto.
4405 2018-04-04 Richard Biener <rguenther@suse.de>
4407 PR tree-optimization/85168
4408 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4409 propagating abnormals.
4411 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4413 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4415 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4416 Kito Cheng <kito.cheng@gmail.com>
4418 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4419 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4420 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4421 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4422 * config/nds32/nds32.md (sibcall_internal): New.
4423 (sibcall_register): Remove.
4424 (sibcall_immediate): Remove.
4425 (sibcall_value_internal): New.
4426 (sibcall_value_register): Remove.
4427 (sibcall_value_immediate): Remove.
4428 * config/nds32/predicates.md (nds32_general_register_operand): New.
4429 (nds32_call_address_operand): New.
4431 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4433 PR rtl-optimization/85167
4434 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4435 bb_defs if *split_p, instead preinitialize it to NULL.
4437 PR tree-optimization/85156
4438 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4439 evaluating the argument multiple times.
4441 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4443 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4445 (_mm_cvtpd_ps): Likewise.
4446 (_mm_cvttpd_epi32): Likewise.
4447 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4448 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4449 vector, pixel, and bool following altivec.h include.
4451 2018-04-03 Martin Sebor <msebor@redhat.com>
4453 * doc/extend.texi (Common Function Attributes): Clarify.
4454 (const attribute): Likewise.
4455 (pure attribute): Likewise.
4457 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4460 * config/i386/i386.c (ix86_expand_vector_set): Use
4461 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4463 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4465 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4466 instructions when changing rounding bits to preserve precision bits
4467 in the x87 control word.
4469 2018-04-03 Martin Liska <mliska@suse.cz>
4471 PR tree-optimization/82491
4472 * rtl.h (strip_offset_and_add): Replace += suboffset with
4473 poly_uint64 () + suboffset.
4475 2018-03-29 Martin Liska <mliska@suse.cz>
4476 Martin Jambor <mjambor@suse.cz>
4479 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4480 param_type is not an integral or pointer type.
4482 2018-04-03 Richard Biener <rguenther@suse.de>
4484 * sese.h (recompute_all_dominators): Remove.
4486 2018-04-02 Martin Sebor <msebor@redhat.com>
4488 * doc/invoke.texi (-Wrestrict): Fix typos.
4490 2018-04-02 Jim Wilson <jimw@sifive.com>
4492 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4493 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4494 (<optab>di3, <optab>si3_extend): Likewise.
4495 (<optab>si3_mask, <optab>si3_mask_1): New.
4496 (<optab>di3_mask, <optab>di3_mask_1): New.
4497 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4498 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4499 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4501 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4503 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4506 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4508 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4509 (nds32_canonicalize_comparison): New function.
4511 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4512 Kito Cheng <kito.cheng@gmail.com>
4513 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4515 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4516 * config/nds32/constants.md (unspec_volatile_element): Add
4517 UNSPEC_VOLATILE_RELAX_GROUP.
4518 * config/nds32/nds32-relax-opt.c: New file.
4519 * config/nds32/nds32-predicates.c
4520 (nds32_symbol_load_store_p): New function.
4521 * config/nds32/nds32-protos.h
4522 (nds32_symbol_load_store_p): Declare function.
4523 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4524 * config/nds32/nds32.c
4525 (nds32_register_pass): New function to register pass.
4526 (nds32_register_passes): New function to register passes.
4527 * config/nds32/nds32.md (relax_group): New pattern.
4528 * config/nds32/nds32.opt (mrelax-hint): New option.
4529 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4531 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4533 * config/nds32/t-nds32: Modify files dependency.
4535 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4537 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4538 (PROFILE_HOOK): Define its implementation.
4540 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4542 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4543 type and 32-bit size.
4545 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4548 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4549 (V_128_256): New mode iterator.
4550 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4551 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4552 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4554 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4555 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4557 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4560 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4561 NaN inputs correctly.
4563 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4566 * config/rs6000/vsx.md (??r): New mode attribute.
4567 (*vsx_mov<mode>_64bit): Use it.
4568 (*vsx_mov<mode>_32bit): Likewise.
4570 2018-03-30 Martin Sebor <msebor@redhat.com>
4572 PR tree-optimization/84818
4573 * builtins.c (check_access): Use warning_n.
4575 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4578 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4580 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4583 2018-03-30 Julia Koval <julia.koval@intel.com>
4586 * x86-tune.def (movx, partial_reg_dependency): Enable for
4589 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4592 * lra-constraints.c (process_alt_operands): Move setting
4593 this_alternative_matches below.
4595 2018-03-29 Martin Liska <mliska@suse.cz>
4598 * doc/invoke.texi: Document how LTO works with debug info.
4599 Describe auto-load support of binutils. Mention 'x86-64'
4600 as valid option value of -march option.
4602 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4604 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4607 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4608 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4609 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4612 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4615 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4616 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4617 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4618 for RS6000_BTM_POWERPC64.
4619 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4620 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4621 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4626 2018-03-28 Carl Love <cel@us.ibm.com>
4629 2017-09-27 Carl Love <cel@us.ibm.com>
4631 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4632 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4633 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4636 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4638 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4639 instead of __vector bool.
4640 (_mm_max_pu8): Likewise.
4641 (_mm_min_pi16): Likewise.
4643 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4646 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4647 (DIVWEUO): Likewise.
4649 (DIVDEUO): Likewise.
4650 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4651 DIVWEUO and DIVDEUO.
4652 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4653 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4654 (div_extend): Likewise.
4655 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4657 (__builtin_divweuo): Likewise.
4658 (__builtin_divdeo): Likewise.
4659 (__builtin_divdeuo): Likewise.
4661 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4664 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4665 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4667 PR tree-optimization/82004
4668 * gimple-match-head.c (optimize_pow_to_exp): New function.
4669 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4670 Don't fold to exp if optimize_pow_to_exp is false.
4672 2018-03-28 Martin Liska <mliska@suse.cz>
4675 * calls.c (initialize_argument_information): Fix trailing space.
4676 * common.opt: Fix typo and provide better explanation for
4677 -fsanitize-coverage option.
4678 * config/i386/i386.opt: Fix typo.
4680 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4681 Martin Liska <mliska@suse.cz>
4684 * gimplify.c (asan_poison_variable): Don't do the check for
4685 gimplify_omp_ctxp here.
4686 (gimplify_decl_expr): Do it here.
4687 (gimplify_target_expr): Likewise.
4689 2018-03-28 Martin Liska <mliska@suse.cz>
4692 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4693 chkp_type_bounds_count if MPX is not enabled.
4695 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4697 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4699 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4702 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4703 function to create the function decl for complex long double
4704 multiply and divide for -mabi=ieeelongdouble.
4705 (init_float128_ieee): Call it.
4707 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4710 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4711 -fcf-protection=branch -mibt.
4712 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4714 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4717 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4719 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4722 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4723 extern array declarations.
4725 2018-03-27 Richard Biener <rguenther@suse.de>
4728 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4729 explicit single_use checks.
4731 2018-03-27 Richard Biener <rguenther@suse.de>
4733 PR tree-optimization/85082
4734 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4737 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4739 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4740 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4741 Turn on fasynchronous-unwind-tables and funwind-tables.
4743 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4746 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4747 (*bmi_blsr_<mode>_ccz): Ditto.
4749 2018-03-26 Tom de Vries <tom@codesourcery.com>
4751 PR tree-optimization/85063
4752 * omp-general.c (offloading_function_p): New function. Factor out
4754 * omp-offload.c (pass_omp_target_link::gate): ... here.
4755 * omp-general.h (offloading_function_p): Declare.
4756 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4757 with attribute omp declare target for offloading functions.
4759 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4761 PR tree-optimization/84005
4762 * tree-data-ref.h (get_base_for_alignment): Declare.
4763 * tree-data-ref.c (get_base_for_alignment_1): New function.
4764 (get_base_for_alignment): Likewise.
4765 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4766 get_base_for_alignment to find a suitable base object, instead
4767 of always using drb->base_address.
4769 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4772 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4773 known size by default.
4775 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4778 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4779 and non BLKmode operands.
4781 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4784 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4785 Clean up attributes.
4787 2018-03-23 Richard Biener <rguenther@suse.de>
4790 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4791 we are going to emit early debug for LTO.
4793 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4796 * function.c (match_asm_constraints_1): Don't optimize if input
4797 doesn't satisfy general_operand predicate for output's mode.
4800 * alias.c (write_dependence_p): Don't require for x_canonicalized
4801 non-VOIDmode if x has VOIDmode.
4804 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4805 just don't try to optimize it rather than assert it never happens.
4807 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4809 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4810 macro expansions for definition of ST_INTERNAL_<mode> and
4811 LD_INTERNAL_<mode> builtins.
4812 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4814 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4816 (altivec_expand_st_builtin): Likewise.
4817 (altivec_expand_builtin): Remove calls to deleted functions.
4818 (rs6000_address_for_altivec): Delete this function.
4819 * config/rs6000/vector.md: Remove expands for
4820 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4822 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4825 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4826 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4827 re-computing once computed.
4828 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4829 (arm_init_machine_status): Initialize
4830 machine->static_chain_stack_bytes.
4832 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4835 * doc/extend.texi: Add four new prototypes for vec_ld.
4836 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4837 definitions for more logical presentation.
4838 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4839 entries for V1TI variants of __builtin_altivec_ld builtin.
4840 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4841 handling of V1TI variant of LVX icode pattern.
4842 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4843 (rs6000_gimple_fold_builtin): Likewise.
4844 (altivec_init_builtins): Add code to define
4845 __builtin_altivec_lvx_v1ti function.
4847 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4850 * function.c (match_asm_constraints_1): Don't do the optimization
4851 if input isn't a REG, SUBREG, MEM or constant.
4853 2018-03-22 Tom de Vries <tom@codesourcery.com>
4855 PR tree-optimization/84956
4856 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4857 bb_has_abnormal_pred.
4859 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4862 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4863 DECL_INITIAL (decl) to decl at the end.
4864 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4867 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4869 * doc/extend.texi (__builtin_tgmath): Document when complex
4870 integer types are treated as _Complex _Float64.
4872 2018-03-21 Tom de Vries <tom@codesourcery.com>
4874 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4876 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4878 PR tree-optimization/84960
4879 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4880 if it is ENTRY block, move them into single succ of ENTRY in that case.
4882 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4884 PR tree-optimization/84811
4885 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4886 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4887 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4888 (known_subrange_p): Update accordingly. Cast each value involved
4889 in the size comparison, rather than casting the result of the
4892 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4894 PR tree-optimization/84982
4895 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4896 by flipping the least significant bit rather than all bits from
4897 bitpos to bitpos + bitsize - 1.
4899 2018-03-21 Nathan Sidwell <nathan@acm.org>
4901 * doc/extend.texi (Deprecated Features): Remove mention of
4902 long-deleted deprecations.
4904 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4907 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4908 * configure: Regenerate.
4910 2018-03-21 Tom de Vries <tom@codesourcery.com>
4912 PR tree-optimization/83126
4913 * tree-parloops.c (num_phis): New function.
4914 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4916 2018-03-21 Nathan Sidwell <nathan@acm.org>
4918 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4919 mention anon-struct/union members and trailing attributes.
4921 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4923 PR tree-optimization/84969
4924 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4925 builtin memset partitions if they set different rhs values.
4927 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4929 PR rtl-optimization/84989
4930 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4931 VEC_DUPLICATE with scalar result mode.
4933 2018-03-21 Martin Liska <mliska@suse.cz>
4936 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4937 not intended return statement.
4939 2018-03-21 Martin Liska <mliska@suse.cz>
4942 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4943 (chkp_find_bound_slots_1): Limit number of iterations.
4945 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4948 * Minor grammar fixes for x86 options.
4950 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4953 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4954 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4956 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4959 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4960 (altivec_lvx_<mode>_1op): Likewise.
4961 (altivec_stvx_<mode>_2op): Likewise.
4962 (altivec_stvx_<mode>_1op): Likewise.
4963 (altivec_lvx_<VM2:mode>): New define_expand.
4964 (altivec_stvx_<VM2:mode>): Likewise.
4965 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4966 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4967 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4968 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4969 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4970 (rs6000_gen_lvx): Likewise.
4971 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4972 (altivec_expand_stv_builtin): Likewise.
4973 (altivec_expand_builtin): Likewise.
4974 * config/rs6000/vector.md: Likewise.
4976 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4979 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4982 2018-03-20 Richard Biener <rguenther@suse.de>
4985 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4986 sign-conversions as zero, fall back to standard scalar_stmt
4989 2018-03-20 Martin Liska <mliska@suse.cz>
4992 * predict.c (rebuild_frequencies): Handle case when we have
4993 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4995 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4998 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4999 flag_section_anchors.
5000 * varasm.c (use_blocks_for_decl_p): Remove hack for
5001 dw2_force_const_mem.
5004 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5006 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5007 be created, use lowpart_subreg of operands[0] rather than operands[0]
5009 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5010 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5011 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5012 and n constraint instead of aarch64_shift_imm_di and Usd.
5013 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5014 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5016 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5019 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5020 to favor GPR over NEON registers.
5021 (<shift>di3_neon): Likewise.
5023 2018-03-20 Tom de Vries <tom@codesourcery.com>
5026 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5027 (nvptx_process_pars): Emit bar.sync asap and alap.
5029 2018-03-20 Tom de Vries <tom@codesourcery.com>
5032 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5033 seen_label if seen_label is already set.
5035 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5038 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5039 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5040 Use 1U instead of 1. Formatting fixes.
5043 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5044 instead of TREE_TYPE (s1) for the return value.
5046 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5048 PR tree-optimization/84946
5049 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5050 bitsize + bitsize in poly_uint64 rather than poly_int64.
5053 * dwarf2asm.c: Include fold-const.c.
5054 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5055 of decl rather than decl itself.
5057 PR rtl-optimization/84643
5058 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5060 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5063 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5064 calling assemble_variable.
5066 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5069 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5070 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5072 2018-03-19 Jim Wilson <jimw@sifive.com>
5075 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5076 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5077 (riscv_first_stack_step): Likewise.
5078 (riscv_option_override): Use STACK_BOUNDARY instead of
5080 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5082 (BIGGEST_ALIGNMENT): Set to 128.
5083 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5084 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5087 2018-03-19 Richard Biener <rguenther@suse.de>
5089 PR tree-optimization/84933
5090 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5091 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5093 2018-03-19 Richard Biener <rguenther@suse.de>
5095 PR tree-optimization/84859
5096 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5097 (cond_if_else_store_replacement): Perform sinking operation on
5098 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5099 Generalize what a BB with a single eligible store is.
5101 2018-03-19 Richard Biener <rguenther@suse.de>
5103 PR tree-optimization/84929
5104 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5105 chrec_is_positive against non-chrec arg.
5107 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5110 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5112 2018-03-18 Martin Liska <mliska@suse.cz>
5114 PR rtl-optimization/84635
5115 * regrename.c (build_def_use): Use matches_mode only when
5118 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5120 PR tree-optimization/84913
5121 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5122 vectorize chains of COND_EXPRs.
5124 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5126 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5128 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5130 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5132 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5134 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5136 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5137 Kito Cheng <kito.cheng@gmail.com>
5139 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5140 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5141 (nds32_adjust_reg_alloc_order): New function.
5142 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5144 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5146 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5147 nds32_print_operand, nds32_print_operand_address): Use
5148 HOST_WIDE_INT_PRINT_DEC instead.
5150 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5152 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5154 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5157 * config/i386/i386.c (initial_ix86_tune_features,
5158 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5160 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5161 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5162 rather than 1u << ix86_tune. Formatting fix.
5163 (ix86_option_override_internal): Change ix86_arch_mask from
5164 unsigned int to unsigned HOST_WIDE_INT, initialize to
5165 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5166 (ix86_function_specific_restore): Likewise.
5168 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5171 * postreload.c (reload_combine_recognize_pattern): Perform
5172 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5173 truncate_int_for_mode the result for the destination's mode.
5176 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5177 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5179 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5182 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5185 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5186 regno_allocno_class_array and sorted_pseudos.
5187 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5188 insns where regno is used.
5190 2018-03-16 Martin Liska <mliska@suse.cz>
5193 * multiple_target.c (create_dispatcher_calls): Redirect
5194 reference in the symbol table.
5196 2018-03-16 Martin Liska <mliska@suse.cz>
5199 * multiple_target.c (create_dispatcher_calls): Redirect also
5202 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5206 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5207 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5208 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5210 2018-03-16 Julia Koval <julia.koval@intel.com>
5212 * doc/invoke.texi (Skylake Server): Add CLWB.
5213 Cannonlake): Remove CLWB.
5215 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5217 PR tree-optimization/84841
5218 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5220 (FLOAT_ONE_CONST_TYPE): Define.
5221 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5222 (sort_by_operand_rank): Put entries with higher constant_type last
5223 rather than first to match comments.
5225 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5227 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5230 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5233 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5234 check and use address of target always.
5236 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5239 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5240 (indirect_thunk_bnd_needed): Likewise.
5241 (indirect_thunks_used): Likewise.
5242 (indirect_thunks_bnd_used): Likewise.
5243 (indirect_return_needed): New.
5244 (indirect_return_bnd_needed): Likewise.
5245 (output_indirect_thunk_function): Add a bool argument for
5247 (output_indirect_thunk_function): Don't generate alias for
5248 function return thunk.
5249 (ix86_code_end): Call output_indirect_thunk_function to generate
5250 function return thunks.
5251 (ix86_output_function_return): Set indirect_return_bnd_needed
5252 and indirect_return_needed instead of indirect_thunk_bnd_needed
5253 and indirect_thunk_needed.
5255 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5257 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5258 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5259 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5261 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5262 Paul Hua <paul.hua.gm@gmail.com>
5265 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5267 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5269 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5270 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5273 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5276 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5277 instead of GET_MODE_SIZE when comparing Units.
5279 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5282 * varasm.c (hash_section): Return an unchangeble hash value
5283 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5284 Return !aarch64_can_use_per_function_literal_pools_p ().
5286 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5289 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5290 rather than address of cmode as last argument to prepare_cmp_insn.
5292 2018-03-15 Julia Koval <julia.koval@intel.com>
5294 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5295 F_AVX512VNNI, F_AVX512BITALG): New.
5297 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5300 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5301 insn for floating-point loads and stores.
5303 2018-03-14 Carl Love <cel@us.ibm.com>
5305 * config/rs6000/rs6000-c.c: Add macro definitions for
5306 ALTIVEC_BUILTIN_VEC_PERMXOR.
5307 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5308 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5309 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5310 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5312 * config/doc/extend.texi: Add prototypes for vec_permxor.
5314 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5317 * diagnostic-show-locus.c (class layout_point): Convert m_line
5318 from int to linenum_type.
5319 (line_span::comparator): Use linenum "compare" function when
5320 comparing line numbers.
5321 (test_line_span): New function.
5322 (layout_range::contains_point): Convert param "row" from int to
5324 (layout_range::intersects_line_p): Likewise.
5325 (layout::will_show_line_p): Likewise.
5326 (layout::print_source_line): Likewise.
5327 (layout::should_print_annotation_line_p): Likewise.
5328 (layout::print_annotation_line): Likewise.
5329 (layout::print_leading_fixits): Likewise.
5330 (layout::annotation_line_showed_range_p): Likewise.
5331 (struct line_corrections): Likewise for field m_row.
5332 (line_corrections::line_corrections): Likewise for param "row".
5333 (layout::print_trailing_fixits): Likewise.
5334 (layout::get_state_at_point): Likewise.
5335 (layout::get_x_bound_for_row): Likewise.
5336 (layout::print_line): Likewise.
5337 (diagnostic_show_locus): Likewise for locals "last_line" and
5339 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5340 * input.c (selftest::test_linenum_comparisons): New function.
5341 (selftest::input_c_tests): Call it.
5342 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5343 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5344 * selftest.h (ASSERT_GT): New macro.
5345 (ASSERT_GT_AT): New macro.
5346 (ASSERT_LT): New macro.
5347 (ASSERT_LT_AT): New macro.
5349 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5351 PR rtl-optimization/84780
5352 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5354 2018-03-14 Martin Liska <mliska@suse.cz>
5356 * tree.c (record_node_allocation_statistics): Use
5357 get_stats_node_kind.
5358 (get_stats_node_kind): New function extracted from
5359 record_node_allocation_statistics.
5360 (free_node): Use get_stats_node_kind.
5362 2018-03-14 Richard Biener <rguenther@suse.de>
5364 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5365 that the value-set of ANTIC_IN doesn't grow.
5368 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5370 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5371 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5373 2018-03-14 Julia Koval <julia.koval@intel.com>
5375 * config.gcc (icelake-client, icelake-server): New.
5377 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5378 (initial_ix86_arch_features): Ditto.
5379 (PTA_SKYLAKE): Add SGX.
5380 (PTA_ICELAKE): Remove.
5381 (PTA_ICELAKE_CLIENT): New.
5382 (PTA_ICELAKE_SERVER): New.
5383 (ix86_option_override_internal): Split up icelake on icelake client and
5385 (get_builtin_code_for_version): Ditto.
5386 (fold_builtin_cpu): Ditto.
5387 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5388 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5389 * config/i386/i386.h (processor_type): Ditto.
5390 * doc/invoke.texi: Ditto.
5392 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5395 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5396 INTEGER_CST offset, add it together with bitpos / 8 and
5397 sign extend based on POINTER_SIZE.
5401 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5404 * config/i386/constraints.md (Yc): New register constraint.
5405 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5406 Use Yc constraint for alternative 2 of operand 0. Remove
5407 preferred_for_speed attribute.
5409 2018-03-14 Richard Biener <rguenther@suse.de>
5411 PR tree-optimization/84830
5412 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5413 with the old one to avoid oscillations.
5415 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5418 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5420 (assign_by_spills): Return a flag of reload assignment failure.
5421 Do not process the reload assignment failures. Do not spill other
5422 reload pseudos if they has the same reg class. Update n if
5424 (lra_assign): Add a return arg. Set up from the result of
5425 assign_by_spills call.
5426 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5427 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5428 usage_insns if it is not NULL.
5429 (spill_hard_reg_in_range): New function.
5430 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5431 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5432 function prototypes.
5433 (lra_assign): Change prototype.
5434 * lra.c (lra): Add code to deal with fails by splitting hard reg
5437 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5439 * config/riscv/riscv.opt (mrelax): New option.
5440 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5441 "norelax" when riscv_mrelax is disabled.
5442 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5444 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5447 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5448 reassociation for int modes.
5450 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5452 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5453 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5455 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5456 * config/aarch64/aarch64-sve.md
5457 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5458 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5459 (*extend<mode><Vwide>2): Rename to...
5460 (aarch64_sve_extend<mode><Vwide>2): ...this.
5461 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5462 renaming the old pattern to...
5463 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5465 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5466 define_expand, renaming the old pattern to...
5467 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5468 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5469 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5470 account when deciding which SVE instruction the optab should use.
5471 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5473 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5475 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5476 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5477 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5478 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5479 (tlsdesc_small_<mode>): Turn a define_expand and use
5480 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5481 (tlsdesc_small_advsimd_<mode>): ...this.
5482 (tlsdesc_small_sve_<mode>): New pattern.
5484 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5486 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5487 (UNSPEC_UMUL_HIGHPART): New constants.
5488 (MUL_HIGHPART): New int iteraor.
5489 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5490 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5492 (*<su>mul<mode>3_highpart): New define_insn.
5494 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5497 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5500 2018-03-13 Martin Liska <mliska@suse.cz>
5503 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5505 (sem_item_optimizer::~sem_item_optimizer): Release it.
5506 (sem_item_optimizer::merge_classes): Register variable aliases.
5507 (sem_item_optimizer::fixup_pt_set): New function.
5508 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5509 * ipa-icf.h: Declare new variables and functions.
5511 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5514 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5515 integer_pow2p@2 and test integer_pow2p in condition.
5516 (A < 0 ? C : 0): Similarly for @1.
5519 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5520 characters starting at p contain '\0' character, don't look beyond
5524 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5525 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5528 * reg-stack.c (change_stack): Change update_end var from int to
5529 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5530 also call set_block_for_insn on the newly added insns and rescan.
5533 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5534 on the last operand.
5537 * tree.c (stabilize_reference_1): Return save_expr (e) for
5538 STATEMENT_LIST even if it doesn't have side-effects.
5540 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5542 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5544 2018-03-12 Renlin Li <renlin.li@arm.com>
5546 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5547 aarch64_output_scalar_simd_mov_immediate.
5549 2018-03-12 Martin Sebor <msebor@redhat.com>
5551 PR tree-optimization/83456
5552 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5553 for perfectly overlapping calls to memcpy.
5554 (gimple_fold_builtin_memory_chk): Same.
5555 (gimple_fold_builtin_strcpy): Handle no-warning.
5556 (gimple_fold_builtin_stxcpy_chk): Same.
5557 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5559 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5561 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5562 parameter. Use it for SFmode.
5563 (rs6000_function_arg_advance_1): Adjust.
5564 (rs6000_function_arg): Adjust.
5565 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5567 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5569 PR rtl-optimization/84169
5570 PR rtl-optimization/84780
5571 * combine.c (can_combine_p): Check for a 2-insn combination whether
5572 the destination register is used between the two insns, too.
5574 2018-03-12 Richard Biener <rguenther@suse.de>
5576 PR tree-optimization/84803
5577 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5578 for refs DR analysis didn't process.
5580 2018-03-12 Richard Biener <rguenther@suse.de>
5582 PR tree-optimization/84777
5583 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5584 force-vectorize loops ignore whether we are optimizing for size.
5586 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5588 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5589 (TARGET_MD_ASM_ADJUST): Define.
5591 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5592 Kito Cheng <kito.cheng@gmail.com>
5593 Chung-Ju Wu <jasonwucj@gmail.com>
5595 * config/nds32/nds32.c (nds32_compute_stack_frame,
5596 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5597 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5598 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5599 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5600 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5601 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5602 * config/nds32/nds32.md (prologue, epilogue): Use macro
5603 NDS32_V3PUSH_AVAILABLE_P to do checking.
5605 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5608 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5609 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5610 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5611 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5612 addition of most attributes on !orig_type_die or the attribute not
5613 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5615 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5616 Chung-Ju Wu <jasonwucj@gmail.com>
5618 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5620 * config/nds32/nds32.opt (mvh): New option.
5622 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5623 Chung-Ju Wu <jasonwucj@gmail.com>
5625 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5627 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5628 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5631 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5632 Chung-Ju Wu <jasonwucj@gmail.com>
5634 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5636 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5637 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5639 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5640 Kito Cheng <kito.cheng@gmail.com>
5641 Chung-Ju Wu <jasonwucj@gmail.com>
5643 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5644 UNSPEC_FFMISM and UNSPEC_FLMISM.
5645 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5646 for ffb, ffmism and flmism.
5647 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5648 (unspec_ffmism): Ditto.
5649 (unspec_flmism): Ditto.
5650 (nds32_expand_builtin_impl): Check if string extension is available.
5651 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5652 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5654 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5657 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5660 * lra-assigns.c (assign_by_spills): Return a flag of reload
5661 assignment failure. Do not process the reload assignment
5662 failures. Do not spill other reload pseudos if they has the same
5664 (lra_assign): Add a return arg. Set up from the result of
5665 assign_by_spills call.
5666 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5667 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5668 usage_insns if it is not NULL.
5669 (spill_hard_reg_in_range): New function.
5670 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5671 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5672 function prototypes.
5673 (lra_assign): Change prototype.
5674 * lra.c (lra): Add code to deal with fails by splitting hard reg
5677 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5680 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5682 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5685 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5686 (dw_val_node): Add val_symbolic_view.
5687 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5688 (symview_upper_bound): New.
5689 (new_line_info_table): Initialize symviews_since_reset.
5690 (dwarf2out_source_line): Count symviews_since_reset and set
5691 symview_upper_bound.
5692 (dw_val_equal_p): Handle symview.
5693 (add_AT_symview): New.
5694 (print_dw_val): Handle symview.
5695 (attr_checksum, attr_checksum_ordered): Likewise.
5696 (same_dw_val_p, size_of_die): Likewise.
5697 (value_format, output_die): Likewise.
5698 (add_high_low_attributes): Use add_AT_symview for entry_view.
5699 (dwarf2out_finish): Reset symview_upper_bound, clear
5702 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5705 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5706 Add strict argument and use it.
5707 (rs6000_split_multireg_move): Update for new strict argument.
5708 (mem_operand_gpr): Disallow all non-offsettable addresses.
5709 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5711 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5714 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5715 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5716 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5719 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5720 decl, use remap_type if we want to use the type.
5722 2018-03-09 Martin Sebor <msebor@redhat.com>
5724 PR tree-optimization/84526
5725 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5727 (builtin_access::generic_overlap): Be prepared to handle non-array
5730 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5732 PR rtl-optimization/84682
5733 * lra-constraints.c (process_address_1): Check is_address flag
5734 for address constraints.
5735 (process_alt_operands): Likewise.
5736 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5737 preprocess_constraints.
5738 * recog.h (preprocess_constraints): Add oploc parameter.
5740 * recog.c (preprocess_constraints): Test address_operand for
5741 CT_ADDRESS constraints.
5743 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5746 * lra-assigns.c (assign_by_spills): Return a flag of reload
5747 assignment failure. Do not process the reload assignment
5748 failures. Do not spill other reload pseudos if they has the same
5750 (lra_assign): Add a return arg. Set up from the result of
5751 assign_by_spills call.
5752 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5753 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5754 usage_insns if it is not NULL.
5755 (spill_hard_reg_in_range): New function.
5756 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5757 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5758 function prototypes.
5759 (lra_assign): Change prototype.
5760 * lra.c (lra): Add code to deal with fails by splitting hard reg
5763 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5766 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5767 Accept complain bool parameter. Only emit errors if it is true.
5768 (arm_parse_cpu_option_name): Likewise.
5769 (arm_target_thumb_only): Adjust callers of the above.
5770 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5771 prototype to take a default true bool parameter.
5772 (arm_parse_arch_option_name): Likewise.
5774 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5775 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5779 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5780 * configure: Regenerate.
5781 * configure.ac ("linker --version-script option"): New.
5782 ("linker soname option"): New.
5784 2018-03-09 Richard Biener <rguenther@suse.de>
5786 PR tree-optimization/84775
5787 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5788 immediate uses of predicate stmts and mark them modified.
5791 PR tree-optimization/84178
5792 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5794 (version_loop_for_if_conversion): Delay update_ssa call.
5795 (tree_if_conversion): Delay update_ssa until after predicate
5798 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5801 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5802 when the function accesses prior frames.
5804 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5807 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5808 gen_llsym, otherwise call maybe_gen_llsym.
5811 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5812 has ',' character inside of it.
5814 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5817 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5818 as clobbering CC_REGNUM.
5820 2018-03-08 Richard Biener <rguenther@suse.de>
5823 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5824 (follow_copies_to_constant): Do not follow SSA names registered
5827 2018-03-08 Richard Biener <rguenther@suse.de>
5829 PR tree-optimization/84178
5830 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5832 (version_loop_for_if_conversion): Delay update_ssa call.
5833 (tree_if_conversion): Delay update_ssa until after predicate
5836 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5838 PR tree-optimization/84178
5839 * tree-if-conv.c (release_bb_predicate): Remove the
5840 the assertion that the stmts have NULL use_ops.
5841 Discard the statements, asserting that they haven't
5842 yet been added to a BB.
5844 2018-03-08 Richard Biener <rguenther@suse.de>
5846 PR tree-optimization/84746
5847 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5848 (phi_translate): Pass in destination ANTIC_OUT set.
5849 (phi_translate_1): Likewise. For a simplified result lookup
5850 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5851 (phi_translate_set): Adjust.
5852 (do_pre_regular_insertion): Likewise.
5853 (do_pre_partial_partial_insertion): Likewise.
5855 2018-03-08 Martin Liska <mliska@suse.cz>
5857 PR gcov-profile/84735
5858 * doc/gcov.texi: Document usage of profile files.
5859 * gcov-io.h: Document changes in the format.
5861 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5865 * dwarf2out.c (struct dw_line_info_table): Update comments for
5867 (FORCE_RESET_NEXT_VIEW): New.
5868 (FORCE_RESETTING_VIEW_P): New.
5869 (RESETTING_VIEW_P): Check for -1 too.
5870 (ZERO_VIEW_P): Likewise.
5871 (new_line_info_table): Force-reset next view.
5872 (dwarf2out_begin_function): Likewise.
5873 (dwarf2out_source_line): Simplify zero_view_p initialization.
5874 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5875 view directly. Omit view when omitting .loc at line 0.
5877 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5879 PR tree-optimization/84740
5880 * tree-switch-conversion.c (process_switch): Call build_constructors
5881 only if info.phi_count is non-zero.
5883 PR tree-optimization/84739
5884 * tree-tailcall.c (find_tail_calls): Check call arguments against
5885 DECL_ARGUMENTS (current_function_decl) rather than
5886 DECL_ARGUMENTS (func) when checking for tail recursion.
5888 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5890 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5891 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5892 Volker Reichelt's entry and add entries for people that perform
5893 GCC fuzzy testing and report numerous bugs.
5895 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5898 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5899 readonly data in sdata, if that is disabled.
5900 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5901 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5902 -mreadonly-in-sdata option.
5904 2018-03-07 Martin Sebor <msebor@redhat.com>
5906 PR tree-optimization/84468
5907 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5908 basic block when looking for nul assignment.
5910 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5913 * except.h (output_function_exception_table): Adjust prototype.
5914 * except.c (output_function_exception_table): Remove FNNAME parameter
5915 and add SECTION parameter. Ouput one part of the table at a time.
5916 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5917 the first part of the exception table and emit unwind directives.
5918 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5919 (i386_pe_seh_cold_init): Likewise.
5920 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5921 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5922 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5923 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5924 * config/i386/winnt.c: Include except.h.
5925 (struct seh_frame_state): Add reg_offset, after_prologue and
5926 in_cold_section fields.
5927 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5928 (i386_pe_seh_cold_init): New function.
5929 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5930 to seh->in_cold_section.
5931 (seh_emit_push): Record the offset of the push.
5932 (seh_emit_save): Record the offet of the save.
5933 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5934 Test seh->after_prologue to disregard the epilogue.
5935 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5936 (i386_pe_end_cold_function): New function.
5938 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5941 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5942 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5945 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5947 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5948 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5951 * multiple_target.c: Include tree-inline.h and intl.h.
5952 (expand_target_clones): Diagnose and fail if node->definition and
5953 !tree_versionable_function_p (node->decl).
5955 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5957 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5959 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5960 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5961 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5963 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5966 * combine.c (try_combine): Use reg_or_subregno instead of handling
5967 just paradoxical SUBREGs and REGs.
5969 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5971 * config/arc/arc.c (arc_finalize_pic): Remove function.
5972 (arc_must_save_register): We use single base PIC register, remove
5973 checks to save/restore the PIC register.
5974 (arc_expand_prologue): Likewise.
5975 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5977 (arc_verify_short): Likewise.
5978 (arc_attr_type): Likewise.
5979 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5980 (walk_stores): Likewise.
5981 (arc_address_cost): Make it static.
5982 (arc_verify_short): Likewise.
5983 (branch_dest): Likewise.
5984 (arc_attr_type): Likewise.
5985 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5986 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5987 (arc_final_prescan_insn): Remove inserting the nops due to
5988 hardware hazards. It is done in reorg step.
5989 (insn_length_variant_t): Remove.
5990 (insn_length_parameters_t): Likewise.
5991 (arc_insn_length_parameters): Likewise.
5992 (arc_get_insn_variants): Likewise.
5993 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5995 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5998 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6001 PR tree-optimization/84687
6002 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6004 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6006 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6008 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6009 Rename to ppc_speculation_barrier.
6010 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6011 __builtin_ppc_speculation_barrier.
6013 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6016 * combine.c (combine_simplify_rtx): Don't try to simplify if
6017 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6020 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6022 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6023 to 32 bytes when compiling for POWER9.
6025 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6028 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6029 regparm >= 3 with no arg reg available also for calls with
6030 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6033 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6035 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6037 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6040 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6042 2018-03-05 Richard Biener <rguenther@suse.de>
6044 PR tree-optimization/84486
6045 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6046 When inserting a __builtin_assume_aligned call set the LHS
6047 SSA name alignment info accordingly.
6049 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6051 PR tree-optimization/84114
6052 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6053 Avoid reassociation of FLOAT_MODE addition.
6055 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6057 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6058 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6059 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6060 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6061 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6062 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6063 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6065 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6066 __builtin_ia32_wbinvd): New builtins.
6067 (SPECIAL_ARGS2): New.
6068 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6069 (SPECIAL_ARGS2): New.
6070 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6071 (ix86_valid_target_attribute_inner_p): Ditto.
6072 (ix86_init_mmx_sse_builtins): Add special_args2.
6073 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6074 TARGET_WBNOINVD_P): New.
6075 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6076 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6077 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6078 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6079 * config/i386/pconfigintrin.h: New file.
6080 * config/i386/wbnoinvdintrin.h: Ditto.
6081 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6082 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6084 2018-03-05 Richard Biener <rguenther@suse.de>
6086 PR tree-optimization/84670
6087 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6089 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6090 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6091 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6092 doesn't grow if all successors (recursively) were visited at least
6095 2018-03-05 Richard Biener <rguenther@suse.de>
6097 PR tree-optimization/84650
6098 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6099 if executed in the loop pipeline.
6101 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6103 * doc/configfiles.texi (Configuration Files): Move info about
6104 conditionalizing $target-protos.h to...
6105 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6106 differs from $target-protos.h.
6108 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6109 Chung-Ju Wu <jasonwucj@gmail.com>
6111 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6112 * config/nds32/nds32-multiple.md (setmemsi): Define.
6113 * config/nds32/nds32-memory-manipulation.c
6114 (nds32_gen_dup_4_byte_to_word_value): New.
6115 (emit_setmem_word_loop): New.
6116 (emit_setmem_byte_loop): New.
6117 (nds32_expand_setmem_loop): New.
6118 (nds32_expand_setmem_loop_v3m): New.
6119 (nds32_expand_setmem_unroll): New.
6120 (nds32_expand_setmem): New.
6122 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6123 Chung-Ju Wu <jasonwucj@gmail.com>
6125 * config/nds32/nds32-memory-manipulation.c
6126 (nds32_emit_load_store): New.
6127 (nds32_emit_post_inc_load_store): New.
6128 (nds32_emit_mem_move): New.
6129 (nds32_emit_mem_move_block): New.
6130 (nds32_expand_movmemsi_loop_unknown_size): New.
6131 (nds32_expand_movmemsi_loop_known_size): New.
6132 (nds32_expand_movmemsi_loop): New.
6133 (nds32_expand_movmemsi_unroll): New.
6134 (nds32_expand_movmemqi): Rename ...
6135 (nds32_expand_movmemsi): ... to this.
6136 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6137 (movmemsi): ... to this.
6138 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6139 (nds32_expand_movmemsi): ... to this.
6141 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6142 Monk Chiang <sh.chiang04@gmail.com>
6143 Chung-Ju Wu <jasonwucj@gmail.com>
6145 * config/nds32/nds32-protos.h
6146 (nds32_expand_load_multiple): New arguments.
6147 (nds32_expand_store_multiple): Ditto.
6148 (nds32_valid_multiple_load_store): Rename ...
6149 (nds32_valid_multiple_load_store_p): ... to this.
6150 * config/nds32/nds32-memory-manipulation.c
6151 (nds32_expand_load_multiple): Refine implementation.
6152 (nds32_expand_store_multiple): Ditto.
6153 * config/nds32/nds32-multiple.md
6154 (load_multiple): Update nds32_expand_load_multiple interface.
6155 (store_multiple): Update nds32_expand_store_multiple interface.
6156 * config/nds32/nds32-predicates.c
6157 (nds32_valid_multiple_load_store): Rename ...
6158 (nds32_valid_multiple_load_store_p): ... to this and refine
6160 * config/nds32/predicates.md
6161 (nds32_load_multiple_and_update_address_operation): New predicate.
6162 (nds32_store_multiple_and_update_address_operation): New predicate.
6164 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6165 Chung-Ju Wu <jasonwucj@gmail.com>
6167 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6168 (combo): New attribute.
6169 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6171 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6173 * config/nds32/nds32.opt: Change -mcmodel= default value.
6175 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6176 Monk Chiang <sh.chiang04@gmail.com>
6177 Chung-Ju Wu <jasonwucj@gmail.com>
6179 * config/nds32/constants.md (unspec_element): New enum.
6180 * config/nds32/constraints.md (Umw): New constraint.
6181 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6182 * config/nds32/nds32-intrinsic.md: Likewise.
6183 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6184 (nds32_valid_smw_lwm_base_p): New.
6185 (nds32_output_smw_single_word): New.
6186 (nds32_output_lmw_single_word): New.
6187 (nds32_expand_unaligned_load): New.
6188 (nds32_expand_unaligned_store): New.
6189 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6190 (nds32_output_smw_single_word): Declare.
6191 (nds32_output_lmw_single_word): Declare.
6192 (nds32_expand_unaligned_load): Declare.
6193 (nds32_expand_unaligned_store): Declare.
6194 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6195 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6196 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6197 NDS32_BUILTIN_UASTORE_DW.
6198 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6201 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6202 Kito Cheng <kito.cheng@gmail.com>
6203 Chung-Ju Wu <jasonwucj@gmail.com>
6205 * config/nds32/nds32-intrinsic.c
6206 (nds32_expand_builtin_null_ftype_reg): Delete.
6207 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6208 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6209 (nds32_read_argument): New.
6210 (nds32_legitimize_target): Ditto.
6211 (nds32_legitimize_argument): Ditto.
6212 (nds32_check_constant_argument): Ditto.
6213 (nds32_expand_unop_builtin): Ditto.
6214 (nds32_expand_unopimm_builtin): Ditto.
6215 (nds32_expand_binop_builtin): Ditto.
6216 (nds32_builtin_decl_impl): Ditto.
6217 (builtin_description): Ditto.
6218 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6219 (nds32_init_builtins_impl): Ditto.
6220 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6221 (nds32_builtin_decl): New.
6222 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6223 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6225 2018-03-02 Jeff Law <law@redhat.com>
6227 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6228 (redundant_insn, fill_simple_delay_slots): Likewise.
6229 (fill_slots_from_thread): Likewise.
6230 * resource.c (mark_referenced_resources): Likewise.
6231 (mark_set_resources, find_dead_or_set_registers): Likewise.
6233 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6235 * substring-locations.h (format_warning_va): Formatting fix for
6237 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6239 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6240 * substring-locations.c: Include intl.h.
6241 (format_warning_va): Turned into small wrapper around
6242 format_warning_n_va, renamed to ...
6243 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6244 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6246 (format_warning_at_substring_n): New function.
6247 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6248 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6249 format_warning_at_substring with just a shorter name instead of
6250 const function pointer.
6251 (fmtwarn_n): New function.
6252 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6253 appropriate, get rid of all the fmtstr temporaries, move conditionals
6254 with G_() wrapped string literals directly into fmtwarn arguments,
6255 cast dir.len to (int), formatting fixes.
6257 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6259 * doc/invoke.texi: Remove "Cilk Plus" references.
6261 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6262 Richard Biener <rguenther@suse.de>
6265 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6266 for error or warning attributes if CALL_FROM_THUNK_P is set.
6269 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6272 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6273 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6276 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6277 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6279 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6280 instead of -1U in last predictors element's probability member.
6282 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6285 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6286 arguments if they are comparable.
6288 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6290 PR tree-optimization/84634
6291 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6292 masks and masked_loop_p with a single loop_masks, making sure it's
6293 null for bb vectorization.
6295 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6297 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6298 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6299 loop->force_vectorize to check whether there is no alias.
6301 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6304 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6306 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6308 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6309 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6310 instead of a loop around prev_real_insn.
6311 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6315 * config/i386/i386.c (ix86_print_operand): Use conditional
6316 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6319 2018-03-02 Richard Biener <rguenther@suse.de>
6321 PR tree-optimization/84427
6322 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6323 (bitmap_set_subtract_values): Rewrite to handle multiple
6326 (prune_clobbered_mems): Likewise.
6327 (phi_translate): Take edge instead of pred/phiblock.
6328 (phi_translate_1): Likewise.
6329 (phi_translate_set): Likewise. Insert all translated
6330 exprs for a value into the set, keeping possibly multiple
6331 expressions per value.
6332 (compute_antic_aux): Adjust for phi_translate changes.
6333 When intersecting union the expressions and prune those
6334 not in the final value set, keeping possibly multiple
6335 expressions per value. Do not use value-insertion
6336 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6337 all expressions. Add verification that the value-sets
6338 only shrink during iteration.
6339 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6340 (do_pre_regular_insertion): Likewise.
6341 (do_pre_partial_partial_insertion): Likewise.
6343 2018-03-02 Richard Biener <rguenther@suse.de>
6346 * config/darwin.c (saved_debug_info_level): New static global.
6347 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6348 (darwin_asm_lto_end): Restore debug info generation settings.
6350 2018-03-01 Martin Liska <mliska@suse.cz>
6353 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6356 2018-03-01 Richard Biener <rguenther@suse.de>
6359 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6360 type annotation with LTO when debug was disabled at compile-time.
6362 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6364 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6366 (mips_final_postscan_insn): Likewise.
6368 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6370 PR rtl-optimization/84528
6371 * alias.c (init_alias_target): Add commentary.
6372 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6373 a unique base value if the frame pointer is not eliminated
6374 to the stack pointer.
6376 2018-03-01 Tom de Vries <tom@codesourcery.com>
6378 PR rtl-optimization/83327
6379 * lra-int.h (hard_regs_spilled_into): Declare.
6380 * lra.c (hard_regs_spilled_into): Define.
6381 (init_reg_info): Init hard_regs_spilled_into.
6382 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6383 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6384 (process_bb_lives): Handle hard_regs_spilled_into.
6385 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6386 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6388 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6390 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6391 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6392 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6393 * config/rs6000/aix72.h: New file.
6395 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6397 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6398 instead of warning_at with conditional singular and plural messages
6402 * stor-layout.c (update_alignment_for_field): For
6403 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6404 && !DECL_PACKED (field), do the alignment update, just use
6405 only desired_align instead of MAX (type_align, desired_align)
6407 (place_field): Don't do known_align < desired_align handling
6408 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6409 is non-NULL, instead do it after rli->prev_field handling and
6410 only if not within a bitfield word. For DECL_PACKED (field)
6411 use type_align of BITS_PER_UNIT.
6413 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6415 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6416 superfluous parentheses and trailing spaces.
6418 2018-02-28 Richard Biener <rguenther@suse.de>
6420 PR tree-optimization/84584
6421 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6422 SCoPs with fake exit edge.
6424 2018-02-28 Martin Liska <mliska@suse.cz>
6427 * timevar.c (timer::print): Fix format to properly print 100%
6430 2018-02-28 Richard Biener <rguenther@suse.de>
6433 * genmatch.c (capture_info::walk_match): Do not mark
6434 captured expressions without operands as expr_p given
6435 they act more like predicates and should be subject to
6436 "lost tail" side-effect preserving.
6438 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6440 PR rtl-optimization/81611
6441 * auto-inc-dec.c (attempt_change): Move dead note from
6442 mem_insn if it's the next use of regno
6443 (find_address): Take address use of reg holding
6444 non-incremented value. Add parm to limit search to the named
6446 (merge_in_block): Attempt to use a mem insn that is the next
6447 use of the original regno.
6449 2018-02-27 Martin Sebor <msebor@redhat.com>
6452 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6453 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6455 2018-02-27 Martin Sebor <msebor@redhat.com>
6457 PR translation/84207
6458 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6459 n argument to unsigned HOST_WIDE_INT.
6460 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6461 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6462 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6463 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6465 2018-02-27 Richard Biener <rguenther@suse.de>
6467 PR tree-optimization/84512
6468 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6469 Do not use the estimate returned from record_stmt_cost for
6470 the scalar iteration cost but sum properly using add_stmt_cost.
6472 2018-02-27 Richard Biener <rguenther@suse.de>
6474 PR tree-optimization/84466
6475 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6476 Adjust last change to less strictly validate use operands.
6478 2018-02-27 Martin Liska <mliska@suse.cz>
6480 PR gcov-profile/84548
6481 * gcov.c (process_file): Allow partial overlap and consider it
6482 also as group functions.
6483 (output_lines): Properly calculate range of lines for a group.
6485 2018-02-27 Martin Liska <mliska@suse.cz>
6487 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6488 'ggc' suffixes. Change first column width.
6489 (timer::print): Fix formatting of the column.
6491 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6493 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6494 preserve inline entry blocks for the sake of debug inline
6495 entry point markers alone.
6496 (remove_unused_locals): Suggest in comments a better place to
6497 force the preservation of inline entry blocks that are
6498 otherwise unused, but do not preserve them.
6500 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6502 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6504 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6507 * config/i386/constraints.md (Bs): Replace
6508 ix86_indirect_branch_register with
6509 TARGET_INDIRECT_BRANCH_REGISTER.
6511 * config/i386/i386.md (indirect_jump): Likewise.
6512 (tablejump): Likewise.
6513 (*sibcall_memory): Likewise.
6514 (*sibcall_value_memory): Likewise.
6515 Peepholes of indirect call and jump via memory: Likewise.
6516 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6517 (*sibcall_value_GOT_32): Likewise.
6518 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6519 (GOT_memory_operand): Likewise.
6520 (call_insn_operand): Likewise.
6521 (sibcall_insn_operand): Likewise.
6522 (GOT32_symbol_operand): Likewise.
6523 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6525 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6527 PR rtl-optimization/83496
6528 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6529 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6530 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6531 redundant insn, if any.
6532 (relax_delay_slots): Likewise.
6533 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6535 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6537 PR tree-optimization/83965
6538 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6539 that grouped statements are part of a reduction chain. Return
6540 true if the statement is not marked as a reduction itself but
6542 (vect_recog_dot_prod_pattern): Don't check whether the statement
6543 is part of a group here.
6544 (vect_recog_sad_pattern): Likewise.
6545 (vect_recog_widen_sum_pattern): Likewise.
6547 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6550 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6552 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6555 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6557 (ix86_output_indirect_function_return): New prototype.
6558 (ix86_split_simple_return_pop_internal): Likewise.
6559 * config/i386/i386.c (indirect_return_via_cx): New.
6560 (indirect_return_via_cx_bnd): Likewise.
6561 (indirect_thunk_name): Handle return va CX_REG.
6562 (output_indirect_thunk_function): Create alias for
6563 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6564 (ix86_output_indirect_jmp): Remove the bool argument.
6565 (ix86_output_indirect_function_return): New function.
6566 (ix86_split_simple_return_pop_internal): Likewise.
6567 * config/i386/i386.md (*indirect_jump): Don't pass false
6568 to ix86_output_indirect_jmp.
6569 (*tablejump_1): Likewise.
6570 (simple_return_pop_internal): Change it to define_insn_and_split.
6571 Call ix86_split_simple_return_pop_internal to split it for
6573 (simple_return_indirect_internal): Call
6574 ix86_output_indirect_function_return instead of
6575 ix86_output_indirect_jmp.
6577 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6580 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6581 memset and value initialization afterwards.
6583 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6585 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6587 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6590 * common/config/aarch64/aarch64-common.c
6591 (aarch_option_optimization_table[]): Switch
6592 off fomit-frame-pointer
6594 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6595 Chung-Ju Wu <jasonwucj@gmail.com>
6597 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6599 (store_multiple): Ditto.
6601 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6603 * config.gcc: Add --with-cpu support for nds32 target.
6604 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6605 * config/nds32/nds32.opt: Add -mcpu= option.
6607 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6609 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6610 isel=yes): Warn for these deprecated options.
6612 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6614 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6615 ISA_2_5_MASKS_EMBEDDED.
6617 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6619 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6620 p->max as pointers rather than using iterative_hash_expr.
6622 2018-02-23 Carl Love <cel@us.ibm.com>
6624 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6625 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6627 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6628 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6629 P8V_BUILTIN_VEC_VUNSIGNED2.
6631 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6634 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6635 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6637 * lra-constraints.c (curr_insn_transform): Set up
6638 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6640 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6641 * lra-eliminations.c (spill_pseudos): Ditto.
6642 (process_insn_for_elimination): Ditto.
6643 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6644 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6645 LRA_NON_CLOBBERED_ALT.
6647 2018-02-22 Martin Sebor <msebor@redhat.com>
6649 PR tree-optimization/84480
6650 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6651 to maybe_diag_stxncpy_trunc. Call it.
6652 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6653 from gimple_fold_builtin_strcpy. Print inlining stack.
6654 (handle_builtin_stxncpy): Print inlining stack.
6655 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6657 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6660 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6661 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6662 and -fcheck-pointer-bounds are used together.
6663 (indirect_thunk_prefix): New enum.
6664 (indirect_thunk_need_prefix): New function.
6665 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6666 "_nt" instead of "_bnd" for NOTRACK prefix.
6667 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6668 (output_indirect_thunk_function): Likewise.
6670 (ix86_code_end): Update output_indirect_thunk_function calls.
6671 (ix86_output_indirect_branch_via_reg): Replace
6672 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6673 (ix86_output_indirect_branch_via_push): Likewise.
6674 (ix86_output_function_return): Likewise.
6675 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6676 incompatible with -fcf-protection=branch and
6677 -fcheck-pointer-bounds.
6679 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6682 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6683 Change gcc_assert call to output_operand_lossage.
6685 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6687 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6689 2018-02-22 DJ Delorie <dj@redhat.com>
6690 Sebastian Perta <sebastian.perta@renesas.com>
6691 Oleg Endo <olegendo@gcc.gnu.org>
6693 * config/rx/rx.c (rx_rtx_costs): New function.
6694 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6696 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6698 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6700 2018-02-22 Martin Liska <mliska@suse.cz>
6703 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6704 Add "native" as a possible value.
6706 2018-02-22 Martin Liska <mliska@suse.cz>
6709 * config/i386/i386.c (ix86_option_override_internal):
6710 Add "native" as a possible value for -march and -mtune.
6712 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6715 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6716 to all type variants.
6718 PR tree-optimization/84503
6719 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6720 width as info->bitpos + info->bitsize - start.
6721 (merged_store_group::merge_overlapping): Simplify width computation.
6722 (check_no_overlap): New function.
6723 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6724 start + width and last_order of the group, fail if check_no_overlap
6726 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6727 to group if check_no_overlap fails.
6729 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6731 * config/rs6000/altivec.md: Delete contraint arguments to
6732 define_expand, define_split, and define_peephole2, and in
6733 define_insn_and_split if always unused.
6734 * config/rs6000/darwin.md: Ditto.
6735 * config/rs6000/dfp.md: Ditto.
6736 * config/rs6000/rs6000.md: Ditto.
6737 * config/rs6000/sync.md: Ditto.
6738 * config/rs6000/vector.md: Ditto.
6739 * config/rs6000/vsx.md: Ditto.
6741 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6743 * config/rs6000/altivec.md: Write output control strings as braced
6744 blocks instead of double-quoted strings.
6745 * config/rs6000/darwin.md: Ditto.
6746 * config/rs6000/rs6000.md: Ditto.
6747 * config/rs6000/vector.md: Ditto.
6748 * config/rs6000/vsx.md: Ditto.
6750 2018-02-21 Jason Merrill <jason@redhat.com>
6752 PR c++/84314 - ICE with templates and fastcall attribute.
6753 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6755 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6757 * ipa-cp.c (determine_versionability): Fix comment typos.
6759 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6762 * ipa-cp.c (determine_versionability): Do not version functions caling
6765 2018-02-21 Martin Liska <mliska@suse.cz>
6768 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6769 Add "native" as a possible value.
6770 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6771 the macro when native cpu detection is available.
6773 2018-02-21 Martin Liska <mliska@suse.cz>
6776 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6777 Add "native" as a possible value.
6778 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6779 when native cpu detection is available.
6781 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6782 Martin Sebor <msebor@redhat.com>
6784 PR tree-optimization/84478
6785 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6787 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6788 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6789 support which is conservatively correct, for 2 only stay conservative
6790 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6791 argument to the 2 argument get_range_strlen, adjust 6 arg
6792 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6794 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6795 (gimple_fold_builtin_strlen): Pass true as last argument to
6798 2018-02-20 Martin Sebor <msebor@redhat.com>
6801 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6802 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6803 (builtin_memref::builtin_memref): Factor out parts into
6804 set_base_and_offset and call it.
6806 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6809 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6810 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6811 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6812 search at the associated MODE_INT.
6814 2018-02-20 Jeff Law <law@redhat.com>
6817 PR tree-optimization/81592
6819 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6820 for range data rather than using global data.
6821 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6822 range data rather than using global data.
6823 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6824 pass it to children as needed.
6825 (struct directive::fmtresult): Similarly.
6826 (struct directive::set_width): Similarly.
6827 (struct directive::set_precision): Similarly.
6828 (format_integer, format_directive, parse_directive): Similarly.
6829 (format_none): Accept unnamed vr_values parameter.
6830 (format_percent, format_floating, format_character): Similarly.
6831 (format_string, format_plain): Similarly.
6832 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6833 the EVRP range analyzer for range data rather than using global data.
6834 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6835 gimple-ssa-evrp-analyze.h
6836 (class sprintf_dom_walker): Add after_dom_children member function.
6837 Add evrp_range_analyzer member.
6838 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6839 range analyzer as needed.
6840 (sprintf_dom_walker::after_dom_children): New member function.
6841 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6843 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6844 (evrp_range_analyzer::pop_to_marker): Likewise.
6846 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6848 PR tree-optimization/84419
6849 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6850 with the required type if its current type is compatible but
6853 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6856 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6857 after vectorization.
6859 2018-02-20 Martin Liska <mliska@suse.cz>
6862 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6863 possible values if we don't have a hint.
6865 2018-02-20 Martin Liska <mliska@suse.cz>
6869 * final.c (shorten_branches): Build align_tab array with one
6871 * opts.c (finish_options): Add alignment option limit check.
6872 (MAX_CODE_ALIGN): Likewise.
6873 (MAX_CODE_ALIGN_VALUE): Likewise.
6874 * doc/invoke.texi: Document maximum allowed option value for
6875 all -falign-* options.
6877 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6880 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6881 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6882 * var-tracking.c (emit_note_insn_var_location): Remove all references
6883 to NOTE_INSN_CALL_ARG_LOCATION.
6884 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6885 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6886 Use copy_rtx_if_shared.
6887 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6888 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6889 (dwarf2out_var_location): Remove handling of
6890 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6892 * final.c (final_scan_insn): Remove all references to
6893 NOTE_INSN_CALL_ARG_LOCATION.
6894 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6895 before dumping final insns.
6896 * except.c (emit_note_eh_region_end): Remove all references to
6897 NOTE_INSN_CALL_ARG_LOCATION.
6898 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6899 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6900 * config/arc/arc.c (hwloop_optimize): Likewise.
6901 * config/arm/arm.c (create_fix_barrier): Likewise.
6902 * config/s390/s390.c (s390_chunkify_start): Likewise.
6903 * config/sh/sh.c (find_barrier): Likewise.
6904 * config/i386/i386.c (rest_of_insert_endbranch,
6905 ix86_seh_fixup_eh_fallthru): Likewise.
6906 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6907 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6908 * config/frv/frv.c (frv_function_prologue): Likewise.
6909 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6911 (note_outside_basic_block_p): Remove all references to
6912 NOTE_INSN_CALL_ARG_LOCATION.
6913 * gengtype.c (adjust_field_rtx_def): Likewise.
6914 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6916 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6917 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6920 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6923 PR tree-optimization/84452
6924 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6925 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6928 2018-02-19 Martin Liska <mliska@suse.cz>
6931 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6933 2018-02-19 Martin Liska <mliska@suse.cz>
6934 Richard Sandiford <richard.sandiford@linaro.org>
6936 PR tree-optimization/82491
6937 * gimple-fold.c (get_base_constructor): Make earlier bail out
6940 2018-02-19 Carl Love <cel@us.ibm.com>
6942 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6943 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6945 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6946 P8V_BUILTIN_VEC_NEG.
6948 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6950 * config/rl78/rl78.md (movdf): New define expand.
6952 2018-02-19 Martin Liska <mliska@suse.cz>
6955 * doc/invoke.texi: Fix typo.
6956 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6958 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6960 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6961 handle rs6000_single_float and rs6000_double_float specially for
6964 2018-02-16 Jeff Law <law@redhat.com>
6966 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6969 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6972 * ipa-inline.c (inline_small_functions): Fix a typo.
6974 2018-02-16 Nathan Sidwell <nathan@acm.org>
6976 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6978 2018-02-16 Carl Love <cel@us.ibm.com>
6980 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6981 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6982 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6983 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6984 expansion to P8V_BUILTIN_VEC_FLOAT2.
6986 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6988 PR rtl-optimization/70023
6989 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6990 src_regno into account.
6992 2018-02-16 Carl Love <cel@us.ibm.com>
6994 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6995 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6996 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6997 * config/rs6000/rs6000.c: Remove case statements for
6998 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6999 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7000 and P9V_BUILTIN_VEC_VINSERT4B.
7001 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7002 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7003 * config/rs6000/vsx.md:
7004 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7007 2018-02-16 Carl Love <cel@us.ibm.com>
7009 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7011 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7013 * config/rs6000/rs6000-c.c: Add the definitions for
7014 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7015 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7016 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7017 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7018 definition for insert4b and define insn *insert3b_internal.
7019 * doc/extend.texi: Add documentation for vec_extract4b.
7021 2018-02-16 Nathan Sidwell <nathan@acm.org>
7023 * doc/extend.texi (Backwards Compatibility): Mention friend
7024 injection. Note for-scope is deprecated.
7025 * doc/invoke.texi (-ffriend-injection): Deprecate.
7027 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7029 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7030 that moved to I2, also allow destinations that are a paradoxical
7031 subreg (instead of a normal reg).
7033 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7036 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7039 2018-02-16 Richard Biener <rguenther@suse.de>
7041 PR tree-optimization/84037
7042 PR tree-optimization/84016
7044 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7045 Adjust vec_construct for the fact we need additional higher latency
7046 128bit inserts for AVX256 and AVX512 vector builds.
7047 (ix86_add_stmt_cost): Scale vector construction cost for
7050 2018-02-16 Richard Biener <rguenther@suse.de>
7052 PR tree-optimization/84417
7053 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7054 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7055 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7057 2018-02-16 Martin Liska <mliska@suse.cz>
7060 * internal-fn.def (ASAN_CHECK): Set proper flags.
7061 (ASAN_MARK): Likewise.
7063 2018-02-16 Julia Koval <julia.koval@intel.com>
7065 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7066 from PTA_CANNONLAKE.
7068 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7071 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7072 Use ++iter rather than iter++ for std::list iterators.
7073 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7074 defer deleting them until all nodes in the forest are processed. Do
7075 free even leaf nodes. Change to_process into auto_vec.
7078 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7079 * vec.h (vec_default_construct): Use memset instead of placement new
7080 if BROKEN_VALUE_INITIALIZATION is defined.
7081 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7082 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7085 PR rtl-optimization/83723
7086 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7087 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7088 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7090 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7092 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7094 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7096 PR rtl-optimization/81443
7097 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7098 from inner REGs to paradoxical SUBREGs.
7100 2018-02-16 Richard Biener <rguenther@suse.de>
7102 PR tree-optimization/84399
7103 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7104 For operands we can analyze at their definition make sure we can
7105 analyze them at each use as well.
7107 2018-02-16 Richard Biener <rguenther@suse.de>
7109 PR tree-optimization/84190
7110 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7111 volatile accesses if the decl isn't volatile.
7113 2018-02-15 Jason Merrill <jason@redhat.com>
7115 PR c++/84314 - ICE with templates and fastcall attribute.
7116 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7117 TYPE_CANONICAL on an existing type.
7119 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7121 PR tree-optimization/84383
7122 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7123 dstoff nor call operand_equal_p if dstbase is NULL.
7125 PR tree-optimization/84334
7126 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7127 also a CONSTANT_CLASS_P, punt.
7129 2018-02-14 Jim Wilson <jimw@sifive.com>
7131 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7132 first SMALL_OPERAND check. New local min_second_step. Move assert
7133 to where locals are set. Add TARGET_RVC support.
7134 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7136 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7138 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7140 2018-02-14 Martin Sebor <msebor@redhat.com>
7142 PR tree-optimization/83698
7143 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7144 arrays constrain the offset range to their bounds.
7145 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7146 (builtin_access::overlap): Avoid setting the size of overlap if it's
7148 (maybe_diag_overlap): Also consider arrays when deciding what values
7149 of offsets to include in diagnostics.
7151 2018-02-14 Martin Sebor <msebor@redhat.com>
7154 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7155 that correspond to the kind of a declaration.
7157 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7160 * config/pa/pa.md: Load address of PIC label using the linkage table
7161 if the label is nonlocal.
7163 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7165 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7166 warning message if user requests -maltivec=be.
7167 * doc/invoke.texi: Document deprecation of -maltivec=be.
7169 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7172 * config/rs6000/rs6000-c.c: Update definitions for
7173 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7174 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7176 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7179 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7180 add _get_ssp intrinsics. Remove argument from
7181 __builtin_ia32_rdssp[d|q].
7182 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7183 * config/i386/i386-builtin.def: Remove argument from
7184 __builtin_ia32_rdssp[d|q].
7185 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7186 ix86_expand_special_args_builtin for _rdssp[d|q].
7187 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7188 Clear register before usage.
7189 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7190 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7192 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7194 PR tree-optimization/84357
7195 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7196 operand 1 of an ARRAY_REF too.
7198 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7201 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7202 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7204 (set_of_reg): New struct.
7205 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7206 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7207 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7209 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7210 Split into bitclr, bitset, bitinvert patterns if appropriate.
7211 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7212 use rx_fuse_in_memory_bitop.
7213 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7214 to named insn, correct maximum insn length.
7216 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7219 * machmode.def: Define a complex mode for PARTIAL_INT.
7220 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7222 * doc/rtl.texi: Document CSPImode.
7223 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7225 (msp430_hard_regno_nregs_with_padding): Likewise.
7227 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7230 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7232 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7234 PR rtl-optimization/84169
7235 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7236 we generated a parallel as new i3 and we split that to new i2 and i3
7237 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7238 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7239 those to i2, not i1. Partially rewrite this scan code.
7241 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7244 * stor-layout.c (place_field): For variable length fields, adjust
7245 offset_align afterwards not just based on the field's alignment,
7246 but also on the size.
7249 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7250 of exps and logs in the use_exp2 case.
7252 2018-02-13 Jeff Law <law@redhat.com>
7254 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7257 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7260 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7264 * common.opt (gas-loc-support, gas-locview-support): New.
7265 (ginline-points, ginternal-reset-location-views): New.
7266 * doc/invoke.texi: Document them. Use @itemx where intended.
7267 (gvariable-location-views): Adjust.
7268 * target.def (reset_location_view): New.
7269 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7270 (TARGET_RESET_LOCATION_VIEW): New.
7271 * doc/tm.texi: Rebuilt.
7272 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7273 (dwarf2out_default_as_locview_support): New.
7274 (output_asm_line_debug_info): Use option variables.
7275 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7276 (output_loc_list): Likewise.
7277 (add_high_low_attributes): Check option variables.
7278 Don't output entry view attribute in strict mode.
7279 (gen_inlined_subroutine_die): Check option variables.
7280 (dwarf2out_inline_entry): Likewise.
7281 (init_sections_and_labels): Likewise.
7282 (dwarf2out_early_finish): Likewise.
7283 (maybe_reset_location_view): New, from...
7284 (dwarf2out_var_location): ... here. Call it.
7285 * debug.h (dwarf2out_default_as_loc_support): Declare.
7286 (dwarf2out_default_as_locview_support): Declare.
7287 * hooks.c (hook_int_rtx_insn_0): New.
7288 * hooks.h (hook_int_rtx_insn_0): Declare.
7289 * toplev.c (process_options): Take -gas-loc-support and
7290 -gas-locview-support from dwarf2out. Enable
7291 -gvariable-location-views by default only with locview
7292 assembler support. Enable -ginternal-reset-location-views by
7293 default only if the target defines the corresponding hook.
7294 Enable -ginline-points by default if location views are
7295 enabled; force it disabled if statement frontiers are
7297 * tree-inline.c (expand_call_inline): Check option variables.
7298 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7300 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7302 PR tree-optimization/84321
7303 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7304 handling. Also check whether the anti-range contains any values
7305 that satisfy the mask; switch to a VR_RANGE if not.
7307 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7310 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7312 2018-02-13 Martin Jambor <mjambor@suse.cz>
7315 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7316 of call statements, also set location of a load to a temporary.
7318 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7320 * config/rl78/rl78.c (add_vector_labels): New function.
7321 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7322 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7323 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7324 which checks that no arguments are passed.
7325 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7326 * doc/extend.texi: Documentation for the new attribute.
7328 2018-02-13 Andreas Schwab <schwab@suse.de>
7330 * config/riscv/linux.h (CPP_SPEC): Define.
7332 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7335 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7336 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7337 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7338 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7339 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7340 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7341 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7342 temporarily for AES and PCLMUL builtins.
7344 PR tree-optimization/84339
7345 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7346 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7350 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7351 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7352 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7354 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7356 * omp-simd-clone.h: New file.
7357 * omp-simd-clone.c: Include omp-simd-clone.h.
7358 (expand_simd_clones): No longer static.
7359 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7360 cgraph.h and omp-simd-clone.h.
7361 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7362 (vect_recog_widen_shift_pattern): Formatting fix.
7363 (vect_pattern_recog_1): Don't check optab for calls.
7366 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7367 operands[2] into a REG before using gen_lowpart on it.
7369 2018-02-12 Jeff Law <law@redhat.com>
7372 * config/sh/sh.c (find_barrier): Consider a sibling call
7375 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7376 successfully back substituting a reg.
7378 2018-02-12 Richard Biener <rguenther@suse.de>
7380 PR tree-optimization/84037
7381 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7382 parameter, move visited init to caller.
7383 (vect_slp_analyze_operations): Separate cost from validity
7384 check, initialize visited once for all instances.
7385 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7387 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7389 (vect_model_promotion_demotion_cost): Likewise.
7390 (vectorizable_bswap): Guard cost modeling with !slp_node
7391 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7393 (vectorizable_call): Likewise.
7394 (vectorizable_conversion): Likewise.
7395 (vectorizable_assignment): Likewise.
7396 (vectorizable_shift): Likewise.
7397 (vectorizable_operation): Likewise.
7398 (vectorizable_store): Likewise.
7399 (vectorizable_load): Likewise.
7400 (vectorizable_condition): Likewise.
7401 (vectorizable_comparison): Likewise.
7403 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7406 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7407 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7408 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7410 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7413 * params.def (inline-min-speedup): Increase from 8 to 15.
7414 (max-inline-insns-auto): Decrease from 40 to 30.
7415 * ipa-split.c (consider_split): Add some buffer for function to
7416 be considered inlining candidate.
7417 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7420 2018-02-12 Richard Biener <rguenther@suse.de>
7422 PR tree-optimization/84037
7423 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7424 matched stmts if we cannot swap the non-matched ones.
7426 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7428 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7429 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7430 _mm_maskz_scalef_round_ss): New intrinsics.
7431 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7432 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7433 __builtin_ia32_scalefss_round): Remove.
7434 (__builtin_ia32_scalefsd_mask_round,
7435 __builtin_ia32_scalefss_mask_round): New intrinsics.
7436 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7437 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7438 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7439 "<round_constraint>")): Changed to ...
7440 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7441 "<round_scalar_constraint>")): ... this.
7442 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7443 %0, %1, %2<round_op3>}"): Changed to ...
7444 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7445 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7446 %2<round_scalar_mask_op3>}"): ... this.
7447 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7449 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7451 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7452 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7453 (_mm_maskz_sqrt_round_ss): New intrinsics.
7454 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7455 (__builtin_ia32_sqrtsd_mask_round)
7456 (__builtin_ia32_sqrtss_mask_round): New builtins.
7457 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7458 (__builtin_ia32_sqrtss_round): Remove.
7459 (__builtin_ia32_sqrtsd_mask_round)
7460 (__builtin_ia32_sqrtss_mask_round): New builtins.
7461 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7462 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7463 ((match_operand:VF_128 1 "vector_operand"
7464 "xBm,<round_constraint>")): Changed to ...
7465 ((match_operand:VF_128 1 "vector_operand"
7466 "xBm,<round_scalar_constraint>")): ... this.
7467 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7468 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7469 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7470 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7471 %<iptr>1<round_scalar_mask_op3>}): ... this.
7472 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7473 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7475 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7478 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7479 Cast vec_cmpeq result to correct type.
7480 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7481 Cast vec_cmpgt result to correct type.
7483 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7485 * final.c (final_scan_insn_1): Renamed from...
7486 (final_scan_insn): ... this. New wrapper, to recover
7487 seen from the outermost call in recursive ones.
7488 * config/sparc/sparc.c (output_return): Drop seen from call.
7489 (output_sibcall): Likewise.
7490 * config/visium/visium.c (output_branch): Likewise.
7492 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7494 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7497 2018-02-10 Alan Modra <amodra@gmail.com>
7500 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7501 Specify LR as an input.
7503 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7506 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7507 remove_member_access_dummy_vars): New functions.
7508 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7509 lower_omp_1, execute_lower_omp): Use them.
7511 PR rtl-optimization/84308
7512 * shrink-wrap.c (spread_components): Release todo vector.
7514 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7516 PR rtl-optimization/57193
7517 * ira-color.c (struct allocno_color_data): Add member
7518 conflict_allocno_hard_prefs.
7519 (update_conflict_allocno_hard_prefs): New.
7520 (bucket_allocno_compare_func): Add a preference based on
7521 conflict_allocno_hard_prefs.
7522 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7523 (color_allocnos): Remove a dead code. Initiate
7524 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7526 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7529 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7530 constraint from =wa to wa. Avoid a subreg on the output operand,
7531 instead use a pseudo and subreg it in a move.
7532 (p9_xxbrd_<mode>): Changed to ...
7533 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7534 (p9_xxbrd_v2df): New expander.
7535 (p9_xxbrw_<mode>): Changed to ...
7536 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7537 (p9_xxbrw_v4sf): New expander.
7539 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7541 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7543 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7546 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7547 multiply in 32-bit mode.
7548 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7549 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7552 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7554 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7555 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7556 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7557 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7559 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7562 * dwarf2out.c (is_trivial_indirect_ref): New function.
7563 (dwarf2out_late_global_decl): Do not generate a location
7564 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7565 and that are not defined in the current unit.
7567 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7569 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7570 instead of a libcall for UNORDERED.
7572 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7575 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7576 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7578 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7581 * config/s390/s390.c (s390_set_current_function): Invoke
7582 s390_indirect_branch_settings also if fndecl didn't change.
7584 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7586 * config/rs6000/rs6000.md (blockage): Set length to zero.
7588 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7590 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7592 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7595 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7596 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7600 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7601 PARALLEL incoming that failed vt_get_decl_and_offset check.
7604 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7605 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7607 (get_variable_section): For decls in named .bss* sections pass true as
7608 second argument to bss_initializer_p.
7610 2018-02-09 Marek Polacek <polacek@redhat.com>
7611 Jakub Jelinek <jakub@redhat.com>
7614 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7615 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7616 Sync some changes from cxx_fold_indirect_ref.
7618 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7620 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7622 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7623 (BLOCK_INLINE_ENTRY_LABEL): New.
7624 (dwarf2out_var_location): Disregard inline entry markers.
7625 (inline_entry_data): New struct.
7626 (inline_entry_data_hasher): New hashtable type.
7627 (inline_entry_data_hasher::hash): New.
7628 (inline_entry_data_hasher::equal): New.
7629 (inline_entry_data_table): New variable.
7630 (add_high_low_attributes): Add DW_AT_entry_pc and
7631 DW_AT_GNU_entry_view attributes if a pending entry is found
7632 in inline_entry_data_table. Add old entry_pc attribute only
7633 if debug nonbinding markers are disabled.
7634 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7635 markers are enabled.
7636 (block_within_block_p, dwarf2out_inline_entry): New.
7637 (dwarf2out_finish): Check that no entries remained in
7638 inline_entry_data_table.
7639 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7640 (final_scan_insn, notice_source_line): Likewise.
7641 (rest_of_clean_state): Skip inline entry markers.
7642 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7644 * gimple.c (gimple_build_debug_inline_entry): New.
7645 * gimple.h (enum gimple_debug_subcode): Add
7646 GIMPLE_DEBUG_INLINE_ENTRY.
7647 (gimple_build_debug_inline_entry): Declare.
7648 (gimple_debug_inline_entry_p): New.
7649 (gimple_debug_nonbind_marker_p): Adjust.
7650 * insn-notes.def (INLINE_ENTRY): New.
7651 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7652 inline entry marker notes.
7653 (print_insn): Likewise.
7654 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7655 (INSN_DEBUG_MARKER_KIND): Likewise.
7656 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7657 * tree-inline.c (expand_call_inline): Build and insert
7658 debug_inline_entry stmt.
7659 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7660 inline entry blocks early, if nonbind markers are enabled.
7661 (dump_scope_block): Dump fragment info.
7662 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7663 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7664 (gimple_build_debug_inline_entry): New.
7665 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7666 Enable/disable inline entry points too.
7667 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7668 (DEBUG_INSN): Describe inline entry markers.
7670 * common.opt (gvariable-location-views): New.
7671 (gvariable-location-views=incompat5): New.
7672 * config.in: Rebuilt.
7673 * configure: Rebuilt.
7674 * configure.ac: Test assembler for view support.
7675 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7676 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7677 * dwarf2out.c (var_loc_view): New typedef.
7678 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7679 (dwarf2out_locviews_in_attribute): New.
7680 (dwarf2out_locviews_in_loclist): New.
7681 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7682 (enum dw_line_info_opcode): Add LI_adv_address.
7683 (struct dw_line_info_table): Add view.
7684 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7685 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7686 (zero_view_p): New variable.
7687 (ZERO_VIEW_P): New macro.
7688 (output_asm_line_debug_info): New.
7689 (struct var_loc_node): Add view.
7690 (add_AT_view_list, AT_loc_list): New.
7691 (add_var_loc_to_decl): Add view param. Test it against last.
7692 (new_loc_list): Add view params. Record them.
7693 (AT_loc_list_ptr): Handle loc and view lists.
7694 (view_list_to_loc_list_val_node): New.
7695 (print_dw_val): Handle dw_val_class_view_list.
7696 (size_of_die): Likewise.
7697 (value_format): Likewise.
7698 (loc_list_has_views): New.
7699 (gen_llsym): Set vl_symbol too.
7700 (maybe_gen_llsym, skip_loc_list_entry): New.
7701 (dwarf2out_maybe_output_loclist_view_pair): New.
7702 (output_loc_list): Output view list or entries too.
7703 (output_view_list_offset): New.
7704 (output_die): Handle dw_val_class_view_list.
7705 (output_dwarf_version): New.
7706 (output_compilation_unit_header): Use it.
7707 (output_skeleton_debug_sections): Likewise.
7708 (output_rnglists, output_line_info): Likewise.
7709 (output_pubnames, output_aranges): Update version comments.
7710 (output_one_line_info_table): Output view numbers in asm comments.
7711 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7712 Call maybe_gen_llsym.
7713 (loc_list_from_tree_1): Adjust.
7714 (add_AT_location_description): Create view list attribute if
7715 needed, check it's absent otherwise.
7716 (convert_cfa_to_fb_loc_list): Adjust.
7717 (maybe_emit_file): Call output_asm_line_debug_info for test.
7718 (dwarf2out_var_location): Reset views as needed. Precompute
7719 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7720 attribute. Set view.
7721 (new_line_info_table): Reset next view.
7722 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7723 (dwarf2out_source_line): Likewise. Output view resets and labels to
7724 the assembler, or select appropriate line info opcodes.
7725 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7726 (optimize_string_length): Catch it. Adjust.
7727 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7728 dw_val_class_view_list, and remove it if no longer needed.
7729 (hash_loc_list): Hash view numbers.
7730 (loc_list_hasher::equal): Compare them.
7731 (optimize_location_lists): Check whether a view list symbol is
7732 needed, and whether the locview attribute is present, and
7733 whether they match. Remove the locview attribute if no longer
7735 (index_location_lists): Call skip_loc_list_entry for test.
7736 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7737 Use output_dwarf_version.
7738 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7739 (struct dw_val_node): Add val_view_list.
7740 * final.c (SEEN_NEXT_VIEW): New.
7741 (set_next_view_needed): New.
7742 (clear_next_view_needed): New.
7743 (maybe_output_next_view): New.
7744 (final_start_function): Rename to...
7745 (final_start_function_1): ... this. Take pointer to FIRST,
7746 add SEEN parameter. Emit param bindings in the initial view.
7747 (final_start_function): Reintroduce SEEN-less interface.
7748 (final): Rename to...
7749 (final_1): ... this. Take SEEN parameter. Output final pending
7750 next view at the end.
7751 (final): Reintroduce seen-less interface.
7752 (final_scan_insn): Output pending next view before switching
7753 sections or ending a block. Mark the next view as needed when
7754 outputting variable locations. Notify debug backend of section
7755 changes, and of location view changes.
7756 (rest_of_handle_final): Adjust.
7757 * toplev.c (process_options): Autodetect value for debug variable
7758 location views option. Warn on incompat5 without -gdwarf-5.
7759 * doc/invoke.texi (gvariable-location-views): New.
7760 (gvariable-location-views=incompat5): New.
7761 (gno-variable-location-views): New.
7763 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7765 PR tree-optimization/84136
7766 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7767 that the result of find_edge is non-NULL.
7769 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7772 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7773 storing integer register in SImode. Fix cost of 256 and 512
7774 byte aligned SSE register store.
7776 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7778 * config/i386/i386.c (ix86_multiplication_cost): Fix
7779 multiplication cost for TARGET_AVX512DQ.
7781 2018-02-08 Marek Polacek <polacek@redhat.com>
7783 PR tree-optimization/84238
7784 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7787 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7789 PR tree-optimization/84265
7790 * tree-vect-stmts.c (vectorizable_store): Don't treat
7791 VMAT_CONTIGUOUS accesses as grouped.
7792 (vectorizable_load): Likewise.
7794 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7796 PR tree-optimization/81635
7797 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7798 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7799 (test_round_for_mask): New functions.
7800 (wide_int_cc_tests): Call test_round_for_mask.
7801 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7802 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7803 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7804 range returned by get_range_info.
7806 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7809 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7810 * symtab.c: Include builtins.h
7811 (symtab_node::output_to_lto_symbol_table_p): Move here
7812 from lto-streamer-out.c:output_symbol_p.
7813 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7814 (output_symbol_p): Move all logic to symtab.c
7815 (produce_symtab): Update.
7817 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7819 * config/s390/s390-opts.h (enum indirect_branch): Define.
7820 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7821 (s390_indirect_branch_via_thunk)
7822 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7823 (enum s390_indirect_branch_type): Define.
7824 * config/s390/s390.c (struct s390_frame_layout, struct
7825 machine_function): Remove.
7826 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7827 (indirect_branch_table_label_no, indirect_branch_table_name):
7829 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7830 (enum s390_indirect_branch_option): Define.
7831 (s390_return_addr_from_memory): New function.
7832 (s390_handle_string_attribute): New function.
7833 (s390_attribute_table): Add new attribute handler.
7834 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7835 (s390_indirect_branch_via_thunk): New function.
7836 (s390_indirect_branch_via_inline_thunk): New function.
7837 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7838 sibling call optimization for non z10 compiles.
7839 (s390_emit_call): Force indirect branch target to be a single
7840 register. Add r1 clobber for non-z10 compiles.
7841 (s390_emit_epilogue): Emit return jump via return_use expander.
7842 (s390_reorg): Handle JUMP_INSNs as execute targets.
7843 (s390_option_override_internal): Perform validity checks for the
7844 new command line options.
7845 (s390_indirect_branch_attrvalue): New function.
7846 (s390_indirect_branch_settings): New function.
7847 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7848 (s390_output_indirect_thunk_function): New function.
7849 (s390_code_end): Implement target hook.
7850 (s390_case_values_threshold): Implement target hook.
7851 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7853 * config/s390/s390.h (struct s390_frame_layout)
7854 (struct machine_function): Move here from s390.c.
7855 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7856 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7857 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7858 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7859 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7860 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7861 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7862 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7863 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7864 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7865 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7866 (mnemonic attribute): Add values which aren't recognized
7868 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7869 pattern for branch conversion. Fix mnemonic attribute.
7870 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7871 indirect branch via thunk if requested.
7872 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7873 ("*indirect_jump"): Disable for branch conversion using out of
7875 ("indirect_jump_via_thunk<mode>_z10")
7876 ("indirect_jump_via_thunk<mode>")
7877 ("indirect_jump_via_inlinethunk<mode>_z10")
7878 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7879 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7880 ("casesi_jump_via_inlinethunk<mode>_z10")
7881 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7882 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7883 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7884 ("*indirect2_jump"): Disable for branch conversion.
7885 ("casesi_jump"): Turn into expander and expand patterns for branch
7887 ("return_use"): New expander.
7888 ("*return"): Emit return via thunk and rename it to ...
7889 ("*return<mode>"): ... this one.
7890 * config/s390/s390.opt: Add new options and and enum for the
7893 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7895 * lra-constraints.c (match_reload): Unconditionally use
7896 gen_lowpart_SUBREG, rather than selecting between that
7897 and equivalent gen_rtx_SUBREG code.
7899 2018-02-08 Richard Biener <rguenther@suse.de>
7901 PR tree-optimization/84233
7902 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7903 changed flag instead of boguously re-using phi_inserted.
7905 2018-02-08 Martin Jambor <mjambor@suse.cz>
7907 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7908 static local variables.
7910 2018-02-08 Richard Biener <rguenther@suse.de>
7912 PR tree-optimization/84278
7913 * tree-vect-stmts.c (vectorizable_store): When looking for
7914 smaller vector types to perform grouped strided loads/stores
7915 make sure the mode is supported by the target.
7916 (vectorizable_load): Likewise.
7918 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7920 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7921 Increase LDP/STP opportunities by adding adjacent callee-saves.
7923 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7925 PR rtl-optimization/84068
7926 PR rtl-optimization/83459
7927 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7929 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7931 PR tree-optimization/84224
7932 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7933 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7936 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7939 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7940 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7941 count, remove one USE.
7943 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7945 * doc/install.texi (Configuration): Document the
7946 --with-long-double-format={ibm,ieee} PowerPC configuration
7950 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7951 Convert from define_expand to be define_insn_and_split. Rework
7952 float/double/_Float128 conversions to QI/HI/SImode to work with
7953 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7954 conversions to QI/HImode types did a store and then a load to
7955 truncate the value. For conversions to VSX registers, don't split
7956 the insn, instead emit the code directly. Use the code iterator
7957 any_fix to combine signed and unsigned conversions.
7958 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7959 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7960 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7961 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7962 (fix_<mode>di2_hw): Likewise.
7963 (fixuns_<mode>di2_hw): Likewise.
7964 (fix_<mode>si2_hw): Likewise.
7965 (fixuns_<mode>si2_hw): Likewise.
7966 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7967 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7968 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7969 fix<uns>_trunc<SFDF:mode>si2_p8.
7970 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7972 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7973 (fix<uns>_<mode>_mem): Likewise.
7974 (fctiw<u>z_<mode>_mem): Likewise.
7975 (fix<uns>_<mode>_mem): Likewise.
7976 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7977 the register allocator from doing a direct move to the GPRs to do
7978 a store, and instead use the ISA 3.0 store byte/half-word from
7979 vector register instruction. For IEEE 128-bit floating point,
7980 also optimize stores of 32-bit ints.
7981 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7983 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7985 * genextract.c (push_pathstr_operand): New function to support
7987 (walk_rtx): Call push_pathstr_operand.
7988 (print_path): Support [a-zA-Z].
7990 2018-02-07 Richard Biener <rguenther@suse.de>
7992 PR tree-optimization/84037
7993 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7994 (cse_and_gimplify_to_preheader): Declare.
7995 (vect_get_place_in_interleaving_chain): Likewise.
7996 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7998 (_loop_vec_info::~_loop_vec_info): Delete it.
7999 (cse_and_gimplify_to_preheader): New function.
8000 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8001 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8002 (vectorizable_load): Likewise. For grouped stores always base
8003 the IV on the first element.
8004 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8005 condition before gimplifying.
8007 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8009 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8010 *DIV_EXPR and *MOD_EXPR.
8012 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8015 * config/i386/i386.c (ix86_option_override_internal): Mask out
8016 the CF_SET bit when checking -fcf-protection.
8018 2018-02-07 Tom de Vries <tom@codesourcery.com>
8021 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8024 2018-02-07 Richard Biener <rguenther@suse.de>
8026 PR tree-optimization/84204
8027 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8030 PR tree-optimization/84205
8031 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8032 special-case isl_ast_op_zdiv_r.
8034 PR tree-optimization/84223
8035 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8036 Only add conditions from within the region.
8037 (gather_bbs::after_dom_children): Adjust.
8039 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8042 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8043 * config/avr/avr.md: Only post-reload split REG-REG moves if
8044 either register is GENERAL_REG_P.
8046 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8048 PR tree-optimization/84235
8049 * tree-ssa-scopedtables.c
8050 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8051 if the subtraction is performed in floating point type where NaNs are
8052 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8053 build 1. Formatting fix.
8055 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8058 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8059 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8060 and skip it regardless of bb boundaries. Use CALL_P macro,
8061 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8062 unnecessarily, formatting fix.
8064 2018-02-06 Michael Collison <michael.collison@arm.com>
8066 * config/arm/thumb2.md:
8067 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8068 (*thumb_mov_notscc): Ditto.
8070 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8073 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8074 unsigned_fix, not "s".
8076 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8078 * configure.ac (gcc_fn_eh_frame_ro): New function.
8079 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8080 correct .eh_frame permissions.
8081 * configure: Regenerate.
8083 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8085 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8088 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8090 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8091 Display warning message for -mno-speculate-indirect-jumps.
8093 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8095 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8097 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8099 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8101 PR tree-optimization/84225
8102 * tree-eh.c (find_trapping_overflow): Only call
8103 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8105 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8108 * config/i386/i386.c: Reimplement the check of possible options
8109 -mibt/-mshstk conbination. Change error messages.
8110 * doc/invoke.texi: Fix a typo: remove extra '='.
8112 2018-02-06 Marek Polacek <polacek@redhat.com>
8114 PR tree-optimization/84228
8115 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8117 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8120 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8121 emitted arch directives.
8122 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8123 __ARM_FEATURE_COPROC before changing architectures.
8125 2018-02-06 Richard Biener <rguenther@suse.de>
8127 * config/i386/i386.c (print_reg): Fix typo.
8128 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8130 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8132 * configure: Regenerate.
8134 2018-02-05 Martin Sebor <msebor@redhat.com>
8136 PR tree-optimization/83369
8137 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8140 2018-02-05 Martin Liska <mliska@suse.cz>
8142 * doc/invoke.texi: Cherry-pick upstream r323995.
8144 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8146 * ira.c (ira_init_register_move_cost): Adjust comment.
8148 2018-02-05 Martin Liska <mliska@suse.cz>
8150 PR gcov-profile/84137
8151 * doc/gcov.texi: Fix typo in documentation.
8153 2018-02-05 Martin Liska <mliska@suse.cz>
8155 PR gcov-profile/83879
8156 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8157 using dlopen functionality.
8159 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8161 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8162 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8163 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8164 _mm_maskz_range_round_ss): New intrinsics.
8165 (__builtin_ia32_rangesd128_round)
8166 (__builtin_ia32_rangess128_round): Remove.
8167 (__builtin_ia32_rangesd128_mask_round,
8168 __builtin_ia32_rangess128_mask_round): New builtins.
8169 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8170 __builtin_ia32_rangess128_round): Remove.
8171 (__builtin_ia32_rangesd128_mask_round,
8172 __builtin_ia32_rangess128_mask_round): New builtins.
8173 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8174 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8175 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8176 "<round_saeonly_constraint>")): Changed to ...
8177 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8178 "<round_saeonly_scalar_constraint>")): ... this.
8179 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8180 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8181 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8182 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8183 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8185 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8187 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8189 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8190 Remove all values except native, 8540 and 8548.
8192 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8194 * config/i386/i386.c (ix86_output_function_return): Pass
8195 INVALID_REGNUM, instead of -1, as invalid register number to
8196 indirect_thunk_name and output_indirect_thunk.
8198 2018-02-02 Julia Koval <julia.koval@intel.com>
8200 * config.gcc: Add -march=icelake.
8201 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8202 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8203 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8204 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8205 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8206 (processor_target_table): Add icelake.
8207 (ix86_option_override_internal): Handle new PTAs.
8208 (get_builtin_code_for_version): Handle icelake.
8209 (M_INTEL_COREI7_ICELAKE): New.
8210 (fold_builtin_cpu): Handle icelake.
8211 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8212 * doc/invoke.texi: Add -march=icelake.
8214 2018-02-02 Julia Koval <julia.koval@intel.com>
8216 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8217 to wide_int_bitmask.
8218 * wide-int-bitmask.h: New.
8220 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8223 * config/i386/i386.md: Replace Pmode with word_mode in
8224 builtin_setjmp_setup and builtin_longjmp to support x32.
8226 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8230 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8232 (rs6000_supported_cpu_names): New static variable.
8233 (linux_cpu_translation_table): Likewise.
8234 (elf_platform) <cpu>: Define new static variable and use it.
8235 Translate kernel AT_PLATFORM name to canonical name if needed.
8236 Error if platform name is unknown.
8238 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8241 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8243 2018-02-01 Jeff Law <law@redhat.com>
8246 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8247 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8248 the scratch if RELEASE_VIA_POP is false.
8249 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8250 If we have to save a temporary register, decrement SIZE appropriately.
8251 Pass new arguments to release_scratch_register_on_entry.
8252 (ix86_adjust_stack_and_probe): Likewise.
8253 (ix86_emit_probe_stack_range): Pass new arguments to
8254 release_scratch_register_on_entry.
8256 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8258 PR rtl-optimization/84157
8259 * combine.c (change_zero_ext): Use REG_P predicate in
8260 front of HARD_REGISTER_P predicate.
8262 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8264 * config/avr/avr.c (avr_option_override): Move disabling of
8265 -fdelete-null-pointer-checks to...
8266 * common/config/avr/avr-common.c (avr_option_optimization_table):
8269 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8271 PR tree-optimization/81635
8272 * tree-data-ref.c (split_constant_offset_1): For types that
8273 wrap on overflow, try to use range info to prove that wrapping
8276 2018-02-01 Renlin Li <renlin.li@arm.com>
8279 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8281 (aarch64_register_move_cost): Likewise.
8282 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8284 (REG_CLASS_NAMES): Likewise.
8285 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8286 TAILCALL_ADDR_REGS. Remove IP registers.
8287 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8289 2018-02-01 Richard Biener <rguenther@suse.de>
8291 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8292 for specifying RPO order and allow NULL for that.
8293 * domwalk.c (dom_walker::dom_walker): Likewise.
8294 (dom_walker::walk): Handle NULL RPO order.
8295 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8297 (rewrite_update_dom_walker): Likewise.
8298 (mark_def_dom_walker): Likewise.
8300 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8302 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8303 (aarch64_maybe_expand_sve_subreg_move): Declare.
8304 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8305 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8307 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8308 that are semantically a reverse operation.
8309 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8310 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8311 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8313 (aarch64_can_change_mode_class): For big-endian, forbid changes
8314 between two SVE modes if they have different element sizes.
8316 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8318 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8319 the TImode handling for big-endian targets.
8321 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8323 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8324 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8326 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8327 Remove BSWAP handing for big-endian targets and use the form of
8328 LD1RQ appropariate for the mode.
8330 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8332 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8333 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8336 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8339 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8340 check for operands that need to go through aarch64_sve_reload_be.
8342 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8344 PR tree-optimization/81661
8345 PR tree-optimization/84117
8346 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8347 * tree-eh.c: Include gimplify.h.
8348 (find_trapping_overflow, replace_trapping_overflow,
8349 rewrite_to_non_trapping_overflow): New functions.
8350 * tree-vect-loop.c: Include tree-eh.h.
8351 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8352 * tree-data-ref.c: Include tree-eh.h.
8353 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8355 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8357 PR rtl-optimization/84123
8358 * combine.c (change_zero_ext): Check if hard register satisfies
8359 can_change_dest_mode before calling gen_lowpart_SUBREG.
8361 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8364 * ira.c (ira_init_register_move_cost): Remove assert.
8366 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8368 PR rtl-optimization/84071
8369 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8370 * doc/tm.texi: Regenerate.
8372 2018-01-31 Richard Biener <rguenther@suse.de>
8374 PR tree-optimization/84132
8375 * tree-data-ref.c (analyze_miv_subscript): Properly
8376 check whether evolution_function_is_affine_multivariate_p
8377 before calling gcd_of_steps_may_divide_p.
8379 2018-01-31 Julia Koval <julia.koval@intel.com>
8382 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8383 * config/i386/i386.md (rdpid_rex64) New.
8384 (rdpid): Make 32bit only.
8386 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8389 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8390 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8392 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8395 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8397 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8399 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8401 PR rtl-optimization/84071
8402 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8403 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8405 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8407 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8408 (arc_attribute_table): Add 'aux' attribute.
8409 (arc_in_small_data_p): Consider aux like variables.
8410 (arc_is_aux_reg_p): New function.
8411 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8412 (arc_get_aux_arg): New function.
8413 (prepare_move_operands): Handle aux-register access.
8414 (arc_handle_aux_attribute): New function.
8415 * doc/extend.texi (ARC Variable attributes): Add subsection.
8417 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8419 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8420 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8421 (arc_attribute_table): Add 'uncached' attribute.
8422 (arc_print_operand): Print '.di' flag for uncached memory
8424 (arc_in_small_data_p): Do not consider for small data the uncached
8426 (arc_is_uncached_mem_p): New function.
8427 * config/arc/predicates.md (compact_store_memory_operand): Check
8428 for uncached memory accesses.
8429 (nonvol_nonimm_operand): Likewise.
8430 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8432 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8435 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8436 falign-loops=): Add Optimization flag.
8438 2018-01-30 Jeff Law <law@redhat.com>
8441 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8442 INT_REGISTERS_SAVED. Check it prior to calling
8443 get_scratch_register_on_entry.
8444 (ix86_adjust_stack_and_probe): Similarly.
8445 (ix86_emit_probe_stack_range): Similarly.
8446 (ix86_expand_prologue): Corresponding changes.
8448 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8451 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8452 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8454 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8457 * lra-constraints.c (curr_insn_transform): Process AND in the
8460 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8462 PR rtl-optimization/83986
8463 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8464 dependence against last_pending_memory_flush in addition to
8467 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8469 PR tree-optimization/81611
8470 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8473 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8476 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8479 2018-01-30 Richard Biener <rguenther@suse.de>
8480 Jakub Jelinek <jakub@redhat.com>
8482 PR tree-optimization/84111
8483 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8484 inner loops added during recursion, as they don't have up-to-date
8487 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8490 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8491 (can_inline_edge_by_limits_p): ... here.
8492 (can_early_inline_edge_p, check_callers,
8493 update_caller_keys, update_callee_keys, recursive_inlining,
8494 add_new_edges_to_heap, speculation_useful_p,
8495 inline_small_functions,
8496 inline_small_functions, flatten_function,
8497 inline_to_all_callers_1): Update.
8499 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8501 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8504 2018-01-30 Richard Biener <rguenther@suse.de>
8506 PR tree-optimization/83008
8507 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8508 invariant and constant vector uses in stmts when they need
8511 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8514 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8515 * configure: Regenerate.
8517 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8519 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8521 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8522 Use gen_rtx_REG rather than gen_lowpart.
8524 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8526 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8527 rather than 0 when creating partial subregs.
8529 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8531 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8534 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8537 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8538 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8539 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8540 flags. This restores the settings used before the 2017-07-24.
8541 Turning off pre increment/decrement/modify allows IVOPTS to
8542 optimize DF/SF loops where the index is an int.
8544 2018-01-29 Richard Biener <rguenther@suse.de>
8545 Kelvin Nilsen <kelvin@gcc.gnu.org>
8548 * tree-vect-stmts.c (vectorizable_call): Don't call
8549 targetm.vectorize_builtin_md_vectorized_function if callee is
8552 2018-01-22 Carl Love <cel@us.ibm.com>
8554 * doc/extend.tex: Fix typo in second arg in
8555 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8557 2018-01-29 Richard Biener <rguenther@suse.de>
8559 PR tree-optimization/84086
8560 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8561 (flush_ssaname_freelist): When SSA names were released reset
8562 the SCEV hash table.
8564 2018-01-29 Richard Biener <rguenther@suse.de>
8566 PR tree-optimization/84057
8567 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8568 removed paths when removing edges.
8570 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8572 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8573 -mfunction-return=@var{choice}.
8575 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8578 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8580 (layout::print_source_line): Likewise.
8581 (test_get_line_width_without_trailing_whitespace): Add test cases.
8583 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8586 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8589 2018-01-26 Jim Wilson <jimw@sifive.com>
8591 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8593 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8596 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8598 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8599 and CMP + SUB-immediate -> SUBS.
8601 2018-01-26 Martin Sebor <msebor@redhat.com>
8603 PR tree-optimization/83896
8604 * tree-ssa-strlen.c (get_string_len): Rename...
8605 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8606 Avoid assuming length is constant.
8607 (handle_char_store): Use HOST_WIDE_INT for string length.
8609 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8612 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8613 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8615 2018-01-26 Richard Biener <rguenther@suse.de>
8617 PR rtl-optimization/84003
8618 * dse.c (record_store): Only record redundant stores when
8619 the earlier store aliases at least all accesses the later one does.
8621 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8623 PR rtl-optimization/83985
8624 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8625 REG_CFA_RESTORE insns.
8626 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8627 REG_CFA_RESTORE insns here.
8630 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8631 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8633 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8635 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8636 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8637 (arc_init): Likewise.
8638 (arc_override_options): Likewise.
8639 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8641 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8643 * config/arc/arc.h (TARGET_DBNZ): Define.
8644 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8645 properly set the tune attribute.
8646 (dbnz): Use TARGET_DBNZ guard.
8647 * config/arc/arc.opt (mtune): Add core3 option.
8649 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8651 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8652 recognize new pic like addresses.
8653 (arc_delegitimize_address): Clean up.
8655 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8657 * config/arc/arc-arches.def: Option mrf16 valid for all
8659 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8660 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8661 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8662 * config/arc/arc-tables.opt: Regenerate.
8663 * config/arc/arc.c (arc_conditional_register_usage): Handle
8664 reduced register file case.
8665 (arc_file_start): Set must have build attributes.
8666 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8668 * config/arc/arc.opt (mrf16): Add new option.
8669 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8670 * config/arc/genmultilib.awk: Handle new mrf16 option.
8671 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8672 * config/arc/t-multilib: Regenerate.
8673 * doc/invoke.texi (ARC Options): Document mrf16 option.
8675 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8677 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8678 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8679 (arc_attribute_table): Add 'secure_call' attribute.
8680 (arc_print_operand): Print secure call operand.
8681 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8683 (arc_is_secure_call_p): New function. * config/arc/arc.md
8684 (call_i): Add support for sjli instruction.
8685 (call_value_i): Likewise.
8686 * config/arc/constraints.md (Csc): New constraint.
8688 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8689 John Eric Martin <John.Martin@emmicro-us.com>
8691 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8692 * config/arc/arc.c (_arc_jli_section): New struct.
8693 (arc_jli_section): New type.
8694 (rc_jli_sections): New static variable.
8695 (arc_handle_jli_attribute): New function.
8696 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8697 (arc_file_end): New function.
8698 (TARGET_ASM_FILE_END): Define.
8699 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8700 (arc_add_jli_section): New function.
8701 (jli_call_scan): Likewise.
8702 (arc_reorg): Call jli_call_scan.
8703 (arc_output_addsi): Remove 'S' from printing asm operand.
8704 (arc_is_jli_call_p): New function.
8705 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8707 (movhi_insn): Likewise.
8708 (movsi_insn): Likewise.
8709 (movsi_set_cc_insn): Likewise.
8710 (loadqi_update): Likewise.
8711 (load_zeroextendqisi_update): Likewise.
8712 (load_signextendqisi_update): Likewise.
8713 (loadhi_update): Likewise.
8714 (load_zeroextendhisi_update): Likewise.
8715 (load_signextendhisi_update): Likewise.
8716 (loadsi_update): Likewise.
8717 (loadsf_update): Likewise.
8718 (movsicc_insn): Likewise.
8719 (bset_insn): Likewise.
8720 (bxor_insn): Likewise.
8721 (bclr_insn): Likewise.
8722 (bmsk_insn): Likewise.
8723 (bicsi3_insn): Likewise.
8724 (cmpsi_cc_c_insn): Likewise.
8725 (movsi_ne): Likewise.
8726 (movsi_cond_exec): Likewise.
8727 (clrsbsi2): Likewise.
8738 (call_i): Remove 'S' asm letter, add jli instruction.
8739 (call_value_i): Likewise.
8740 * config/arc/arc.op (mjli-always): New option.
8741 * config/arc/constraints.md (Cji): New constraint.
8742 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8744 (subsf3_fpx): Likewise.
8745 (mulsf3_fpx): Likewise.
8746 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8748 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8750 * doc/invoke.texi (ARC): Document mjli-always option.
8752 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8754 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8755 avoid addition with 0 and use incw and decw where possible.
8757 2018-01-26 Richard Biener <rguenther@suse.de>
8759 PR tree-optimization/81082
8760 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8761 association if it requires casting to unsigned.
8762 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8763 from fold_plusminus_mult_expr to catch important cases late when
8764 range info is available.
8766 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8768 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8769 * configure.ac (hidden_linkonce): New test.
8770 * configure: Regenerate.
8771 * config.in: Regenerate.
8773 2018-01-26 Julia Koval <julia.koval@intel.com>
8775 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8776 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8777 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8778 _mm_mask_bitshuffle_epi64_mask): Fix type.
8779 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8780 USI_FTYPE_V4DI_V4DI_USI): Remove.
8781 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8782 __builtin_ia32_vpshufbitqmb256_mask,
8783 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8784 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8785 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8787 2018-01-26 Alan Modra <amodra@gmail.com>
8790 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8791 UNSPEC_VBPERMQ. Sort other unspecs.
8793 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8795 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8797 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8800 * predict.c (drop_profile): Do not push/pop cfun; update also
8802 (handle_missing_profiles): Fix logic looking for zero profiles.
8804 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8807 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8808 on functions with #pragma omp declare simd or functions with simd
8810 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8811 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8812 Remove trailing \n from warning_at calls.
8814 2018-01-25 Tom de Vries <tom@codesourcery.com>
8817 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8818 for neutered workers.
8820 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8823 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8824 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8826 2018-01-24 Jeff Law <law@redhat.com>
8829 * i386.c (get_probe_interval): Move to earlier point.
8830 (ix86_compute_frame_layout): If -fstack-clash-protection and
8831 the frame is larger than the probe interval, then use pushes
8832 to save registers rather than reg->mem moves.
8833 (ix86_expand_prologue): Remove conditional for int_registers_saved
8836 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8839 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8840 min/max for never referenced object.
8842 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8845 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8847 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8848 attributes from DECL_ATTRIBUTES (decl) without affecting
8849 DECL_ATTRIBUTES (current_function_decl).
8850 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8851 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8853 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8855 PR tree-optimization/83979
8856 * fold-const.c (fold_comparison): Use constant_boolean_node
8857 instead of boolean_{true,false}_node.
8859 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8861 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8864 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8866 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8867 Simplify the clause that sets the length attribute.
8868 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8869 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8870 clause that sets the length attribute.
8871 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8873 2018-01-24 Tom de Vries <tom@codesourcery.com>
8876 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8877 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8878 Add strict parameter.
8879 (prevent_branch_around_nothing): Insert dummy insn between branch to
8880 label and label with no ptx insn inbetween.
8881 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8883 2018-01-24 Tom de Vries <tom@codesourcery.com>
8886 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8887 for neutered threads in warp.
8888 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8890 2018-01-24 Richard Biener <rguenther@suse.de>
8892 PR tree-optimization/83176
8893 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8896 2018-01-24 Richard Biener <rguenther@suse.de>
8898 PR tree-optimization/82819
8899 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8900 code generating pluses that are no-ops in the target precision.
8902 2018-01-24 Richard Biener <rguenther@suse.de>
8905 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8907 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8909 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8910 to merge probabilities.
8911 * predict.c (probably_never_executed): Also mark as cold functions
8912 with global 0 profile and guessed local profile.
8913 * profile-count.c (profile_probability::combine_with_count): New
8915 * profile-count.h (profile_probability::operator*,
8916 profile_probability::operator*=, profile_probability::operator/,
8917 profile_probability::operator/=): Reduce precision to adjusted
8918 and set value to guessed on contradictory divisions.
8919 (profile_probability::combine_with_freq): Remove.
8920 (profile_probability::combine_wiht_count): Declare.
8921 (profile_count::force_nonzero):: Set to adjusted.
8922 (profile_count::probability_in):: Set quality to adjusted.
8923 * tree-ssa-tail-merge.c (replace_block_by): Use
8926 2018-01-23 Andrew Waterman <andrew@sifive.com>
8927 Jim Wilson <jimw@sifive.com>
8929 * config/riscv/riscv.c (riscv_stack_boundary): New.
8930 (riscv_option_override): Set riscv_stack_boundary. Handle
8931 riscv_preferred_stack_boundary_arg.
8932 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8933 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8934 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8935 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8936 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8937 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8939 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8942 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8943 of struct ix86_frame.
8944 (ix86_expand_epilogue): Likewise. Add a local variable for
8945 the reg_save_offset field in struct ix86_frame.
8947 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8949 PR tree-optimization/82604
8950 * tree-loop-distribution.c (enum partition_kind): New enum item
8951 PKIND_PARTIAL_MEMSET.
8952 (partition_builtin_p): Support above new enum item.
8953 (generate_code_for_partition): Ditto.
8954 (compute_access_range): Differentiate cases that equality can be
8955 proven at all loops, the innermost loops or no loops.
8956 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8957 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8958 (finalize_partitions, distribute_loop): Don't fuse partition of
8959 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8960 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8963 2018-01-23 Martin Liska <mliska@suse.cz>
8965 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8966 order to ignore the predictor.
8967 (PRED_POLYMORPHIC_CALL): Likewise.
8968 (PRED_RECURSIVE_CALL): Likewise.
8970 2018-01-23 Martin Liska <mliska@suse.cz>
8972 * tree-profile.c (tree_profiling): Print function header to
8973 aware reader which function we are working on.
8974 * value-prof.c (gimple_find_values_to_profile): Do not print
8975 not interesting value histograms.
8977 2018-01-23 Martin Liska <mliska@suse.cz>
8979 * profile-count.h (enum profile_quality): Add
8980 profile_uninitialized as the first value. Do not number values
8981 as they are zero based.
8982 (profile_count::verify): Update sanity check.
8983 (profile_probability::verify): Likewise.
8985 2018-01-23 Nathan Sidwell <nathan@acm.org>
8987 * doc/invoke.texi (ffor-scope): Deprecate.
8989 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8991 PR tree-optimization/83510
8992 * domwalk.c (set_all_edges_as_executable): New function.
8993 (dom_walker::dom_walker): Convert bool param
8994 "skip_unreachable_blocks" to enum reachability. Move setup of
8995 edge flags to set_all_edges_as_executable and only do it when
8996 reachability is REACHABLE_BLOCKS.
8997 * domwalk.h (enum dom_walker::reachability): New enum.
8998 (dom_walker::dom_walker): Convert bool param
8999 "skip_unreachable_blocks" to enum reachability.
9000 (set_all_edges_as_executable): New decl.
9001 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9002 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9004 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9005 but converting true to REACHABLE_BLOCKS.
9006 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9008 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9009 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9010 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9012 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9013 if check_all_array_refs will be called.
9015 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9017 * tree.c (selftest::test_location_wrappers): Add more test
9020 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9022 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9023 (selftest::test_bit_in_range): Likewise.
9025 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9028 * doc/sourcebuild.texi (vect_float): Say that the selector
9029 only describes the situation when -funsafe-math-optimizations is on.
9030 (vect_float_strict): Document.
9032 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9034 PR tree-optimization/83965
9035 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9036 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9037 instead of checking only for a reduction.
9038 (vect_recog_widen_sum_pattern): Likewise.
9040 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9042 * predict.c (probably_never_executed): Only use precise profile info.
9043 (compute_function_frequency): Skip after inlining hack since we now
9044 have quality checking.
9046 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9048 * profile-count.h (profile_probability::very_unlikely,
9049 profile_probability::unlikely, profile_probability::even): Set
9050 precision to guessed.
9052 2018-01-23 Richard Biener <rguenther@suse.de>
9054 PR tree-optimization/83963
9055 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9056 Properly terminate dominator walk when crossing the exit edge not
9057 when visiting its source block.
9059 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9062 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9063 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9065 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9067 PR tree-optimization/83957
9068 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9069 semicolon after for body surrounded by braces.
9071 PR tree-optimization/83081
9072 * profile-count.h (profile_probability::split): New method.
9073 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9074 Use profile_probability::split.
9075 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9076 when splitting a single conditional jump into 2.
9078 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9080 PR tree-optimization/69452
9081 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9084 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9086 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9087 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9088 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9090 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9092 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9093 * config/rl78/rl78.md (movdi): New define_expand.
9094 * config/rl78/rl78.c (rl78_split_movdi): New function.
9096 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9099 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9101 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9102 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9103 128-bit to produce an UNSPEC move to get the double word with the
9104 signbit and then a shift directly to do signbit.
9105 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9106 implementation with a new version that just does either a direct
9107 move or a regular move. Move memory interface to separate insns.
9108 Move insns so they are next to the expander.
9109 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9110 with signbit move. Split big and little endian case.
9111 (signbit<mode>2_dm_mem_le): Likewise.
9112 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9113 (signbit<mode>2_dm2): Likewise.
9115 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9117 * config/rl78/rl78.md (anddi3): New define_expand.
9119 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9121 * config/rl78/rl78.md (umindi3): New define_expand.
9123 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9125 * config/rl78/rl78.md (smindi3): New define_expand.
9127 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9129 * config/rl78/rl78.md (smaxdi3): New define_expand.
9131 2018-01-22 Carl Love <cel@us.ibm.com>
9133 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9134 LVX_V1TI): Add macro expansion.
9135 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9136 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9137 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9138 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9139 Change check to determine if the instruction is a byte reversing
9140 entry. Fix typo in comment.
9141 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9142 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9143 Add def_builtin calls for new builtins.
9144 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9145 Add define_insn expansion.
9147 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9149 * config/rl78/rl78.md (umaxdi3): New define_expand.
9151 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9153 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9154 for non-QImode registers.
9156 2018-01-22 Richard Biener <rguenther@suse.de>
9158 PR tree-optimization/83963
9159 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9160 including the loop exit block.
9161 (scop_detection::merge_sese): Likewise.
9162 (scop_detection::add_scop): Do it here instead.
9164 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9166 * doc/sourcebuild.texi (arm_softfloat): Document.
9168 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9171 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9172 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9173 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9175 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9176 David Edelsohn <dje.gcc@gmail.com>
9179 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9180 Change "crset eq" to "crset 2".
9181 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9182 (*call_indirect_aix<mode>_nospec): Likewise.
9183 (*call_value_indirect_aix<mode>_nospec): Likewise.
9184 (*call_indirect_elfv2<mode>_nospec): Likewise.
9185 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9186 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9187 change assembly output from . to $.
9188 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9189 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9190 (*tablejump<mode>_internal1_nospec): Likewise.
9192 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9195 * config/sh/sh_optimize_sett_clrt.cc:
9196 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9198 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9200 PR tree-optimization/83940
9201 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9202 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9203 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9204 use it to pass back the definition type.
9205 (vect_check_store_rhs): Likewise rhs_dt_out.
9206 (vect_build_gather_load_calls): Add a mask_dt argument and use
9207 it instead of a call to vect_is_simple_use.
9208 (vectorizable_store): Update calls to vect_check_load_store_mask
9209 and vect_check_store_rhs. Use the dt returned by the latter instead
9210 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9211 instead of calls to vect_is_simple_use. Pass the scalar rather
9212 than the vector operand to vect_is_simple_use when handling
9213 second and subsequent copies of an rhs value.
9214 (vectorizable_load): Update calls to vect_check_load_store_mask
9215 and vect_build_gather_load_calls. Use the cached mask_dt and
9216 gs_info.offset_dt instead of calls to vect_is_simple_use.
9218 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9221 * tree-emutls.c: Include gimplify.h.
9222 (lower_emutls_2): New function.
9223 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9224 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9225 it before further processing.
9228 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9229 UINTVAL (trueop1) instead of INTVAL (op1).
9231 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9235 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9236 INCOMING_FRAME_SP_OFFSET if not defined.
9237 (scan_trace): Add ENTRY argument. If true and
9238 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9239 emit a note to adjust the CFA offset.
9240 (create_cfi_notes): Adjust scan_trace callers.
9241 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9242 INCOMING_FRAME_SP_OFFSET in the CIE.
9243 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9244 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9246 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9247 * doc/tm.texi: Regenerated.
9249 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9251 PR rtl-optimization/83147
9252 * lra-constraints.c (remove_inheritance_pseudos): Use
9253 lra_substitute_pseudo_within_insn.
9255 2018-01-19 Tom de Vries <tom@codesourcery.com>
9256 Cesar Philippidis <cesar@codesourcery.com>
9259 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9261 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9264 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9265 spaces for function labels.
9267 2018-01-19 Martin Liska <mliska@suse.cz>
9269 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9270 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9271 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9272 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9273 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9274 (PRED_CONST_RETURN): Change from 69 to 65.
9275 (PRED_NULL_RETURN): Change from 91 to 71.
9276 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9277 (PRED_LOOP_GUARD): Change from 66 to 73.
9279 2018-01-19 Martin Liska <mliska@suse.cz>
9281 * predict.c (predict_insn_def): Add new assert.
9282 (struct branch_predictor): Change type to signed integer.
9283 (test_prediction_value_range): Amend test to cover
9285 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9286 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9287 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9288 (PRED_LOOP_IV_COMPARE): Likewise.
9289 * predict.h (PROB_UNINITIALIZED): Define new constant.
9291 2018-01-19 Martin Liska <mliska@suse.cz>
9293 * predict.c (dump_prediction): Add new format for
9294 analyze_brprob.py script which is enabled with -details
9296 * profile-count.h (precise_p): New function.
9298 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9300 PR tree-optimization/83922
9301 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9302 there are no statements that need masking.
9303 (vect_active_double_reduction_p): New function.
9304 (vect_analyze_loop_operations): Use it when handling phis that
9305 are not in the loop header.
9307 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9309 PR tree-optimization/83914
9310 * tree-vect-loop.c (vectorizable_induction): Don't convert
9311 init_expr or apply the peeling adjustment for inductions
9312 that are nested within the vectorized loop.
9314 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9316 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9319 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9323 * function.h (gimplify_parameters): Add gimple_seq * argument.
9324 * function.c: Include gimple.h and options.h.
9325 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9326 for the added local temporaries if needed.
9327 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9328 if there are any parameter cleanups, wrap whole body into a
9329 try/finally with the cleanups.
9331 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9334 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9335 Use GET_MODE_CLASS for scalar floating point.
9337 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9341 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9342 Fix call of call_cgraph_insertion_hooks.
9344 2018-01-18 Martin Sebor <msebor@redhat.com>
9346 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9348 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9351 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9354 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9357 * common.opt: (-ffile-prefix-map): New option.
9358 * opts.c (common_handle_option): Defer it.
9359 * opts-global.c (handle_common_deferred_options): Handle it.
9360 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9361 * file-prefix-map.h: New file.
9362 (remap_debug_filename, add_debug_prefix_map): ...here.
9363 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9364 * final.c (debug_prefix_map, add_debug_prefix_map
9365 remap_debug_filename): Move to...
9366 * file-prefix-map.c: New file.
9367 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9368 generalize, get rid of alloca(), use strrchr() instead of strchr().
9369 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9370 Implement in terms of add_prefix_map().
9371 (remap_macro_filename, remap_debug_filename): Implement in term of
9373 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9374 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9375 * dbxout.c: Include file-prefix-map.h.
9376 * varasm.c: Likewise.
9377 * vmsdbgout.c: Likewise.
9378 * xcoffout.c: Likewise.
9379 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9380 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9381 * doc/invoke.texi (-ffile-prefix-map): Document.
9382 (-fdebug-prefix-map): Update description.
9384 2018-01-18 Martin Liska <mliska@suse.cz>
9386 * config/i386/i386.c (indirect_thunk_name): Document that also
9388 (output_indirect_thunk): Document why both instructions
9389 (pause and lfence) are generated.
9391 2018-01-18 Richard Biener <rguenther@suse.de>
9393 PR tree-optimization/83887
9394 * graphite-scop-detection.c
9395 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9396 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9397 (scop_detection::merge_sese): Re-implement with a flood-fill
9398 algorithm that properly finds a SESE region if it exists.
9400 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9403 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9404 pointer_diff optimizations use view_convert instead of convert.
9406 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9408 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9409 Generate different code for -mno-speculate-indirect-jumps.
9410 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9411 (*call_indirect_aix<mode>): Disable for
9412 -mno-speculate-indirect-jumps.
9413 (*call_indirect_aix<mode>_nospec): New define_insn.
9414 (*call_value_indirect_aix<mode>): Disable for
9415 -mno-speculate-indirect-jumps.
9416 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9417 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9418 -mno-speculate-indirect-jumps.
9419 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9421 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9423 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9424 long double type, set the flags for noting the default long double
9425 type, even if we don't pass or return a long double type.
9427 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9430 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9433 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9435 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9436 support for merge[hl].
9437 (fold_mergehl_helper): New helper function.
9438 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9439 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9440 (altivec_vmrglw_direct): Add xxmrglw insn.
9442 2018-01-17 Andrew Waterman <andrew@sifive.com>
9444 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9445 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9447 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9450 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9451 call the lto_location_cache before reading the
9452 DECL_SOURCE_LOCATION of the types.
9454 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9455 Richard Sandiford <richard.sandiford@linaro.org>
9457 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9458 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9459 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9460 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9461 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9463 * config/aarch64/constraints.md (aarch64_movti_operand):
9465 * config/aarch64/predicates.md (Uti): Add new constraint.
9467 2018-01-17 Carl Love <cel@us.ibm.com>
9469 * config/rs6000/vsx.md (define_expand xl_len_r,
9470 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9471 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9473 (define_expand, define_insn): Move the shift left from the
9474 define_insn to the define_expand for lxvl and stxvl instructions.
9475 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9476 and XL_LEN_R definitions to PURE.
9478 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9480 * config/i386/i386.c (indirect_thunk_name): Declare regno
9481 as unsigned int. Compare regno with INVALID_REGNUM.
9482 (output_indirect_thunk): Ditto.
9483 (output_indirect_thunk_function): Ditto.
9484 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9485 in the call to output_indirect_thunk_function.
9487 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9490 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9491 rather than the size of inner_type to determine the stack slot size
9492 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9494 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9497 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9500 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9502 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9503 endian Linux systems to optionally enable multilibs for selecting
9504 the long double type if the user configured an explicit type.
9505 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9506 have no long double multilibs if not defined.
9507 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9508 warn if the user used -mabi={ieee,ibm}longdouble and we built
9509 multilibs for long double.
9510 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9511 appropriate multilib option.
9512 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9514 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9515 for building long double multilibs.
9516 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9518 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9520 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9523 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9525 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9528 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9531 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9534 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9536 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9537 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9539 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9541 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9542 different rtl trees depending on TARGET_64BIT.
9543 (rs6000_gen_lvx): Likewise.
9545 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9547 * config/visium/visium.md (nop): Tweak comment.
9548 (hazard_nop): Likewise.
9550 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9552 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9553 -mspeculate-indirect-jumps.
9554 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9555 for -mno-speculate-indirect-jumps.
9556 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9557 (*call_value_indirect_elfv2<mode>): Disable for
9558 -mno-speculate-indirect-jumps.
9559 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9560 (indirect_jump): Emit different RTL for
9561 -mno-speculate-indirect-jumps.
9562 (*indirect_jump<mode>): Disable for
9563 -mno-speculate-indirect-jumps.
9564 (*indirect_jump<mode>_nospec): New define_insn.
9565 (tablejump): Emit different RTL for
9566 -mno-speculate-indirect-jumps.
9567 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9568 (tablejumpsi_nospec): New define_expand.
9569 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9570 (tablejumpdi_nospec): New define_expand.
9571 (*tablejump<mode>_internal1): Disable for
9572 -mno-speculate-indirect-jumps.
9573 (*tablejump<mode>_internal1_nospec): New define_insn.
9574 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9577 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9579 * caller-save.c (insert_save): Drop unnecessary parameter. All
9582 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9583 Richard Biener <rguenth@suse.de>
9586 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9587 return early, inline manually is_gimple_sizepos. Make sure if we
9588 call gimplify_expr we don't end up with a gimple constant.
9589 * tree.c (variably_modified_type_p): Don't return true for
9590 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9591 * gimplify.h (is_gimple_sizepos): Remove.
9593 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9595 PR tree-optimization/83857
9596 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9597 vectorizable_live_operation for pure SLP statements.
9598 (vectorizable_live_operation): Handle PHIs.
9600 2018-01-16 Richard Biener <rguenther@suse.de>
9602 PR tree-optimization/83867
9603 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9604 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9606 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9609 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9610 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9611 If off is not INTEGER_CST, issue a may not be aligned warning
9612 rather than isn't aligned. Use isn%'t rather than isn't.
9613 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9615 <case MULT_EXPR>: Improve the case when bottom and one of the
9616 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9617 operand, in that case check if the other operand is multiple of
9618 bottom divided by the INTEGER_CST operand.
9620 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9623 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9624 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9625 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9626 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9627 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9628 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9629 (pa_function_arg_size): New function.
9631 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9633 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9634 in a separate statement.
9636 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9638 PR tree-optimization/83847
9639 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9640 group gathers and scatters.
9642 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9644 PR rtl-optimization/86620
9645 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9647 PR rtl-optimization/83213
9648 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9649 to last if both are JUMP_INSNs.
9651 PR tree-optimization/83843
9652 * gimple-ssa-store-merging.c
9653 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9654 store_immediate_info for bswap/nop orig_stores.
9656 2018-01-15 Andrew Waterman <andrew@sifive.com>
9658 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9660 <UDIV>: Increase cost if !TARGET_DIV.
9662 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9664 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9665 (define_attr "cr_logical_3op"): New.
9666 (cceq_ior_compare): Adjust.
9667 (cceq_ior_compare_complement): Adjust.
9668 (*cceq_rev_compare): Adjust.
9669 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9670 (is_cracked_insn): Adjust.
9671 (insn_must_be_first_in_group): Adjust.
9672 * config/rs6000/40x.md: Adjust.
9673 * config/rs6000/440.md: Adjust.
9674 * config/rs6000/476.md: Adjust.
9675 * config/rs6000/601.md: Adjust.
9676 * config/rs6000/603.md: Adjust.
9677 * config/rs6000/6xx.md: Adjust.
9678 * config/rs6000/7450.md: Adjust.
9679 * config/rs6000/7xx.md: Adjust.
9680 * config/rs6000/8540.md: Adjust.
9681 * config/rs6000/cell.md: Adjust.
9682 * config/rs6000/e300c2c3.md: Adjust.
9683 * config/rs6000/e500mc.md: Adjust.
9684 * config/rs6000/e500mc64.md: Adjust.
9685 * config/rs6000/e5500.md: Adjust.
9686 * config/rs6000/e6500.md: Adjust.
9687 * config/rs6000/mpc.md: Adjust.
9688 * config/rs6000/power4.md: Adjust.
9689 * config/rs6000/power5.md: Adjust.
9690 * config/rs6000/power6.md: Adjust.
9691 * config/rs6000/power7.md: Adjust.
9692 * config/rs6000/power8.md: Adjust.
9693 * config/rs6000/power9.md: Adjust.
9694 * config/rs6000/rs64.md: Adjust.
9695 * config/rs6000/titan.md: Adjust.
9697 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9699 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9700 ix86_indirect_branch_register logic.
9702 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9704 * config/i386/constraints.md (Bs): Update
9705 ix86_indirect_branch_register check. Don't check
9706 ix86_indirect_branch_register with GOT_memory_operand.
9708 * config/i386/predicates.md (GOT_memory_operand): Don't check
9709 ix86_indirect_branch_register here.
9710 (GOT32_symbol_operand): Likewise.
9712 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9714 * config/i386/predicates.md (constant_call_address_operand):
9715 Rewrite ix86_indirect_branch_register logic.
9716 (sibcall_insn_operand): Likewise.
9718 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9720 * config/i386/constraints.md (Bs): Replace
9721 ix86_indirect_branch_thunk_register with
9722 ix86_indirect_branch_register.
9724 * config/i386/i386.md (indirect_jump): Likewise.
9725 (tablejump): Likewise.
9726 (*sibcall_memory): Likewise.
9727 (*sibcall_value_memory): Likewise.
9728 Peepholes of indirect call and jump via memory: Likewise.
9729 * config/i386/i386.opt: Likewise.
9730 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9731 (GOT_memory_operand): Likewise.
9732 (call_insn_operand): Likewise.
9733 (sibcall_insn_operand): Likewise.
9734 (GOT32_symbol_operand): Likewise.
9736 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9739 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9740 type rather than type addr's type points to.
9741 (expand_omp_atomic_mutex): Likewise.
9742 (expand_omp_atomic): Likewise.
9744 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9747 * config/i386/i386.c (output_indirect_thunk_function): Use
9748 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9749 for __x86_return_thunk.
9751 2018-01-15 Richard Biener <rguenther@suse.de>
9754 * expmed.c (extract_bit_field_1): Fix typo.
9756 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9759 * config/arm/iterators.md (VF): New mode iterator.
9760 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9761 Remove integer-related logic from pattern.
9762 (neon_vabd<mode>_3): Likewise.
9764 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9767 * common.opt (fstrict-overflow): No longer an alias.
9768 (fwrapv-pointer): New option.
9769 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9770 also for pointer types based on flag_wrapv_pointer.
9771 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9772 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9773 opts->x_flag_wrapv got set.
9774 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9775 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9776 POINTER_TYPE_OVERFLOW_UNDEFINED.
9777 * match.pd: Likewise in address comparison pattern.
9778 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9780 2018-01-15 Richard Biener <rguenther@suse.de>
9783 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9784 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9785 Reset type names to their identifier if their TYPE_DECL doesn't
9786 have linkage (and thus is used for ODR and devirt).
9787 (save_debug_info_for_decl): Remove.
9788 (save_debug_info_for_type): Likewise.
9789 (add_tree_to_fld_list): Adjust.
9790 * tree-pretty-print.c (dump_generic_node): Make dumping of
9791 type names more robust.
9793 2018-01-15 Richard Biener <rguenther@suse.de>
9795 * BASE-VER: Bump to 8.0.1.
9797 2018-01-14 Martin Sebor <msebor@redhat.com>
9800 * builtins.c (check_access): Avoid warning when the no-warning bit
9803 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9805 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9806 * ira-color (allocno_hard_regs_compare): Likewise.
9808 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9811 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9812 Use .pushsection/.popsection.
9814 2018-01-14 Martin Sebor <msebor@redhat.com>
9817 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9819 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9821 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9822 entry from extra_headers.
9823 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9824 extra_headers, make the list bitwise identical to the i?86-*-* one.
9826 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9828 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9829 -mcmodel=large with -mindirect-branch=thunk,
9830 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9831 -mfunction-return=thunk-extern.
9832 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9833 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9834 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9836 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9838 * config/i386/i386.c (print_reg): Print the name of the full
9839 integer register without '%'.
9840 (ix86_print_operand): Handle 'V'.
9841 * doc/extend.texi: Document 'V' modifier.
9843 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9845 * config/i386/constraints.md (Bs): Disallow memory operand for
9846 -mindirect-branch-register.
9848 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9849 (GOT_memory_operand): Likewise.
9850 (call_insn_operand): Likewise.
9851 (sibcall_insn_operand): Likewise.
9852 (GOT32_symbol_operand): Likewise.
9853 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9854 for -mindirect-branch-register.
9855 (tablejump): Likewise.
9856 (*sibcall_memory): Likewise.
9857 (*sibcall_value_memory): Likewise.
9858 Disallow peepholes of indirect call and jump via memory for
9859 -mindirect-branch-register.
9860 (*call_pop): Replace m with Bw.
9861 (*call_value_pop): Likewise.
9862 (*sibcall_pop_memory): Replace m with Bs.
9863 * config/i386/i386.opt (mindirect-branch-register): New option.
9864 * doc/invoke.texi: Document -mindirect-branch-register option.
9866 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9868 * config/i386/i386-protos.h (ix86_output_function_return): New.
9869 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9870 set function_return_type.
9871 (indirect_thunk_name): Add ret_p to indicate thunk for function
9873 (output_indirect_thunk_function): Pass false to
9874 indirect_thunk_name.
9875 (ix86_output_indirect_branch_via_reg): Likewise.
9876 (ix86_output_indirect_branch_via_push): Likewise.
9877 (output_indirect_thunk_function): Create alias for function
9878 return thunk if regno < 0.
9879 (ix86_output_function_return): New function.
9880 (ix86_handle_fndecl_attribute): Handle function_return.
9881 (ix86_attribute_table): Add function_return.
9882 * config/i386/i386.h (machine_function): Add
9883 function_return_type.
9884 * config/i386/i386.md (simple_return_internal): Use
9885 ix86_output_function_return.
9886 (simple_return_internal_long): Likewise.
9887 * config/i386/i386.opt (mfunction-return=): New option.
9888 (indirect_branch): Mention -mfunction-return=.
9889 * doc/extend.texi: Document function_return function attribute.
9890 * doc/invoke.texi: Document -mfunction-return= option.
9892 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9894 * config/i386/i386-opts.h (indirect_branch): New.
9895 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9896 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9897 with local indirect jump when converting indirect call and jump.
9898 (ix86_set_indirect_branch_type): New.
9899 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9900 (indirectlabelno): New.
9901 (indirect_thunk_needed): Likewise.
9902 (indirect_thunk_bnd_needed): Likewise.
9903 (indirect_thunks_used): Likewise.
9904 (indirect_thunks_bnd_used): Likewise.
9905 (INDIRECT_LABEL): Likewise.
9906 (indirect_thunk_name): Likewise.
9907 (output_indirect_thunk): Likewise.
9908 (output_indirect_thunk_function): Likewise.
9909 (ix86_output_indirect_branch_via_reg): Likewise.
9910 (ix86_output_indirect_branch_via_push): Likewise.
9911 (ix86_output_indirect_branch): Likewise.
9912 (ix86_output_indirect_jmp): Likewise.
9913 (ix86_code_end): Call output_indirect_thunk_function if needed.
9914 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9916 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9917 (ix86_attribute_table): Add indirect_branch.
9918 * config/i386/i386.h (machine_function): Add indirect_branch_type
9919 and has_local_indirect_jump.
9920 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9922 (tablejump): Likewise.
9923 (*indirect_jump): Use ix86_output_indirect_jmp.
9924 (*tablejump_1): Likewise.
9925 (simple_return_indirect_internal): Likewise.
9926 * config/i386/i386.opt (mindirect-branch=): New option.
9927 (indirect_branch): New.
9930 (thunk-inline): Likewise.
9931 (thunk-extern): Likewise.
9932 * doc/extend.texi: Document indirect_branch function attribute.
9933 * doc/invoke.texi: Document -mindirect-branch= option.
9935 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9938 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9940 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9942 * ipa-inline.c (want_inline_small_function_p): Return false if
9943 inlining has already failed with CIF_FINAL_ERROR.
9944 (update_caller_keys): Call want_inline_small_function_p before
9946 (update_callee_keys): Likewise.
9948 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9950 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9952 (rs6000_quadword_masked_address_p): Likewise.
9953 (quad_aligned_load_p): Likewise.
9954 (quad_aligned_store_p): Likewise.
9955 (const_load_sequence_p): Add comment to describe the outer-most loop.
9956 (mimic_memory_attributes_and_flags): New function.
9957 (rs6000_gen_stvx): Likewise.
9958 (replace_swapped_aligned_store): Likewise.
9959 (rs6000_gen_lvx): Likewise.
9960 (replace_swapped_aligned_load): Likewise.
9961 (replace_swapped_load_constant): Capitalize argument name in
9962 comment describing this function.
9963 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9964 and stores that access quad-word aligned addresses and replace
9965 with stvx or lvx instructions when appropriate.
9966 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9967 New function prototype.
9968 (rs6000_quadword_masked_address_p): Likewise.
9969 (rs6000_gen_lvx): Likewise.
9970 (rs6000_gen_stvx): Likewise.
9971 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9972 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9973 when memory address is aligned.
9974 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9975 this split to select lvx instruction when memory address is aligned.
9976 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9977 instruction when memory address is aligned.
9978 (*vsx_le_perm_load_v16qi): Likewise.
9979 (four unnamed splitters): Modify to select the stvx instruction
9980 when memory is aligned.
9982 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9984 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9985 which appears in the queue multiple times.
9987 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9988 Alan Hayward <alan.hayward@arm.com>
9989 David Sherwood <david.sherwood@arm.com>
9991 * tree-vectorizer.h (vec_lower_bound): New structure.
9992 (_loop_vec_info): Add check_nonzero and lower_bounds.
9993 (LOOP_VINFO_CHECK_NONZERO): New macro.
9994 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9995 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9996 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9997 fields. Make seg_len the distance travelled, not including the
9999 (dr_direction_indicator): Declare.
10000 (dr_zero_step_indicator): Likewise.
10001 (dr_known_forward_stride_p): Likewise.
10002 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10004 (runtime_alias_check_p): Allow runtime alias checks with
10006 (operator ==): Compare access_size and align.
10007 (prune_runtime_alias_test_list): Rework for new distinction between
10008 the access_size and seg_len.
10009 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10011 (get_segment_min_max): New function.
10012 (create_intersect_range_checks): Use it.
10013 (dr_step_indicator): New function.
10014 (dr_direction_indicator): Likewise.
10015 (dr_zero_step_indicator): Likewise.
10016 (dr_known_forward_stride_p): Likewise.
10017 * tree-loop-distribution.c (data_ref_segment_size): Return
10018 DR_STEP * (niters - 1).
10019 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10021 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10022 (vect_preserves_scalar_order_p): New function, split out from...
10023 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10024 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10025 (vect_vfa_access_size): New function.
10026 (vect_vfa_align): Likewise.
10027 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10028 (dump_lower_bound): New function.
10029 (vect_check_lower_bound): Likewise.
10030 (vect_small_gap_p): Likewise.
10031 (vectorizable_with_step_bound_p): Likewise.
10032 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10033 depencies if the vectorization factor is 1. Convert the checks
10034 for nonzero steps into checks on the bounds of DR_STEP. Try using
10035 a bunds check for variable steps if the minimum required step is
10036 relatively small. Update calls to the dr_with_seg_len
10037 constructor and to vect_compile_time_alias.
10038 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10040 (vect_loop_versioning): Call it.
10041 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10043 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10046 Alan Hayward <alan.hayward@arm.com>
10047 David Sherwood <david.sherwood@arm.com>
10049 * doc/sourcebuild.texi (vect_scatter_store): Document.
10050 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10052 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10054 * genopinit.c (main): Add supports_vec_scatter_store and
10055 supports_vec_scatter_store_cached to target_optabs.
10056 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10057 IFN_MASK_SCATTER_STORE.
10058 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10060 * internal-fn.h (internal_store_fn_p): Declare.
10061 (internal_fn_stored_value_index): Likewise.
10062 * internal-fn.c (scatter_store_direct): New macro.
10063 (expand_scatter_store_optab_fn): New function.
10064 (direct_scatter_store_optab_supported_p): New macro.
10065 (internal_store_fn_p): New function.
10066 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10067 IFN_MASK_SCATTER_STORE.
10068 (internal_fn_mask_index): Likewise.
10069 (internal_fn_stored_value_index): New function.
10070 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10071 for scatter stores.
10072 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10073 * optabs-query.c (supports_vec_scatter_store_p): New function.
10074 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10075 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10076 true for scatter stores.
10077 (vect_gather_scatter_fn_p): Handle scatter stores too.
10078 (vect_check_gather_scatter): Consider using scatter stores if
10079 supports_vec_scatter_store_p.
10080 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10081 scatter stores too.
10082 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10083 internal_fn_stored_value_index.
10084 (check_load_store_masking): Handle scatter stores too.
10085 (vect_get_store_rhs): Make public.
10086 (vectorizable_call): Use internal_store_fn_p.
10087 (vectorizable_store): Handle scatter store internal functions.
10088 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10089 when deciding whether the end of the group has been reached.
10090 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10091 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10092 (mask_scatter_store<mode>): New insns.
10094 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10095 Alan Hayward <alan.hayward@arm.com>
10096 David Sherwood <david.sherwood@arm.com>
10098 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10099 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10100 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10102 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10103 Use vect_truncate_gather_scatter_offset if we can't treat the
10104 operation as a normal gather load or scatter store.
10105 (get_group_load_store_type): Take the gather_scatter_info
10106 as argument. Try using a gather load or scatter store for
10107 single-element groups.
10108 (get_load_store_type): Update calls to get_group_load_store_type
10109 and vect_use_strided_gather_scatters_p.
10111 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10112 Alan Hayward <alan.hayward@arm.com>
10113 David Sherwood <david.sherwood@arm.com>
10115 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10116 optional tree argument.
10117 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10119 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10120 but continue to use the current value as a fallback.
10121 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10122 to compare the updates.
10123 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10124 (get_load_store_type): Use it when handling a strided access.
10125 (vect_get_strided_load_store_ops): New function.
10126 (vect_get_data_ptr_increment): Likewise.
10127 (vectorizable_load): Handle strided gather loads. Always pass
10128 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10130 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10131 Alan Hayward <alan.hayward@arm.com>
10132 David Sherwood <david.sherwood@arm.com>
10134 * doc/md.texi (gather_load@var{m}): Document.
10135 (mask_gather_load@var{m}): Likewise.
10136 * genopinit.c (main): Add supports_vec_gather_load and
10137 supports_vec_gather_load_cached to target_optabs.
10138 * optabs-tree.c (init_tree_optimization_optabs): Use
10139 ggc_cleared_alloc to allocate target_optabs.
10140 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10141 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10143 * internal-fn.h (internal_load_fn_p): Declare.
10144 (internal_gather_scatter_fn_p): Likewise.
10145 (internal_fn_mask_index): Likewise.
10146 (internal_gather_scatter_fn_supported_p): Likewise.
10147 * internal-fn.c (gather_load_direct): New macro.
10148 (expand_gather_load_optab_fn): New function.
10149 (direct_gather_load_optab_supported_p): New macro.
10150 (direct_internal_fn_optab): New function.
10151 (internal_load_fn_p): Likewise.
10152 (internal_gather_scatter_fn_p): Likewise.
10153 (internal_fn_mask_index): Likewise.
10154 (internal_gather_scatter_fn_supported_p): Likewise.
10155 * optabs-query.c (supports_at_least_one_mode_p): New function.
10156 (supports_vec_gather_load_p): Likewise.
10157 * optabs-query.h (supports_vec_gather_load_p): Declare.
10158 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10159 and memory_type field.
10160 (NUM_PATTERNS): Bump to 15.
10161 * tree-vect-data-refs.c: Include internal-fn.h.
10162 (vect_gather_scatter_fn_p): New function.
10163 (vect_describe_gather_scatter_call): Likewise.
10164 (vect_check_gather_scatter): Try using internal functions for
10165 gather loads. Recognize existing calls to a gather load function.
10166 (vect_analyze_data_refs): Consider using gather loads if
10167 supports_vec_gather_load_p.
10168 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10169 (vect_get_gather_scatter_offset_type): Likewise.
10170 (vect_convert_mask_for_vectype): Likewise.
10171 (vect_add_conversion_to_patterm): Likewise.
10172 (vect_try_gather_scatter_pattern): Likewise.
10173 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10174 (vect_vect_recog_func_ptrs): Add it.
10175 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10176 internal_fn_mask_index and internal_gather_scatter_fn_p.
10177 (check_load_store_masking): Take the gather_scatter_info as an
10178 argument and handle gather loads.
10179 (vect_get_gather_scatter_ops): New function.
10180 (vectorizable_call): Check internal_load_fn_p.
10181 (vectorizable_load): Likewise. Handle gather load internal
10183 (vectorizable_store): Update call to check_load_store_masking.
10184 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10185 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10186 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10187 (aarch64_gather_scale_operand_d): New predicates.
10188 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10189 (mask_gather_load<mode>): New insns.
10191 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10192 Alan Hayward <alan.hayward@arm.com>
10193 David Sherwood <david.sherwood@arm.com>
10195 * optabs.def (fold_left_plus_optab): New optab.
10196 * doc/md.texi (fold_left_plus_@var{m}): Document.
10197 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10198 * internal-fn.c (fold_left_direct): Define.
10199 (expand_fold_left_optab_fn): Likewise.
10200 (direct_fold_left_optab_supported_p): Likewise.
10201 * fold-const-call.c (fold_const_fold_left): New function.
10202 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10203 * tree-parloops.c (valid_reduction_p): New function.
10204 (gather_scalar_reductions): Use it.
10205 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10206 (vect_finish_replace_stmt): Declare.
10207 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10208 (needs_fold_left_reduction_p): New function, split out from...
10209 (vect_is_simple_reduction): ...here. Accept reductions that
10210 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10211 (vect_force_simple_reduction): Also store the reduction type in
10212 the assignment's STMT_VINFO_REDUC_TYPE.
10213 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10214 (merge_with_identity): New function.
10215 (vect_expand_fold_left): Likewise.
10216 (vectorize_fold_left_reduction): Likewise.
10217 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10218 scalar phi in place for it. Check for target support and reject
10219 cases that would reassociate the operation. Defer the transform
10220 phase to vectorize_fold_left_reduction.
10221 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10222 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10223 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10225 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10227 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10228 call to ifc_temp_var.
10230 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10231 Alan Hayward <alan.hayward@arm.com>
10232 David Sherwood <david.sherwood@arm.com>
10234 * target.def (legitimize_address_displacement): Take the original
10235 offset as a poly_int.
10236 * targhooks.h (default_legitimize_address_displacement): Update
10238 * targhooks.c (default_legitimize_address_displacement): Likewise.
10239 * doc/tm.texi: Regenerate.
10240 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10241 as an argument, moving assert of ad->disp == ad->disp_term to...
10242 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10243 Try calling targetm.legitimize_address_displacement before expanding
10244 the address rather than afterwards, and adjust for the new interface.
10245 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10246 Match the new hook interface. Handle SVE addresses.
10247 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10248 new hook interface.
10250 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10252 * Makefile.in (OBJS): Add early-remat.o.
10253 * target.def (select_early_remat_modes): New hook.
10254 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10255 * doc/tm.texi: Regenerate.
10256 * targhooks.h (default_select_early_remat_modes): Declare.
10257 * targhooks.c (default_select_early_remat_modes): New function.
10258 * timevar.def (TV_EARLY_REMAT): New timevar.
10259 * passes.def (pass_early_remat): New pass.
10260 * tree-pass.h (make_pass_early_remat): Declare.
10261 * early-remat.c: New file.
10262 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10264 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10266 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10267 Alan Hayward <alan.hayward@arm.com>
10268 David Sherwood <david.sherwood@arm.com>
10270 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10271 vfm1 with a bound_epilog parameter.
10272 (vect_do_peeling): Update calls accordingly, and move the prologue
10273 call earlier in the function. Treat the base bound_epilog as 0 for
10274 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10275 this base when peeling for gaps.
10276 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10277 with fully-masked loops.
10278 (vect_estimate_min_profitable_iters): Handle the single peeled
10279 iteration in that case.
10281 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10282 Alan Hayward <alan.hayward@arm.com>
10283 David Sherwood <david.sherwood@arm.com>
10285 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10286 single-element interleaving even if the size is not a power of 2.
10287 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10288 accesses for single-element interleaving if the group size is
10291 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10292 Alan Hayward <alan.hayward@arm.com>
10293 David Sherwood <david.sherwood@arm.com>
10295 * doc/md.texi (fold_extract_last_@var{m}): Document.
10296 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10297 * optabs.def (fold_extract_last_optab): New optab.
10298 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10299 * internal-fn.c (fold_extract_direct): New macro.
10300 (expand_fold_extract_optab_fn): Likewise.
10301 (direct_fold_extract_optab_supported_p): Likewise.
10302 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10303 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10304 EXTRACT_LAST_REDUCTION.
10305 (get_initial_def_for_reduction): Do not create an initial vector
10306 for EXTRACT_LAST_REDUCTION reductions.
10307 (vectorizable_reduction): Leave the scalar phi in place for
10308 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10309 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10310 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10311 transform phase to vectorizable_condition.
10312 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10314 (vect_finish_stmt_generation): ...here.
10315 (vect_finish_replace_stmt): New function.
10316 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10317 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10319 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10321 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10322 Alan Hayward <alan.hayward@arm.com>
10323 David Sherwood <david.sherwood@arm.com>
10325 * doc/md.texi (extract_last_@var{m}): Document.
10326 * optabs.def (extract_last_optab): New optab.
10327 * internal-fn.def (EXTRACT_LAST): New internal function.
10328 * internal-fn.c (cond_unary_direct): New macro.
10329 (expand_cond_unary_optab_fn): Likewise.
10330 (direct_cond_unary_optab_supported_p): Likewise.
10331 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10332 loops using EXTRACT_LAST.
10333 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10334 (extract_last_<mode>): ...this optab.
10335 (vec_extract<mode><Vel>): Update accordingly.
10337 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10338 Alan Hayward <alan.hayward@arm.com>
10339 David Sherwood <david.sherwood@arm.com>
10341 * target.def (empty_mask_is_expensive): New hook.
10342 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10343 * doc/tm.texi: Regenerate.
10344 * targhooks.h (default_empty_mask_is_expensive): Declare.
10345 * targhooks.c (default_empty_mask_is_expensive): New function.
10346 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10347 if the target says that empty masks are expensive.
10348 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10350 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10352 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10353 Alan Hayward <alan.hayward@arm.com>
10354 David Sherwood <david.sherwood@arm.com>
10356 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10357 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10358 (vect_use_loop_mask_for_alignment_p): New function.
10359 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10360 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10361 niters_skip argument. Make sure that the first niters_skip elements
10362 of the first iteration are inactive.
10363 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10364 Update call to vect_set_loop_masks_directly.
10365 (get_misalign_in_elems): New function, split out from...
10366 (vect_gen_prolog_loop_niters): ...here.
10367 (vect_update_init_of_dr): Take a code argument that specifies whether
10368 the adjustment should be added or subtracted.
10369 (vect_update_init_of_drs): Likewise.
10370 (vect_prepare_for_masked_peels): New function.
10371 (vect_do_peeling): Skip prologue peeling if we're using a mask
10372 instead. Update call to vect_update_inits_of_drs.
10373 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10375 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10376 alignment. Do not include the number of peeled iterations in
10377 the minimum threshold in that case.
10378 (vectorizable_induction): Adjust the start value down by
10379 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10380 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10381 Take the number of skipped iterations into account when calculating
10383 * tree-vect-stmts.c (vect_gen_while_not): New function.
10385 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10386 Alan Hayward <alan.hayward@arm.com>
10387 David Sherwood <david.sherwood@arm.com>
10389 * doc/sourcebuild.texi (vect_fully_masked): Document.
10390 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10391 default value to 0.
10392 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10394 (vect_analyze_loop_2): ...here. Don't check the vectorization
10395 factor against the number of loop iterations if the loop is
10398 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10399 Alan Hayward <alan.hayward@arm.com>
10400 David Sherwood <david.sherwood@arm.com>
10402 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10403 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10404 (dump_groups): Update accordingly.
10405 (iv_use::mem_type): New member variable.
10406 (address_p): New function.
10407 (record_use): Add a mem_type argument and initialize the new
10409 (record_group_use): Add a mem_type argument. Use address_p.
10410 Remove obsolete null checks of base_object. Update call to record_use.
10411 (find_interesting_uses_op): Update call to record_group_use.
10412 (find_interesting_uses_cond): Likewise.
10413 (find_interesting_uses_address): Likewise.
10414 (get_mem_type_for_internal_fn): New function.
10415 (find_address_like_use): Likewise.
10416 (find_interesting_uses_stmt): Try find_address_like_use before
10417 calling find_interesting_uses_op.
10418 (addr_offset_valid_p): Use the iv mem_type field as the type
10419 of the addressed memory.
10420 (add_autoinc_candidates): Likewise.
10421 (get_address_cost): Likewise.
10422 (split_small_address_groups_p): Use address_p.
10423 (split_address_groups): Likewise.
10424 (add_iv_candidate_for_use): Likewise.
10425 (autoinc_possible_for_pair): Likewise.
10426 (rewrite_groups): Likewise.
10427 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10428 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10429 (get_alias_ptr_type_for_ptr_address): New function.
10430 (rewrite_use_address): Rewrite address uses in calls that were
10431 identified by find_address_like_use.
10433 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10434 Alan Hayward <alan.hayward@arm.com>
10435 David Sherwood <david.sherwood@arm.com>
10437 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10439 * gimple-expr.h (is_gimple_addressable: Likewise.
10440 * gimple-expr.c (is_gimple_address): Likewise.
10441 * internal-fn.c (expand_call_mem_ref): New function.
10442 (expand_mask_load_optab_fn): Use it.
10443 (expand_mask_store_optab_fn): Likewise.
10445 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10446 Alan Hayward <alan.hayward@arm.com>
10447 David Sherwood <david.sherwood@arm.com>
10449 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10450 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10451 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10452 (cond_umax@var{mode}): Document.
10453 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10454 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10455 (cond_umin_optab, cond_umax_optab): New optabs.
10456 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10457 (COND_IOR, COND_XOR): New internal functions.
10458 * internal-fn.h (get_conditional_internal_fn): Declare.
10459 * internal-fn.c (cond_binary_direct): New macro.
10460 (expand_cond_binary_optab_fn): Likewise.
10461 (direct_cond_binary_optab_supported_p): Likewise.
10462 (get_conditional_internal_fn): New function.
10463 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10464 Cope with reduction statements that are vectorized as calls rather
10466 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10467 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10468 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10469 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10470 (UNSPEC_COND_EOR): New unspecs.
10471 (optab): Add mappings for them.
10472 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10473 (sve_int_op, sve_fp_op): New int attributes.
10475 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10476 Alan Hayward <alan.hayward@arm.com>
10477 David Sherwood <david.sherwood@arm.com>
10479 * optabs.def (while_ult_optab): New optab.
10480 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10481 * internal-fn.def (WHILE_ULT): New internal function.
10482 * internal-fn.h (direct_internal_fn_supported_p): New override
10483 that takes two types as argument.
10484 * internal-fn.c (while_direct): New macro.
10485 (expand_while_optab_fn): New function.
10486 (convert_optab_supported_p): Likewise.
10487 (direct_while_optab_supported_p): New macro.
10488 * wide-int.h (wi::udiv_ceil): New function.
10489 * tree-vectorizer.h (rgroup_masks): New structure.
10490 (vec_loop_masks): New typedef.
10491 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10492 and fully_masked_p.
10493 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10494 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10495 (vect_max_vf): New function.
10496 (slpeel_make_loop_iterate_ntimes): Delete.
10497 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10498 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10499 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10500 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10501 internal-fn.h, stor-layout.h and optabs-query.h.
10502 (vect_set_loop_mask): New function.
10503 (add_preheader_seq): Likewise.
10504 (add_header_seq): Likewise.
10505 (interleave_supported_p): Likewise.
10506 (vect_maybe_permute_loop_masks): Likewise.
10507 (vect_set_loop_masks_directly): Likewise.
10508 (vect_set_loop_condition_masked): Likewise.
10509 (vect_set_loop_condition_unmasked): New function, split out from
10510 slpeel_make_loop_iterate_ntimes.
10511 (slpeel_make_loop_iterate_ntimes): Rename to..
10512 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10513 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10514 (vect_do_peeling): Update call accordingly.
10515 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10517 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10518 mask_compare_type, can_fully_mask_p and fully_masked_p.
10519 (release_vec_loop_masks): New function.
10520 (_loop_vec_info): Use it to free the loop masks.
10521 (can_produce_all_loop_masks_p): New function.
10522 (vect_get_max_nscalars_per_iter): Likewise.
10523 (vect_verify_full_masking): Likewise.
10524 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10525 retries, and free the mask rgroups before retrying. Check loop-wide
10526 reasons for disallowing fully-masked loops. Make the final decision
10527 about whether use a fully-masked loop or not.
10528 (vect_estimate_min_profitable_iters): Do not assume that peeling
10529 for the number of iterations will be needed for fully-masked loops.
10530 (vectorizable_reduction): Disable fully-masked loops.
10531 (vectorizable_live_operation): Likewise.
10532 (vect_halve_mask_nunits): New function.
10533 (vect_double_mask_nunits): Likewise.
10534 (vect_record_loop_mask): Likewise.
10535 (vect_get_loop_mask): Likewise.
10536 (vect_transform_loop): Handle the case in which the final loop
10537 iteration might handle a partial vector. Call vect_set_loop_condition
10538 instead of slpeel_make_loop_iterate_ntimes.
10539 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10540 (check_load_store_masking): New function.
10541 (prepare_load_store_mask): Likewise.
10542 (vectorizable_store): Handle fully-masked loops.
10543 (vectorizable_load): Likewise.
10544 (supportable_widening_operation): Use vect_halve_mask_nunits for
10546 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10547 (vect_gen_while): New function.
10548 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10549 (aarch64_uqdec<mode>): New insn.
10551 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10552 Alan Hayward <alan.hayward@arm.com>
10553 David Sherwood <david.sherwood@arm.com>
10555 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10556 (reduc_xor_scal_optab): New optabs.
10557 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10558 (reduc_xor_scal_@var{m}): Document.
10559 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10560 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10561 internal functions.
10562 * fold-const-call.c (fold_const_call): Handle them.
10563 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10564 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10565 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10566 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10567 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10568 (UNSPEC_XORV): New unspecs.
10569 (optab): Add entries for them.
10570 (BITWISEV): New int iterator.
10571 (bit_reduc_op): New int attributes.
10573 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10574 Alan Hayward <alan.hayward@arm.com>
10575 David Sherwood <david.sherwood@arm.com>
10577 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10578 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10579 * optabs.def (vec_shl_insert_optab): New optab.
10580 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10581 (duplicate_and_interleave): Likewise.
10582 * tree-vect-loop.c: Include internal-fn.h.
10583 (neutral_op_for_slp_reduction): New function, split out from
10584 get_initial_defs_for_reduction.
10585 (get_initial_def_for_reduction): Handle option 2 for variable-length
10586 vectors by loading the neutral value into a vector and then shifting
10587 the initial value into element 0.
10588 (get_initial_defs_for_reduction): Replace the code argument with
10589 the neutral value calculated by neutral_op_for_slp_reduction.
10590 Use gimple_build_vector for constant-length vectors.
10591 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10592 but the first group_size elements have a neutral value.
10593 Use duplicate_and_interleave otherwise.
10594 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10595 Update call to get_initial_defs_for_reduction. Handle SLP
10596 reductions for variable-length vectors by creating one vector
10597 result for each scalar result, with the elements associated
10598 with other scalar results stubbed out with the neutral value.
10599 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10600 Require IFN_VEC_SHL_INSERT for double reductions on
10601 variable-length vectors, or SLP reductions that have
10602 a neutral value. Require can_duplicate_and_interleave_p
10603 support for variable-length unchained SLP reductions if there
10604 is no neutral value, such as for MIN/MAX reductions. Also require
10605 the number of vector elements to be a multiple of the number of
10606 SLP statements when doing variable-length unchained SLP reductions.
10607 Update call to vect_create_epilog_for_reduction.
10608 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10609 and remove initial values.
10610 (duplicate_and_interleave): Make public.
10611 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10612 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10615 Alan Hayward <alan.hayward@arm.com>
10616 David Sherwood <david.sherwood@arm.com>
10618 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10619 (can_duplicate_and_interleave_p): New function.
10620 (vect_get_and_check_slp_defs): Take the vector of statements
10621 rather than just the current one. Remove excess parentheses.
10622 Restriction rejectinon of vect_constant_def and vect_external_def
10623 for variable-length vectors to boolean types, or types for which
10624 can_duplicate_and_interleave_p is false.
10625 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10626 (duplicate_and_interleave): New function.
10627 (vect_get_constant_vectors): Use gimple_build_vector for
10628 constant-length vectors and suitable variable-length constant
10629 vectors. Use duplicate_and_interleave for other variable-length
10630 vectors. Don't defer the update when inserting new statements.
10632 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10633 Alan Hayward <alan.hayward@arm.com>
10634 David Sherwood <david.sherwood@arm.com>
10636 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10637 min_profitable_iters doesn't go negative.
10639 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10640 Alan Hayward <alan.hayward@arm.com>
10641 David Sherwood <david.sherwood@arm.com>
10643 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10644 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10645 * optabs.def (vec_mask_load_lanes_optab): New optab.
10646 (vec_mask_store_lanes_optab): Likewise.
10647 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10648 (MASK_STORE_LANES): Likewise.
10649 * internal-fn.c (mask_load_lanes_direct): New macro.
10650 (mask_store_lanes_direct): Likewise.
10651 (expand_mask_load_optab_fn): Handle masked operations.
10652 (expand_mask_load_lanes_optab_fn): New macro.
10653 (expand_mask_store_optab_fn): Handle masked operations.
10654 (expand_mask_store_lanes_optab_fn): New macro.
10655 (direct_mask_load_lanes_optab_supported_p): Likewise.
10656 (direct_mask_store_lanes_optab_supported_p): Likewise.
10657 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10659 (vect_load_lanes_supported): Likewise.
10660 * tree-vect-data-refs.c (strip_conversion): New function.
10661 (can_group_stmts_p): Likewise.
10662 (vect_analyze_data_ref_accesses): Use it instead of checking
10663 for a pair of assignments.
10664 (vect_store_lanes_supported): Take a masked_p parameter.
10665 (vect_load_lanes_supported): Likewise.
10666 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10667 vect_store_lanes_supported and vect_load_lanes_supported.
10668 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10669 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10670 parameter. Don't allow gaps for masked accesses.
10671 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10672 and vect_load_lanes_supported.
10673 (get_load_store_type): Take a masked_p parameter and update
10674 call to get_group_load_store_type.
10675 (vectorizable_store): Update call to get_load_store_type.
10676 Handle IFN_MASK_STORE_LANES.
10677 (vectorizable_load): Update call to get_load_store_type.
10678 Handle IFN_MASK_LOAD_LANES.
10680 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10681 Alan Hayward <alan.hayward@arm.com>
10682 David Sherwood <david.sherwood@arm.com>
10684 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10686 * config/aarch64/aarch64-protos.h
10687 (aarch64_sve_struct_memory_operand_p): Declare.
10688 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10689 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10690 (VPRED, vpred): Handle SVE structure modes.
10691 * config/aarch64/constraints.md (Utx): New constraint.
10692 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10693 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10694 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10695 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10696 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10697 structure modes. Split into pieces after RA.
10698 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10699 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10701 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10702 SVE structure modes.
10703 (aarch64_classify_address): Likewise.
10704 (sizetochar): Move earlier in file.
10705 (aarch64_print_operand): Handle SVE register lists.
10706 (aarch64_array_mode): New function.
10707 (aarch64_sve_struct_memory_operand_p): Likewise.
10708 (TARGET_ARRAY_MODE): Redefine.
10710 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10711 Alan Hayward <alan.hayward@arm.com>
10712 David Sherwood <david.sherwood@arm.com>
10714 * target.def (array_mode): New target hook.
10715 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10716 * doc/tm.texi: Regenerate.
10717 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10718 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10719 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10720 targetm.array_mode.
10721 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10724 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10725 Alan Hayward <alan.hayward@arm.com>
10726 David Sherwood <david.sherwood@arm.com>
10728 * fold-const.c (fold_binary_loc): Check the argument types
10729 rather than the result type when testing for a vector operation.
10731 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10733 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10734 * doc/tm.texi: Regenerate.
10736 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10737 Alan Hayward <alan.hayward@arm.com>
10738 David Sherwood <david.sherwood@arm.com>
10740 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10741 (sve): Document new AArch64 extension.
10742 * doc/md.texi (w): Extend the description of the AArch64
10743 constraint to include SVE vectors.
10744 (Upl, Upa): Document new AArch64 predicate constraints.
10745 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10747 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10748 (msve-vector-bits=): New option.
10749 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10750 SVE when these are disabled.
10751 (sve): New extension.
10752 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10753 modes. Adjust their number of units based on aarch64_sve_vg.
10754 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10755 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10756 aarch64_addr_query_type.
10757 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10758 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10759 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10760 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10761 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10762 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10763 (aarch64_simd_imm_zero_p): Delete.
10764 (aarch64_check_zero_based_sve_index_immediate): Declare.
10765 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10766 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10767 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10768 (aarch64_sve_float_mul_immediate_p): Likewise.
10769 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10770 rather than an rtx.
10771 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10772 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10773 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10774 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10775 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10776 (aarch64_regmode_natural_size): Likewise.
10777 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10778 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10780 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10781 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10782 for VG and the SVE predicate registers.
10783 (V_ALIASES): Add a "z"-prefixed alias.
10784 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10785 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10786 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10787 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10788 (REG_CLASS_NAMES): Add entries for them.
10789 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10790 and the predicate registers.
10791 (aarch64_sve_vg): Declare.
10792 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10793 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10794 (REGMODE_NATURAL_SIZE): Define.
10795 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10797 * config/aarch64/aarch64.c: Include cfgrtl.h.
10798 (simd_immediate_info): Add a constructor for series vectors,
10799 and an associated step field.
10800 (aarch64_sve_vg): New variable.
10801 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10802 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10803 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10804 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10805 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10806 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10807 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10808 (aarch64_get_mask_mode): New functions.
10809 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10810 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10811 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10812 predicate modes and predicate registers. Explicitly restrict
10813 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10814 to store a vector mode if it is recognized by
10815 aarch64_classify_vector_mode.
10816 (aarch64_regmode_natural_size): New function.
10817 (aarch64_hard_regno_caller_save_mode): Return the original mode
10819 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10820 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10821 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10822 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10824 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10825 does not overlap dest if the function is frame-related. Handle
10827 (aarch64_split_add_offset): New function.
10828 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10829 them aarch64_add_offset.
10830 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10831 and update call to aarch64_sub_sp.
10832 (aarch64_add_cfa_expression): New function.
10833 (aarch64_expand_prologue): Pass extra temporary registers to the
10834 functions above. Handle the case in which we need to emit new
10835 DW_CFA_expressions for registers that were originally saved
10836 relative to the stack pointer, but now have to be expressed
10837 relative to the frame pointer.
10838 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10840 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10841 IP0 and IP1 values for SVE frames.
10842 (aarch64_expand_vec_series): New function.
10843 (aarch64_expand_sve_widened_duplicate): Likewise.
10844 (aarch64_expand_sve_const_vector): Likewise.
10845 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10846 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10847 into the register, rather than emitting a SET directly.
10848 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10849 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10850 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10851 (offset_9bit_signed_scaled_p): New functions.
10852 (aarch64_replicate_bitmask_imm): New function.
10853 (aarch64_bitmask_imm): Use it.
10854 (aarch64_cannot_force_const_mem): Reject expressions involving
10855 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10856 (aarch64_classify_index): Handle SVE indices, by requiring
10857 a plain register index with a scale that matches the element size.
10858 (aarch64_classify_address): Handle SVE addresses. Assert that
10859 the mode of the address is VOIDmode or an integer mode.
10860 Update call to aarch64_classify_symbol.
10861 (aarch64_classify_symbolic_expression): Update call to
10862 aarch64_classify_symbol.
10863 (aarch64_const_vec_all_in_range_p): New function.
10864 (aarch64_print_vector_float_operand): Likewise.
10865 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10866 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10867 and the FP immediates 1.0 and 0.5.
10868 (aarch64_print_address_internal): Handle SVE addresses.
10869 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10870 (aarch64_regno_regclass): Handle predicate registers.
10871 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10873 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10874 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10875 (aarch64_convert_sve_vector_bits): New function.
10876 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10877 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10878 rather than an rtx.
10879 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10880 Handle SVE vector and predicate modes. Accept VL-based constants
10881 that need only one temporary register, and VL offsets that require
10882 no temporary registers.
10883 (aarch64_conditional_register_usage): Mark the predicate registers
10884 as fixed if SVE isn't available.
10885 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10886 Return true for SVE vector and predicate modes.
10887 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10888 rather than an unsigned int. Handle SVE modes.
10889 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10891 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10893 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10894 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10895 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10896 (aarch64_sve_float_mul_immediate_p): New functions.
10897 (aarch64_sve_valid_immediate): New function.
10898 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10899 Explicitly reject structure modes. Check for INDEX constants.
10900 Handle PTRUE and PFALSE constants.
10901 (aarch64_check_zero_based_sve_index_immediate): New function.
10902 (aarch64_simd_imm_zero_p): Delete.
10903 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10904 vector modes. Accept constants in the range of CNT[BHWD].
10905 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10906 ask for an Advanced SIMD mode.
10907 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10908 (aarch64_simd_vector_alignment): Handle SVE predicates.
10909 (aarch64_vectorize_preferred_vector_alignment): New function.
10910 (aarch64_simd_vector_alignment_reachable): Use it instead of
10912 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10913 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10915 (MAX_VECT_LEN): Delete.
10916 (expand_vec_perm_d): Add a vec_flags field.
10917 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10918 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10919 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10921 (aarch64_evpc_rev): Rename to...
10922 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10923 (aarch64_evpc_rev_global): New function.
10924 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10925 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10927 (aarch64_evpc_sve_tbl): New function.
10928 (aarch64_expand_vec_perm_const_1): Update after rename of
10929 aarch64_evpc_rev. Handle SVE permutes too, trying
10930 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10931 than aarch64_evpc_tbl.
10932 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10933 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10934 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10935 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10936 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10937 (aarch64_expand_sve_vcond): New functions.
10938 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10939 of aarch64_vector_mode_p.
10940 (aarch64_dwarf_poly_indeterminate_value): New function.
10941 (aarch64_compute_pressure_classes): Likewise.
10942 (aarch64_can_change_mode_class): Likewise.
10943 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10944 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10945 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10946 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10947 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10948 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10949 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10950 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10952 (Dn, Dl, Dr): Accept const as well as const_vector.
10953 (Dz): Likewise. Compare against CONST0_RTX.
10954 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10955 of "vector" where appropriate.
10956 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10957 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10958 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10959 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10960 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10961 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10962 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10963 (v_int_equiv): Extend to SVE modes.
10964 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10966 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10967 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10968 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10969 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10970 (SVE_COND_FP_CMP): New int iterators.
10971 (perm_hilo): Handle the new unpack unspecs.
10972 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10974 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10975 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10976 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10977 (aarch64_equality_operator, aarch64_constant_vector_operand)
10978 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10979 (aarch64_sve_nonimmediate_operand): Likewise.
10980 (aarch64_sve_general_operand): Likewise.
10981 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10982 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10983 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10984 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10985 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10986 (aarch64_sve_float_arith_immediate): Likewise.
10987 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10988 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10989 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10990 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10991 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10992 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10993 (aarch64_sve_float_arith_operand): Likewise.
10994 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10995 (aarch64_sve_float_mul_operand): Likewise.
10996 (aarch64_sve_vec_perm_operand): Likewise.
10997 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10998 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10999 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11000 as well as const_vector.
11001 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11002 in file. Use CONST0_RTX and CONSTM1_RTX.
11003 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11004 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11005 Use aarch64_simd_imm_zero.
11006 * config/aarch64/aarch64-sve.md: New file.
11007 * config/aarch64/aarch64.md: Include it.
11008 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11009 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11010 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11011 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11012 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11013 (sve): New attribute.
11014 (enabled): Disable instructions with the sve attribute unless
11016 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11017 aarch64_expand_mov_immediate.
11018 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11019 CNT[BHSD] immediates.
11020 (movti): Split CONST_POLY_INT moves into two halves.
11021 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11022 Split additions that need a temporary here if the destination
11023 is the stack pointer.
11024 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11025 (*add<mode>3_poly_1): New instruction.
11026 (set_clobber_cc): New expander.
11028 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11030 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11031 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11032 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11033 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11034 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11035 Change innermode from fixed_mode_size to machine_mode.
11036 (simplify_subreg): Update call accordingly. Handle a constant-sized
11037 subreg of a variable-length CONST_VECTOR.
11039 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11040 Alan Hayward <alan.hayward@arm.com>
11041 David Sherwood <david.sherwood@arm.com>
11043 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11044 (add_offset_to_base): New function, split out from...
11045 (create_mem_ref): ...here. When handling a scale other than 1,
11046 check first whether the address is valid without the offset.
11047 Add it into the base if so, leaving the index and scale as-is.
11049 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11052 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11053 fold_for_warn before checking if arg2 is INTEGER_CST.
11055 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11057 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11058 (store_multiple_operation): Delete.
11059 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11060 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11061 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11062 guarded by TARGET_STRING.
11063 (rs6000_output_load_multiple): Delete.
11064 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11065 OPTION_MASK_STRING / TARGET_STRING handling.
11066 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11067 (const rs6000_opt_masks) <"string">: Change mask to 0.
11068 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11069 (MASK_STRING): Delete.
11070 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11072 (load_multiple): Delete.
11079 (store_multiple): Delete.
11086 (movmemsi_8reg): Delete.
11087 (corresponding unnamed define_insn): Delete.
11088 (movmemsi_6reg): Delete.
11089 (corresponding unnamed define_insn): Delete.
11090 (movmemsi_4reg): Delete.
11091 (corresponding unnamed define_insn): Delete.
11092 (movmemsi_2reg): Delete.
11093 (corresponding unnamed define_insn): Delete.
11094 (movmemsi_1reg): Delete.
11095 (corresponding unnamed define_insn): Delete.
11096 * config/rs6000/rs6000.opt (mno-string): New.
11097 (mstring): Replace by deprecation warning stub.
11098 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11100 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11102 * regrename.c (regrename_do_replace): If replacing the same
11103 reg multiple times, try to reuse last created gen_raw_REG.
11106 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11107 main to workaround a bug in GDB.
11109 2018-01-12 Tom de Vries <tom@codesourcery.com>
11112 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11114 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11116 PR rtl-optimization/80481
11117 * ira-color.c (get_cap_member): New function.
11118 (allocnos_conflict_by_live_ranges_p): Use it.
11119 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11120 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11122 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11125 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11126 (*saddl_se_1): Ditto.
11127 (*ssubsi_1): Ditto.
11128 (*ssubl_se_1): Ditto.
11130 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11132 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11133 rather than wi::to_widest for DR_INITs.
11134 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11135 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11136 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11138 (vect_analyze_group_access_1): Note that here.
11140 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11142 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11143 polynomial type sizes.
11145 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11147 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11148 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11149 (gimple_add_tmp_var): Likewise.
11151 2018-01-12 Martin Liska <mliska@suse.cz>
11153 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11154 (gimple_alloc_sizes): Likewise.
11155 (dump_gimple_statistics): Use PRIu64 in printf format.
11156 * gimple.h: Change uint64_t to int.
11158 2018-01-12 Martin Liska <mliska@suse.cz>
11160 * tree-core.h: Use uint64_t instead of int.
11161 * tree.c (tree_node_counts): Likewise.
11162 (tree_node_sizes): Likewise.
11163 (dump_tree_statistics): Use PRIu64 in printf format.
11165 2018-01-12 Martin Liska <mliska@suse.cz>
11167 * Makefile.in: As qsort_chk is implemented in vec.c, add
11168 vec.o to linkage of gencfn-macros.
11169 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11170 passing the info to record_node_allocation_statistics.
11171 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11173 * ggc-common.c (struct ggc_usage): Add operator== and use
11174 it in operator< and compare function.
11175 * mem-stats.h (struct mem_usage): Likewise.
11176 * vec.c (struct vec_usage): Remove operator< and compare
11177 function. Can be simply inherited.
11179 2018-01-12 Martin Jambor <mjambor@suse.cz>
11182 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11183 * tree-ssa-math-opts.c: Include domwalk.h.
11184 (convert_mult_to_fma_1): New function.
11185 (fma_transformation_info): New type.
11186 (fma_deferring_state): Likewise.
11187 (cancel_fma_deferring): New function.
11188 (result_of_phi): Likewise.
11189 (last_fma_candidate_feeds_initial_phi): Likewise.
11190 (convert_mult_to_fma): Added deferring logic, split actual
11191 transformation to convert_mult_to_fma_1.
11192 (math_opts_dom_walker): New type.
11193 (math_opts_dom_walker::after_dom_children): New method, body moved
11194 here from pass_optimize_widening_mul::execute, added deferring logic
11196 (pass_optimize_widening_mul::execute): Moved most of code to
11197 math_opts_dom_walker::after_dom_children.
11198 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11199 * config/i386/i386.c (ix86_option_override_internal): Added
11200 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11202 2018-01-12 Richard Biener <rguenther@suse.de>
11205 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11206 inline instance vars.
11208 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11211 * config/rx/rx.c (rx_is_restricted_memory_address):
11212 Handle SUBREG case.
11214 2018-01-12 Richard Biener <rguenther@suse.de>
11216 PR tree-optimization/80846
11217 * target.def (split_reduction): New target hook.
11218 * targhooks.c (default_split_reduction): New function.
11219 * targhooks.h (default_split_reduction): Declare.
11220 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11221 target requests first reduce vectors by combining low and high
11223 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11224 (get_vectype_for_scalar_type_and_size): Export.
11225 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11226 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11227 * doc/tm.texi: Regenerate.
11228 * config/i386/i386.c (ix86_split_reduction): Implement
11229 TARGET_VECTORIZE_SPLIT_REDUCTION.
11231 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11234 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11235 in PIC mode except for TARGET_VXWORKS_RTP.
11236 * config/sparc/sparc.c: Include cfgrtl.h.
11237 (TARGET_INIT_PIC_REG): Define.
11238 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11239 (sparc_pic_register_p): New predicate.
11240 (sparc_legitimate_address_p): Use it.
11241 (sparc_legitimize_pic_address): Likewise.
11242 (sparc_delegitimize_address): Likewise.
11243 (sparc_mode_dependent_address_p): Likewise.
11244 (gen_load_pcrel_sym): Remove 4th parameter.
11245 (load_got_register): Adjust call to above. Remove obsolete stuff.
11246 (sparc_expand_prologue): Do not call load_got_register here.
11247 (sparc_flat_expand_prologue): Likewise.
11248 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11249 (sparc_use_pseudo_pic_reg): New function.
11250 (sparc_init_pic_reg): Likewise.
11251 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11252 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11254 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11256 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11257 Add item for branch_cost.
11259 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11261 PR rtl-optimization/83565
11262 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11263 not extend the result to a larger mode for rotate operations.
11264 (num_sign_bit_copies1): Likewise.
11266 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11269 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11271 Use values-Xc.o for -pedantic.
11272 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11274 2018-01-12 Martin Liska <mliska@suse.cz>
11277 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11279 (possible_polymorphic_call_targets): Use it.
11280 (ipa_devirt): Likewise.
11282 2018-01-12 Martin Liska <mliska@suse.cz>
11284 * profile-count.h (enum profile_quality): Use 0 as invalid
11285 enum value of profile_quality.
11287 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11289 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11290 -mext-string options.
11292 2018-01-12 Richard Biener <rguenther@suse.de>
11294 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11295 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11296 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11298 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11300 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11302 * configure.ac (--with-long-double-format): Add support for the
11303 configuration option to change the default long double format on
11305 * config.gcc (powerpc*-linux*-*): Likewise.
11306 * configure: Regenerate.
11307 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11308 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11309 used without modification.
11311 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11313 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11314 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11315 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11316 MISC_BUILTIN_SPEC_BARRIER.
11317 (rs6000_init_builtins): Likewise.
11318 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11320 (speculation_barrier): New define_insn.
11321 * doc/extend.texi: Document __builtin_speculation_barrier.
11323 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11326 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11327 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11328 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11330 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11331 integral modes instead of "ss" and "sd".
11332 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11333 vectors with 32-bit and 64-bit elements.
11334 (vecdupssescalarmodesuffix): New mode attribute.
11335 (vec_dup<mode>): Use it.
11337 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11340 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11341 frame if argument is passed on stack.
11343 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11346 * ree.c (combine_reaching_defs): Optimize also
11347 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11348 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11350 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11352 PR middle-end/83189
11353 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11355 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11357 PR middle-end/83718
11358 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11359 after they are computed.
11361 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11363 PR tree-optimization/83695
11364 * gimple-loop-linterchange.cc
11365 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11366 reset cached scev information after interchange.
11367 (pass_linterchange::execute): Remove call to scev_reset_htab.
11369 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11371 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11372 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11373 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11374 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11375 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11376 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11377 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11378 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11379 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11380 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11381 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11382 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11383 (V_lane_reg): Likewise.
11384 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11386 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11387 (vfmal_lane_low<mode>_intrinsic,
11388 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11389 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11390 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11391 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11392 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11393 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11395 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11397 * config/arm/arm-cpus.in (fp16fml): New feature.
11398 (ALL_SIMD): Add fp16fml.
11399 (armv8.2-a): Add fp16fml as an option.
11400 (armv8.3-a): Likewise.
11401 (armv8.4-a): Add fp16fml as part of fp16.
11402 * config/arm/arm.h (TARGET_FP16FML): Define.
11403 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11405 * config/arm/arm-modes.def (V2HF): Define.
11406 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11407 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11408 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11409 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11410 vfmsl_low, vfmsl_high): New set of builtins.
11411 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11412 (vfml_op): New code attribute.
11413 (VFMLHALVES): New int iterator.
11414 (VFML, VFMLSEL): New mode attributes.
11415 (V_reg): Define mapping for V2HF.
11416 (V_hi, V_lo): New mode attributes.
11417 (VF_constraint): Likewise.
11418 (vfml_half, vfml_half_selector): New int attributes.
11419 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11421 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11422 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11424 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11425 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11426 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11427 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11429 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11430 Document new effective target and option set.
11432 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11434 * config/arm/arm-cpus.in (armv8_4): New feature.
11435 (ARMv8_4a): New fgroup.
11436 (armv8.4-a): New arch.
11437 * config/arm/arm-tables.opt: Regenerate.
11438 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11439 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11440 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11441 Add matching rules for -march=armv8.4-a and extensions.
11442 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11444 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11447 * config/rx/rx.md (BW): New mode attribute.
11448 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11450 2018-01-11 Richard Biener <rguenther@suse.de>
11452 PR tree-optimization/83435
11453 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11454 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11455 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11457 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11458 Alan Hayward <alan.hayward@arm.com>
11459 David Sherwood <david.sherwood@arm.com>
11461 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11463 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11464 (aarch64_print_address_internal): Use it to check for a zero offset.
11466 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11467 Alan Hayward <alan.hayward@arm.com>
11468 David Sherwood <david.sherwood@arm.com>
11470 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11471 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11472 Return a poly_int64 rather than a HOST_WIDE_INT.
11473 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11474 rather than a HOST_WIDE_INT.
11475 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11476 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11477 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11478 final_offset from HOST_WIDE_INT to poly_int64.
11479 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11480 to_constant when getting the number of units in an Advanced SIMD
11482 (aarch64_builtin_vectorized_function): Check for a constant number
11484 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11486 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11487 attribute instead of GET_MODE_NUNITS.
11488 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11489 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11490 GET_MODE_SIZE for fixed-size registers.
11491 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11492 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11493 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11494 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11495 (aarch64_print_operand, aarch64_print_address_internal)
11496 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11497 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11498 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11499 Handle polynomial GET_MODE_SIZE.
11500 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11501 wider than SImode without modification.
11502 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11503 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11504 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11505 passing and returning SVE modes.
11506 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11507 rather than GEN_INT.
11508 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11509 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11510 (aarch64_allocate_and_probe_stack_space): Likewise.
11511 (aarch64_layout_frame): Cope with polynomial offsets.
11512 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11513 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11514 polynomial offsets.
11515 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11516 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11517 poly_int64 rather than a HOST_WIDE_INT.
11518 (aarch64_get_separate_components, aarch64_process_components)
11519 (aarch64_expand_prologue, aarch64_expand_epilogue)
11520 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11521 (aarch64_anchor_offset): New function, split out from...
11522 (aarch64_legitimize_address): ...here.
11523 (aarch64_builtin_vectorization_cost): Handle polynomial
11524 TYPE_VECTOR_SUBPARTS.
11525 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11527 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11528 number of elements from the PARALLEL rather than the mode.
11529 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11530 rather than GET_MODE_BITSIZE.
11531 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11532 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11533 (aarch64_expand_vec_perm_const_1): Handle polynomial
11534 d->perm.length () and d->perm elements.
11535 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11536 Apply to_constant to d->perm elements.
11537 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11538 polynomial CONST_VECTOR_NUNITS.
11539 (aarch64_move_pointer): Take amount as a poly_int64 rather
11541 (aarch64_progress_pointer): Avoid temporary variable.
11542 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11543 the mode attribute instead of GET_MODE.
11545 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11546 Alan Hayward <alan.hayward@arm.com>
11547 David Sherwood <david.sherwood@arm.com>
11549 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11550 x exists before using it.
11551 (aarch64_add_constant_internal): Rename to...
11552 (aarch64_add_offset_1): ...this. Replace regnum with separate
11553 src and dest rtxes. Handle the case in which they're different,
11554 including when the offset is zero. Replace scratchreg with an rtx.
11555 Use 2 additions if there is no spare register into which we can
11556 move a 16-bit constant.
11557 (aarch64_add_constant): Delete.
11558 (aarch64_add_offset): Replace reg with separate src and dest
11559 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11560 Use aarch64_add_offset_1.
11561 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11562 an rtx rather than an int. Take the delta as a poly_int64
11563 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11564 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11565 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11566 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11567 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11568 and aarch64_add_sp.
11569 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11570 aarch64_add_constant.
11572 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11574 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11575 Use scalar_float_mode.
11577 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11579 * config/aarch64/aarch64-simd.md
11580 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11581 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11582 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11583 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11584 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11585 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11586 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11587 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11588 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11589 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11591 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11594 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11595 targ_options->x_arm_arch_string is non NULL.
11597 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11599 * config/aarch64/aarch64.h
11600 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11602 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11605 * expmed.c (emit_store_flag_force): Swap if const op0
11606 and change VOIDmode to mode of op0.
11608 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11610 PR rtl-optimization/83761
11611 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11612 than bytes to mode_for_size.
11614 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11616 PR middle-end/83189
11617 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11618 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11621 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11623 PR middle-end/83575
11624 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11625 when in layout mode.
11626 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11627 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11630 2018-01-10 Michael Collison <michael.collison@arm.com>
11632 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11633 * config/aarch64/aarch64-option-extension.def: Add
11634 AARCH64_OPT_EXTENSION of 'fp16fml'.
11635 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11636 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11637 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11638 * config/aarch64/constraints.md (Ui7): New constraint.
11639 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11640 (VFMLA_SEL_W): Ditto.
11643 (VFMLA16_LOW): New int iterator.
11644 (VFMLA16_HIGH): Ditto.
11645 (UNSPEC_FMLAL): New unspec.
11646 (UNSPEC_FMLSL): Ditto.
11647 (UNSPEC_FMLAL2): Ditto.
11648 (UNSPEC_FMLSL2): Ditto.
11649 (f16mac): New code attribute.
11650 * config/aarch64/aarch64-simd-builtins.def
11651 (aarch64_fmlal_lowv2sf): Ditto.
11652 (aarch64_fmlsl_lowv2sf): Ditto.
11653 (aarch64_fmlalq_lowv4sf): Ditto.
11654 (aarch64_fmlslq_lowv4sf): Ditto.
11655 (aarch64_fmlal_highv2sf): Ditto.
11656 (aarch64_fmlsl_highv2sf): Ditto.
11657 (aarch64_fmlalq_highv4sf): Ditto.
11658 (aarch64_fmlslq_highv4sf): Ditto.
11659 (aarch64_fmlal_lane_lowv2sf): Ditto.
11660 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11661 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11662 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11663 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11664 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11665 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11666 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11667 (aarch64_fmlal_lane_highv2sf): Ditto.
11668 (aarch64_fmlsl_lane_highv2sf): Ditto.
11669 (aarch64_fmlal_laneq_highv2sf): Ditto.
11670 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11671 (aarch64_fmlalq_lane_highv4sf): Ditto.
11672 (aarch64_fmlsl_lane_highv4sf): Ditto.
11673 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11674 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11675 * config/aarch64/aarch64-simd.md:
11676 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11677 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11678 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11679 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11680 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11681 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11682 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11683 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11684 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11685 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11686 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11687 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11688 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11689 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11690 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11691 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11692 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11693 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11694 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11695 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11696 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11697 (vfmlsl_low_u32): Ditto.
11698 (vfmlalq_low_u32): Ditto.
11699 (vfmlslq_low_u32): Ditto.
11700 (vfmlal_high_u32): Ditto.
11701 (vfmlsl_high_u32): Ditto.
11702 (vfmlalq_high_u32): Ditto.
11703 (vfmlslq_high_u32): Ditto.
11704 (vfmlal_lane_low_u32): Ditto.
11705 (vfmlsl_lane_low_u32): Ditto.
11706 (vfmlal_laneq_low_u32): Ditto.
11707 (vfmlsl_laneq_low_u32): Ditto.
11708 (vfmlalq_lane_low_u32): Ditto.
11709 (vfmlslq_lane_low_u32): Ditto.
11710 (vfmlalq_laneq_low_u32): Ditto.
11711 (vfmlslq_laneq_low_u32): Ditto.
11712 (vfmlal_lane_high_u32): Ditto.
11713 (vfmlsl_lane_high_u32): Ditto.
11714 (vfmlal_laneq_high_u32): Ditto.
11715 (vfmlsl_laneq_high_u32): Ditto.
11716 (vfmlalq_lane_high_u32): Ditto.
11717 (vfmlslq_lane_high_u32): Ditto.
11718 (vfmlalq_laneq_high_u32): Ditto.
11719 (vfmlslq_laneq_high_u32): Ditto.
11720 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11721 (AARCH64_FL_FOR_ARCH8_4): New.
11722 (AARCH64_ISA_F16FML): New ISA flag.
11723 (TARGET_F16FML): New feature flag for fp16fml.
11724 (doc/invoke.texi): Document new fp16fml option.
11726 2018-01-10 Michael Collison <michael.collison@arm.com>
11728 * config/aarch64/aarch64-builtins.c:
11729 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11730 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11731 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11732 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11733 (AARCH64_ISA_SHA3): New ISA flag.
11734 (TARGET_SHA3): New feature flag for sha3.
11735 * config/aarch64/iterators.md (sha512_op): New int attribute.
11736 (CRYPTO_SHA512): New int iterator.
11737 (UNSPEC_SHA512H): New unspec.
11738 (UNSPEC_SHA512H2): Ditto.
11739 (UNSPEC_SHA512SU0): Ditto.
11740 (UNSPEC_SHA512SU1): Ditto.
11741 * config/aarch64/aarch64-simd-builtins.def
11742 (aarch64_crypto_sha512hqv2di): New builtin.
11743 (aarch64_crypto_sha512h2qv2di): Ditto.
11744 (aarch64_crypto_sha512su0qv2di): Ditto.
11745 (aarch64_crypto_sha512su1qv2di): Ditto.
11746 (aarch64_eor3qv8hi): Ditto.
11747 (aarch64_rax1qv2di): Ditto.
11748 (aarch64_xarqv2di): Ditto.
11749 (aarch64_bcaxqv8hi): Ditto.
11750 * config/aarch64/aarch64-simd.md:
11751 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11752 (aarch64_crypto_sha512su0qv2di): Ditto.
11753 (aarch64_crypto_sha512su1qv2di): Ditto.
11754 (aarch64_eor3qv8hi): Ditto.
11755 (aarch64_rax1qv2di): Ditto.
11756 (aarch64_xarqv2di): Ditto.
11757 (aarch64_bcaxqv8hi): Ditto.
11758 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11759 (vsha512h2q_u64): Ditto.
11760 (vsha512su0q_u64): Ditto.
11761 (vsha512su1q_u64): Ditto.
11762 (veor3q_u16): Ditto.
11763 (vrax1q_u64): Ditto.
11764 (vxarq_u64): Ditto.
11765 (vbcaxq_u16): Ditto.
11766 * config/arm/types.md (crypto_sha512): New type attribute.
11767 (crypto_sha3): Ditto.
11768 (doc/invoke.texi): Document new sha3 option.
11770 2018-01-10 Michael Collison <michael.collison@arm.com>
11772 * config/aarch64/aarch64-builtins.c:
11773 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11774 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11775 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11776 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11777 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11778 (AARCH64_ISA_SM4): New ISA flag.
11779 (TARGET_SM4): New feature flag for sm4.
11780 * config/aarch64/aarch64-simd-builtins.def
11781 (aarch64_sm3ss1qv4si): Ditto.
11782 (aarch64_sm3tt1aq4si): Ditto.
11783 (aarch64_sm3tt1bq4si): Ditto.
11784 (aarch64_sm3tt2aq4si): Ditto.
11785 (aarch64_sm3tt2bq4si): Ditto.
11786 (aarch64_sm3partw1qv4si): Ditto.
11787 (aarch64_sm3partw2qv4si): Ditto.
11788 (aarch64_sm4eqv4si): Ditto.
11789 (aarch64_sm4ekeyqv4si): Ditto.
11790 * config/aarch64/aarch64-simd.md:
11791 (aarch64_sm3ss1qv4si): Ditto.
11792 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11793 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11794 (aarch64_sm4eqv4si): Ditto.
11795 (aarch64_sm4ekeyqv4si): Ditto.
11796 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11797 (sm3part_op): Ditto.
11798 (CRYPTO_SM3TT): Ditto.
11799 (CRYPTO_SM3PART): Ditto.
11800 (UNSPEC_SM3SS1): New unspec.
11801 (UNSPEC_SM3TT1A): Ditto.
11802 (UNSPEC_SM3TT1B): Ditto.
11803 (UNSPEC_SM3TT2A): Ditto.
11804 (UNSPEC_SM3TT2B): Ditto.
11805 (UNSPEC_SM3PARTW1): Ditto.
11806 (UNSPEC_SM3PARTW2): Ditto.
11807 (UNSPEC_SM4E): Ditto.
11808 (UNSPEC_SM4EKEY): Ditto.
11809 * config/aarch64/constraints.md (Ui2): New constraint.
11810 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11811 * config/arm/types.md (crypto_sm3): New type attribute.
11812 (crypto_sm4): Ditto.
11813 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11814 (vsm3tt1aq_u32): Ditto.
11815 (vsm3tt1bq_u32): Ditto.
11816 (vsm3tt2aq_u32): Ditto.
11817 (vsm3tt2bq_u32): Ditto.
11818 (vsm3partw1q_u32): Ditto.
11819 (vsm3partw2q_u32): Ditto.
11820 (vsm4eq_u32): Ditto.
11821 (vsm4ekeyq_u32): Ditto.
11822 (doc/invoke.texi): Document new sm4 option.
11824 2018-01-10 Michael Collison <michael.collison@arm.com>
11826 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11827 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11828 (AARCH64_FL_FOR_ARCH8_4): New.
11829 (AARCH64_FL_V8_4): New flag.
11830 (doc/invoke.texi): Document new armv8.4-a option.
11832 2018-01-10 Michael Collison <michael.collison@arm.com>
11834 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11835 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11836 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11837 * config/aarch64/aarch64-option-extension.def: Add
11838 AARCH64_OPT_EXTENSION of 'sha2'.
11839 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11840 (crypto): Disable sha2 and aes if crypto disabled.
11841 (crypto): Enable aes and sha2 if enabled.
11842 (simd): Disable sha2 and aes if simd disabled.
11843 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11845 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11846 (TARGET_SHA2): New feature flag for sha2.
11847 (TARGET_AES): New feature flag for aes.
11848 * config/aarch64/aarch64-simd.md:
11849 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11850 conditional on TARGET_AES.
11851 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11852 (aarch64_crypto_sha1hsi): Make pattern conditional
11854 (aarch64_crypto_sha1hv4si): Ditto.
11855 (aarch64_be_crypto_sha1hv4si): Ditto.
11856 (aarch64_crypto_sha1su1v4si): Ditto.
11857 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11858 (aarch64_crypto_sha1su0v4si): Ditto.
11859 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11860 (aarch64_crypto_sha256su0v4si): Ditto.
11861 (aarch64_crypto_sha256su1v4si): Ditto.
11862 (doc/invoke.texi): Document new aes and sha2 options.
11864 2018-01-10 Martin Sebor <msebor@redhat.com>
11866 PR tree-optimization/83781
11867 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11870 2018-01-11 Martin Sebor <msebor@gmail.com>
11871 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11873 PR tree-optimization/83501
11874 PR tree-optimization/81703
11876 * tree-ssa-strlen.c (get_string_cst): Rename...
11877 (get_string_len): ...to this. Handle global constants.
11878 (handle_char_store): Adjust.
11880 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11881 Jim Wilson <jimw@sifive.com>
11883 * config/riscv/riscv-protos.h (riscv_output_return): New.
11884 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11885 (riscv_attribute_table, riscv_output_return),
11886 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11887 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11888 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11890 (riscv_expand_prologue): Add early return for naked function.
11891 (riscv_expand_epilogue): Likewise.
11892 (riscv_function_ok_for_sibcall): Return false for naked function.
11893 (riscv_set_current_function): New.
11894 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11895 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11896 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11897 * doc/extend.texi (RISC-V Function Attributes): New.
11899 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11901 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11902 check for 128-bit long double before checking TCmode.
11903 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11904 128-bit long doubles before checking TFmode or TCmode.
11905 (FLOAT128_IBM_P): Likewise.
11907 2018-01-10 Martin Sebor <msebor@redhat.com>
11909 PR tree-optimization/83671
11910 * builtins.c (c_strlen): Unconditionally return zero for the empty
11912 Use -Warray-bounds for warnings.
11913 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11914 for non-constant array indices with COMPONENT_REF, arrays of
11915 arrays, and pointers to arrays.
11916 (gimple_fold_builtin_strlen): Determine and set length range for
11917 non-constant character arrays.
11919 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11921 PR middle-end/81897
11922 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11925 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11927 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11929 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11932 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11933 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11934 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11935 indexed_or_indirect_operand predicate.
11936 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11937 (*vsx_le_perm_load_v8hi): Likewise.
11938 (*vsx_le_perm_load_v16qi): Likewise.
11939 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11940 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11941 (*vsx_le_perm_store_v8hi): Likewise.
11942 (*vsx_le_perm_store_v16qi): Likewise.
11943 (eight unnamed splitters): Likewise.
11945 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11947 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11948 * config/rs6000/emmintrin.h: Likewise.
11949 * config/rs6000/mmintrin.h: Likewise.
11950 * config/rs6000/xmmintrin.h: Likewise.
11952 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11955 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11957 * tree.c (tree_nop_conversion): Return true for location wrapper
11959 (maybe_wrap_with_location): New function.
11960 (selftest::check_strip_nops): New function.
11961 (selftest::test_location_wrappers): New function.
11962 (selftest::tree_c_tests): Call it.
11963 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11964 (maybe_wrap_with_location): New decl.
11965 (EXPR_LOCATION_WRAPPER_P): New macro.
11966 (location_wrapper_p): New inline function.
11967 (tree_strip_any_location_wrapper): New inline function.
11969 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11972 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11973 stack_realign_offset for the largest alignment of stack slot
11975 (ix86_find_max_used_stack_alignment): New function.
11976 (ix86_finalize_stack_frame_flags): Use it. Set
11977 max_used_stack_alignment if we don't realign stack.
11978 * config/i386/i386.h (machine_function): Add
11979 max_used_stack_alignment.
11981 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11983 * config/arm/arm.opt (-mbranch-cost): New option.
11984 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11987 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11990 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11991 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11993 2018-01-10 Richard Biener <rguenther@suse.de>
11996 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11997 early out so it also covers the case where we have a non-NULL
12000 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12002 PR tree-optimization/83753
12003 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12004 for non-strided grouped accesses if the number of elements is 1.
12006 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12009 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12010 * i386.h (TARGET_USE_GATHER): Define.
12011 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12013 2018-01-10 Martin Liska <mliska@suse.cz>
12016 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12017 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12019 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12020 CLEANUP_NO_PARTITIONING is not set.
12022 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12024 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12025 for vectors, as a partial revert of r254296.
12026 * rtl.h (const_vec_p): Delete.
12027 (const_vec_duplicate_p): Don't test for vector CONSTs.
12028 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12029 * expmed.c (make_tree): Likewise.
12032 * common.md (E, F): Use CONSTANT_P instead of checking for
12034 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12035 checking for CONST_VECTOR.
12037 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12039 PR middle-end/83575
12040 * predict.c (force_edge_cold): Handle in more sane way edges
12041 with no prediction.
12043 2018-01-09 Carl Love <cel@us.ibm.com>
12045 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12047 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12048 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12049 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12050 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12051 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12052 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12053 * config/rs6000/rs6000-protos.h: Add extern defition for
12054 rs6000_generate_float2_double_code.
12055 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12057 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12058 (float2_v2df): Add define_expand.
12060 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12063 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12064 op_mode in the force_to_mode call.
12066 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12068 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12069 instead of checking each element individually.
12070 (aarch64_evpc_uzp): Likewise.
12071 (aarch64_evpc_zip): Likewise.
12072 (aarch64_evpc_ext): Likewise.
12073 (aarch64_evpc_rev): Likewise.
12074 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12075 instead of checking each element individually. Return true without
12077 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12078 whether all selected elements come from the same input, instead of
12079 checking each element individually. Remove calls to gen_rtx_REG,
12080 start_sequence and end_sequence and instead assert that no rtl is
12083 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12085 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12086 order of HIGH and CONST checks.
12088 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12090 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12091 if the destination isn't an SSA_NAME.
12093 2018-01-09 Richard Biener <rguenther@suse.de>
12095 PR tree-optimization/83668
12096 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12098 (canonicalize_loop_form): ... here, renamed from ...
12099 (canonicalize_loop_closed_ssa_form): ... this and amended to
12100 swap successor edges for loop exit blocks to make us use
12101 the RPO order we need for initial schedule generation.
12103 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12105 PR tree-optimization/64811
12106 * match.pd: When optimizing comparisons with Inf, avoid
12107 introducing or losing exceptions from comparisons with NaN.
12109 2018-01-09 Martin Liska <mliska@suse.cz>
12112 * asan.c (shadow_mem_size): Add gcc_assert.
12114 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12116 Don't save registers in main().
12119 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12120 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12121 * config/avr/avr.c (avr_set_current_function): Don't error if
12122 naked, OS_task or OS_main are specified at the same time.
12123 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12125 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12127 * common/config/avr/avr-common.c (avr_option_optimization_table):
12128 Switch on -mmain-is-OS_task for optimizing compilations.
12130 2018-01-09 Richard Biener <rguenther@suse.de>
12132 PR tree-optimization/83572
12133 * graphite.c: Include cfganal.h.
12134 (graphite_transform_loops): Connect infinite loops to exit
12135 and remove fake edges at the end.
12137 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12139 * ipa-inline.c (edge_badness): Revert accidental checkin.
12141 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12144 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12145 symbols; not inline clones.
12147 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12150 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12151 hard registers. Formatting fixes.
12153 PR preprocessor/83722
12154 * gcc.c (try_generate_repro): Pass
12155 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12156 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12159 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12160 Kito Cheng <kito.cheng@gmail.com>
12162 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12163 (riscv_leaf_function_p): Delete.
12164 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12166 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12168 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12170 (do_ifelse): New function.
12171 (do_isel): New function.
12172 (do_sub3): New function.
12173 (do_add3): New function.
12174 (do_load_mask_compare): New function.
12175 (do_overlap_load_compare): New function.
12176 (expand_compare_loop): New function.
12177 (expand_block_compare): Call expand_compare_loop() when appropriate.
12178 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12179 option description.
12180 (-mblock-compare-inline-loop-limit): New option.
12182 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12185 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12186 Reverse order of second and third operands in first alternative.
12187 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12188 of first and second elements in UNSPEC_VPERMR vector.
12189 (altivec_expand_vec_perm_le): Likewise.
12191 2018-01-08 Jeff Law <law@redhat.com>
12193 PR rtl-optimizatin/81308
12194 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12195 (process_switch): If group_case_labels makes a change, then set
12197 (pass_convert_switch::execute): If a switch is converted, then
12198 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12200 PR rtl-optimization/81308
12201 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12204 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12206 PR target/83663 - Revert r255946
12207 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12208 generation for cases where splatting a value is not useful.
12209 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12210 across a vec_duplicate and a paradoxical subreg forming a vector
12211 mode to a vec_concat.
12213 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12215 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12216 -march=armv8.3-a variants.
12217 * config/arm/t-multilib: Likewise.
12218 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12220 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12222 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12224 (cceq_ior_compare_complement): Give it a name so I can use it, and
12225 change boolean_or_operator predicate to boolean_operator so it can
12226 be used to generate a crand.
12227 (eqne): New code iterator.
12228 (bd/bd_neg): New code_attrs.
12229 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12230 a single define_insn.
12231 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12232 decrement (bdnzt/bdnzf/bdzt/bdzf).
12233 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12234 with the new names of the branch decrement patterns, and added the
12235 names of the branch decrement conditional patterns.
12237 2018-01-08 Richard Biener <rguenther@suse.de>
12239 PR tree-optimization/83563
12240 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12243 2018-01-08 Richard Biener <rguenther@suse.de>
12245 PR middle-end/83713
12246 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12248 2018-01-08 Richard Biener <rguenther@suse.de>
12250 PR tree-optimization/83685
12251 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12252 references to abnormals.
12254 2018-01-08 Richard Biener <rguenther@suse.de>
12257 * dwarf2out.c (output_indirect_strings): Handle empty
12258 skeleton_debug_str_hash.
12259 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12261 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12263 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12264 (emit_store_direct): Likewise.
12265 (arc_trampoline_adjust_address): Likewise.
12266 (arc_asm_trampoline_template): New function.
12267 (arc_initialize_trampoline): Use asm_trampoline_template.
12268 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12269 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12270 * config/arc/arc.md (flush_icache): Delete pattern.
12272 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12274 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12275 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12278 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12281 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12282 by not USED_FOR_TARGET.
12283 (make_pass_resolve_sw_modes): Likewise.
12285 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12287 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12290 2018-01-08 Richard Biener <rguenther@suse.de>
12292 PR middle-end/83580
12293 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12295 2018-01-08 Richard Biener <rguenther@suse.de>
12297 PR middle-end/83517
12298 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12300 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12302 PR middle-end/81897
12303 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12304 basic blocks with a small number of successors.
12305 (convert_control_dep_chain_into_preds): Improve handling of
12307 (dump_predicates): Split apart into...
12308 (dump_pred_chain): ...here...
12309 (dump_pred_info): ...and here.
12310 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12311 (can_chain_union_be_invalidated_p): Improve check for invalidation
12313 (uninit_uses_cannot_happen): Avoid unnecessary if
12314 convert_control_dep_chain_into_preds yielded nothing.
12316 2018-01-06 Martin Sebor <msebor@redhat.com>
12318 PR tree-optimization/83640
12319 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12320 subtracting negative offset from size.
12321 (builtin_access::overlap): Adjust offset bounds of the access to fall
12322 within the size of the object if possible.
12324 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12326 PR rtl-optimization/83699
12327 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12328 extract_bit_field_as_subreg to cases in which the extracted
12329 value is also a vector.
12331 * lra-constraints.c (process_alt_operands): Test for the equivalence
12332 substitutions when detecting a possible reload cycle.
12334 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12337 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12338 by default if flag_selective_schedling{,2}. Formatting fixes.
12340 PR rtl-optimization/83682
12341 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12342 if it has non-VECTOR_MODE element mode.
12343 (vec_duplicate_p): Likewise.
12345 PR middle-end/83694
12346 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12347 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12349 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12352 * config/i386/i386-builtin.def
12353 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12354 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12355 Require also OPTION_MASK_ISA_AVX512F in addition to
12356 OPTION_MASK_ISA_GFNI.
12357 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12358 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12359 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12360 to OPTION_MASK_ISA_GFNI.
12361 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12362 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12363 OPTION_MASK_ISA_AVX512BW.
12364 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12365 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12366 addition to OPTION_MASK_ISA_GFNI.
12367 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12368 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12369 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12370 to OPTION_MASK_ISA_GFNI.
12371 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12372 a requirement for all ISAs rather than any of them with a few
12374 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12376 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12377 bitmasks to be enabled with 3 exceptions, instead of requiring any
12378 enabled ISA with lots of exceptions.
12379 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12380 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12381 Change avx512bw in isa attribute to avx512f.
12382 * config/i386/sgxintrin.h: Add license boilerplate.
12383 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12384 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12385 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12386 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12388 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12389 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12390 temporarily sse2 rather than sse if not enabled already.
12393 * config/i386/sse.md (VI248_VLBW): Rename to ...
12394 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12395 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12396 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12397 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12398 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12399 mode iterator instead of VI248_VLBW.
12401 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12403 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12404 (record_modified): Skip clobbers; add debug output.
12405 (param_change_prob): Use sreal frequencies.
12407 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12409 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12410 punt for user-aligned variables.
12412 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12414 * tree-chrec.c (chrec_contains_symbols): Return true for
12417 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12420 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12421 of (x|y) == x for BICS pattern.
12423 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12425 PR tree-optimization/83605
12426 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12427 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12430 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12432 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12433 * config/epiphany/rtems.h: New file.
12435 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12436 Uros Bizjak <ubizjak@gmail.com>
12439 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12440 QIreg_operand instead of register_operand predicate.
12441 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12442 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12443 comments instead of -fmitigate[-_]rop.
12445 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12448 * cgraphunit.c (symbol_table::compile): Switch to text_section
12449 before calling assembly_start debug hook.
12450 * run-rtl-passes.c (run_rtl_passes): Likewise.
12453 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12455 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12456 range_int_cst_p rather than !symbolic_range_p before calling
12457 extract_range_from_multiplicative_op_1.
12459 2018-01-04 Jeff Law <law@redhat.com>
12461 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12462 redundant test in assertion.
12464 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12466 * doc/rtl.texi: Document machine_mode wrapper classes.
12468 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12470 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12471 using tree_to_uhwi.
12473 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12475 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12476 the VEC_PERM_EXPR fold to fail.
12478 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12481 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12482 to switched_sections.
12484 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12487 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12488 test for d.testing.
12490 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12493 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12494 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12496 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12499 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12500 is BLKmode and bitpos not zero or mode change is needed.
12502 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12505 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12508 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12511 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12512 instead of MULT rtx. Update all corresponding splitters.
12513 (*saddl_se): Ditto.
12514 (*ssub<modesuffix>): Ditto.
12515 (*ssubl_se): Ditto.
12516 (*cmp_sadd_di): Update split patterns.
12517 (*cmp_sadd_si): Ditto.
12518 (*cmp_sadd_sidi): Ditto.
12519 (*cmp_ssub_di): Ditto.
12520 (*cmp_ssub_si): Ditto.
12521 (*cmp_ssub_sidi): Ditto.
12522 * config/alpha/predicates.md (const23_operand): New predicate.
12523 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12524 Look for ASHIFT, not MULT inner operand.
12525 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12527 2018-01-04 Martin Liska <mliska@suse.cz>
12529 PR gcov-profile/83669
12530 * gcov.c (output_intermediate_file): Add version to intermediate
12532 * doc/gcov.texi: Document new field 'version' in intermediate
12533 file format. Fix location of '-k' option of gcov command.
12535 2018-01-04 Martin Liska <mliska@suse.cz>
12538 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12540 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12542 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12544 2018-01-03 Martin Sebor <msebor@redhat.com>
12546 PR tree-optimization/83655
12547 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12548 checking calls with invalid arguments.
12550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12552 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12553 (vectorizable_mask_load_store): Delete.
12554 (vectorizable_call): Return false for masked loads and stores.
12555 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12556 instead of gimple_assign_rhs1.
12557 (vectorizable_load): Handle IFN_MASK_LOAD.
12558 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12562 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12564 (vectorizable_mask_load_store): ...here.
12565 (vectorizable_load): ...and here.
12567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12569 * tree-vect-stmts.c (vect_build_all_ones_mask)
12570 (vect_build_zero_merge_argument): New functions, split out from...
12571 (vectorizable_load): ...here.
12573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12575 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12577 (vectorizable_mask_load_store): ...here.
12578 (vectorizable_store): ...and here.
12580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12582 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12584 (vectorizable_mask_load_store): ...here.
12586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12588 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12589 (vect_model_store_cost): Take a vec_load_store_type instead of a
12591 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12592 (vect_model_store_cost): Take a vec_load_store_type instead of a
12594 (vectorizable_mask_load_store): Update accordingly.
12595 (vectorizable_store): Likewise.
12596 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12600 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12601 IFN_MASK_LOAD calls here rather than...
12602 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12604 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12605 Alan Hayward <alan.hayward@arm.com>
12606 David Sherwood <david.sherwood@arm.com>
12608 * expmed.c (extract_bit_field_1): For vector extracts,
12609 fall back to extract_bit_field_as_subreg if vec_extract
12612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12613 Alan Hayward <alan.hayward@arm.com>
12614 David Sherwood <david.sherwood@arm.com>
12616 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12617 they are variable or constant sized.
12618 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12619 slots for constant-sized data.
12621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12622 Alan Hayward <alan.hayward@arm.com>
12623 David Sherwood <david.sherwood@arm.com>
12625 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12626 handling COND_EXPRs with boolean comparisons, try to find a better
12627 basis for the mask type than the boolean itself.
12629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12631 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12632 is calculated and how it can be overridden.
12633 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12634 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12636 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12640 Alan Hayward <alan.hayward@arm.com>
12641 David Sherwood <david.sherwood@arm.com>
12643 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12644 Remove the mode argument.
12645 (aarch64_simd_valid_immediate): Remove the mode and inverse
12647 * config/aarch64/iterators.md (bitsize): New iterator.
12648 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12649 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12650 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12651 aarch64_simd_valid_immediate.
12652 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12653 (aarch64_reg_or_bic_imm): Likewise.
12654 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12655 with an insn_type enum and msl with a modifier_type enum.
12656 Replace element_width with a scalar_mode. Change the shift
12657 to unsigned int. Add constructors for scalar_float_mode and
12658 scalar_int_mode elements.
12659 (aarch64_vect_float_const_representable_p): Delete.
12660 (aarch64_can_const_movi_rtx_p)
12661 (aarch64_simd_scalar_immediate_valid_for_move)
12662 (aarch64_simd_make_constant): Update call to
12663 aarch64_simd_valid_immediate.
12664 (aarch64_advsimd_valid_immediate_hs): New function.
12665 (aarch64_advsimd_valid_immediate): Likewise.
12666 (aarch64_simd_valid_immediate): Remove mode and inverse
12667 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12668 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12669 and aarch64_float_const_representable_p on the result.
12670 (aarch64_output_simd_mov_immediate): Remove mode argument.
12671 Update call to aarch64_simd_valid_immediate and use of
12672 simd_immediate_info.
12673 (aarch64_output_scalar_simd_mov_immediate): Update call
12676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12677 Alan Hayward <alan.hayward@arm.com>
12678 David Sherwood <david.sherwood@arm.com>
12680 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12681 (mode_nunits): Likewise CONST_MODE_NUNITS.
12682 * machmode.def (ADJUST_NUNITS): Document.
12683 * genmodes.c (mode_data::need_nunits_adj): New field.
12684 (blank_mode): Update accordingly.
12685 (adj_nunits): New variable.
12686 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12688 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12689 listed in adj_nunits.
12690 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12691 listed in adj_nunits. Don't emit case statements for such modes.
12692 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12693 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12694 nothing if adj_nunits is nonnull.
12695 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12696 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12697 (emit_mode_fbit): Update use of print_maybe_const_decl.
12698 (emit_move_size): Likewise. Treat the array as non-const
12700 (emit_mode_adjustments): Handle adj_nunits.
12702 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12704 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12705 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12706 (VECTOR_MODES): Use it.
12707 (make_vector_modes): Take the prefix as an argument.
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 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12714 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12715 for MODE_VECTOR_BOOL.
12716 * machmode.def (VECTOR_BOOL_MODE): Document.
12717 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12718 (make_vector_bool_mode): New function.
12719 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12721 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12722 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12724 * stor-layout.c (int_mode_for_mode): Likewise.
12725 * tree.c (build_vector_type_for_mode): Likewise.
12726 * varasm.c (output_constant_pool_2): Likewise.
12727 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12728 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12729 for MODE_VECTOR_BOOL.
12730 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12731 of mode class checks.
12732 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12733 instead of a list of mode class checks.
12734 (expand_vector_scalar_condition): Likewise.
12735 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12738 Alan Hayward <alan.hayward@arm.com>
12739 David Sherwood <david.sherwood@arm.com>
12741 * machmode.h (mode_size): Change from unsigned short to
12743 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12744 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12745 or if measurement_type is not polynomial.
12746 (fixed_size_mode::includes_p): Check for constant-sized modes.
12747 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12748 return a poly_uint16 rather than an unsigned short.
12749 (emit_mode_size): Change the type of mode_size from unsigned short
12750 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12751 (emit_mode_adjustments): Cope with polynomial vector sizes.
12752 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12754 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12756 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12757 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12758 * caller-save.c (setup_save_areas): Likewise.
12759 (replace_reg_with_saved_mem): Likewise.
12760 * calls.c (emit_library_call_value_1): Likewise.
12761 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12762 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12763 (gen_lowpart_for_combine): Likewise.
12764 * convert.c (convert_to_integer_1): Likewise.
12765 * cse.c (equiv_constant, cse_insn): Likewise.
12766 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12767 (cselib_subst_to_values): Likewise.
12768 * dce.c (word_dce_process_block): Likewise.
12769 * df-problems.c (df_word_lr_mark_ref): Likewise.
12770 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12771 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12772 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12773 (rtl_for_decl_location): Likewise.
12774 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12775 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12776 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12777 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12778 (expand_expr_real_1): Likewise.
12779 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12780 (pad_below): Likewise.
12781 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12782 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12783 * ira.c (get_subreg_tracking_sizes): Likewise.
12784 * ira-build.c (ira_create_allocno_objects): Likewise.
12785 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12786 (ira_sort_regnos_for_alter_reg): Likewise.
12787 * ira-costs.c (record_operand_costs): Likewise.
12788 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12789 (resolve_simple_move): Likewise.
12790 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12791 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12792 (lra_constraints): Likewise.
12793 (CONST_POOL_OK_P): Reject variable-sized modes.
12794 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12795 (add_pseudo_to_slot, lra_spill): Likewise.
12796 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12797 * optabs-query.c (get_best_extraction_insn): Likewise.
12798 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12799 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12800 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12801 * recog.c (offsettable_address_addr_space_p): Likewise.
12802 * regcprop.c (maybe_mode_change): Likewise.
12803 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12804 * regrename.c (build_def_use): Likewise.
12805 * regstat.c (dump_reg_info): Likewise.
12806 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12807 (find_reloads, find_reloads_subreg_address): Likewise.
12808 * reload1.c (eliminate_regs_1): Likewise.
12809 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12810 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12811 (simplify_binary_operation_1, simplify_subreg): Likewise.
12812 * targhooks.c (default_function_arg_padding): Likewise.
12813 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12814 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12815 (verify_gimple_assign_ternary): Likewise.
12816 * tree-inline.c (estimate_move_cost): Likewise.
12817 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12818 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12819 (get_address_cost_ainc): Likewise.
12820 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12821 (vect_supportable_dr_alignment): Likewise.
12822 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12823 (vectorizable_reduction): Likewise.
12824 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12825 (vectorizable_operation, vectorizable_load): Likewise.
12826 * tree.c (build_same_sized_truth_vector_type): Likewise.
12827 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12828 * var-tracking.c (emit_note_insn_var_location): Likewise.
12829 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12830 (ADDR_VEC_ALIGN): Likewise.
12832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12833 Alan Hayward <alan.hayward@arm.com>
12834 David Sherwood <david.sherwood@arm.com>
12836 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12838 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12839 or if measurement_type is polynomial.
12840 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12841 * combine.c (make_extraction): Likewise.
12842 * dse.c (find_shift_sequence): Likewise.
12843 * dwarf2out.c (mem_loc_descriptor): Likewise.
12844 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12845 (extract_bit_field, extract_low_bits): Likewise.
12846 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12847 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12848 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12849 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12850 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12851 * reload.c (find_reloads): Likewise.
12852 * reload1.c (alter_reg): Likewise.
12853 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12854 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12855 * tree-if-conv.c (predicate_mem_writes): Likewise.
12856 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12857 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12858 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12859 * valtrack.c (dead_debug_insert_temp): Likewise.
12860 * varasm.c (mergeable_constant_section): Likewise.
12861 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12864 Alan Hayward <alan.hayward@arm.com>
12865 David Sherwood <david.sherwood@arm.com>
12867 * expr.c (expand_assignment): Cope with polynomial mode sizes
12868 when assigning to a CONCAT.
12870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12871 Alan Hayward <alan.hayward@arm.com>
12872 David Sherwood <david.sherwood@arm.com>
12874 * machmode.h (mode_precision): Change from unsigned short to
12876 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12878 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12879 or if measurement_type is not polynomial.
12880 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12881 in which the mode is already known to be a scalar_int_mode.
12882 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12883 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12885 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12886 for GET_MODE_PRECISION.
12887 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12888 for GET_MODE_PRECISION.
12889 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12891 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12892 (expand_field_assignment, make_extraction): Likewise.
12893 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12894 (get_last_value): Likewise.
12895 * convert.c (convert_to_integer_1): Likewise.
12896 * cse.c (cse_insn): Likewise.
12897 * expr.c (expand_expr_real_1): Likewise.
12898 * lra-constraints.c (simplify_operand_subreg): Likewise.
12899 * optabs-query.c (can_atomic_load_p): Likewise.
12900 * optabs.c (expand_atomic_load): Likewise.
12901 (expand_atomic_store): Likewise.
12902 * ree.c (combine_reaching_defs): Likewise.
12903 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12904 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12905 * tree.h (type_has_mode_precision_p): Likewise.
12906 * ubsan.c (instrument_si_overflow): Likewise.
12908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12909 Alan Hayward <alan.hayward@arm.com>
12910 David Sherwood <david.sherwood@arm.com>
12912 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12913 polynomial numbers of units.
12914 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12915 (valid_vector_subparts_p): New function.
12916 (build_vector_type): Remove temporary shim and take the number
12917 of units as a poly_uint64 rather than an int.
12918 (build_opaque_vector_type): Take the number of units as a
12919 poly_uint64 rather than an int.
12920 * tree.c (build_vector_from_ctor): Handle polynomial
12921 TYPE_VECTOR_SUBPARTS.
12922 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12923 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12924 (build_vector_from_val): If the number of units is variable,
12925 use build_vec_duplicate_cst for constant operands and
12926 VEC_DUPLICATE_EXPR otherwise.
12927 (make_vector_type): Remove temporary is_constant ().
12928 (build_vector_type, build_opaque_vector_type): Take the number of
12929 units as a poly_uint64 rather than an int.
12930 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12932 * cfgexpand.c (expand_debug_expr): Likewise.
12933 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12934 (store_constructor, expand_expr_real_1): Likewise.
12935 (const_scalar_mask_from_tree): Likewise.
12936 * fold-const-call.c (fold_const_reduction): Likewise.
12937 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12938 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12939 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12940 (fold_relational_const): Likewise.
12941 (native_interpret_vector): Likewise. Change the size from an
12942 int to an unsigned int.
12943 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12944 TYPE_VECTOR_SUBPARTS.
12945 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12946 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12947 duplicating a non-constant operand into a variable-length vector.
12948 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12949 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12950 * ipa-icf.c (sem_variable::equals): Likewise.
12951 * match.pd: Likewise.
12952 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12953 * print-tree.c (print_node): Likewise.
12954 * stor-layout.c (layout_type): Likewise.
12955 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12956 * tree-cfg.c (verify_gimple_comparison): Likewise.
12957 (verify_gimple_assign_binary): Likewise.
12958 (verify_gimple_assign_ternary): Likewise.
12959 (verify_gimple_assign_single): Likewise.
12960 * tree-pretty-print.c (dump_generic_node): Likewise.
12961 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12962 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12963 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12964 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12965 (vect_shift_permute_load_chain): Likewise.
12966 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12967 (expand_vector_condition, optimize_vector_constructor): Likewise.
12968 (lower_vec_perm, get_compute_type): Likewise.
12969 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12970 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12971 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12972 (vect_recog_mask_conversion_pattern): Likewise.
12973 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12974 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12975 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12976 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12977 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12978 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12979 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12980 (supportable_widening_operation): Likewise.
12981 (supportable_narrowing_operation): Likewise.
12982 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12984 * varasm.c (output_constant): Likewise.
12986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12987 Alan Hayward <alan.hayward@arm.com>
12988 David Sherwood <david.sherwood@arm.com>
12990 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12991 so that both the length == 3 and length != 3 cases set up their
12992 own permute vectors. Add comments explaining why we know the
12993 number of elements is constant.
12994 (vect_permute_load_chain): Likewise.
12996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12997 Alan Hayward <alan.hayward@arm.com>
12998 David Sherwood <david.sherwood@arm.com>
13000 * machmode.h (mode_nunits): Change from unsigned char to
13002 (ONLY_FIXED_SIZE_MODES): New macro.
13003 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13004 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13005 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13007 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13008 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13009 or if measurement_type is not polynomial.
13010 * genmodes.c (ZERO_COEFFS): New macro.
13011 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13013 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13014 Use ZERO_COEFFS when emitting initializers.
13015 * data-streamer.h (bp_pack_poly_value): New function.
13016 (bp_unpack_poly_value): Likewise.
13017 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13018 for GET_MODE_NUNITS.
13019 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13020 for GET_MODE_NUNITS.
13021 * tree.c (make_vector_type): Remove temporary shim and make
13022 the real function take the number of units as a poly_uint64
13023 rather than an int.
13024 (build_vector_type_for_mode): Handle polynomial nunits.
13025 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13026 * emit-rtl.c (const_vec_series_p_1): Likewise.
13027 (gen_rtx_CONST_VECTOR): Likewise.
13028 * fold-const.c (test_vec_duplicate_folding): Likewise.
13029 * genrecog.c (validate_pattern): Likewise.
13030 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13031 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13032 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13033 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13034 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13035 * rtlanal.c (subreg_get_info): Likewise.
13036 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13037 (vect_grouped_load_supported): Likewise.
13038 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13039 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13040 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13041 (simplify_const_unary_operation, simplify_binary_operation_1)
13042 (simplify_const_binary_operation, simplify_ternary_operation)
13043 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13044 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13045 instead of CONST_VECTOR_NUNITS.
13046 * varasm.c (output_constant_pool_2): Likewise.
13047 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13048 explicit-encoded elements in the XVEC for variable-length vectors.
13050 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13052 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13055 Alan Hayward <alan.hayward@arm.com>
13056 David Sherwood <david.sherwood@arm.com>
13058 * coretypes.h (fixed_size_mode): Declare.
13059 (fixed_size_mode_pod): New typedef.
13060 * builtins.h (target_builtins::x_apply_args_mode)
13061 (target_builtins::x_apply_result_mode): Change type to
13062 fixed_size_mode_pod.
13063 * builtins.c (apply_args_size, apply_result_size, result_vector)
13064 (expand_builtin_apply_args_1, expand_builtin_apply)
13065 (expand_builtin_return): Update accordingly.
13067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13069 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13070 * cselib.c (cselib_hash_rtx): Likewise.
13071 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13072 CONST_VECTOR encoding.
13074 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13075 Jeff Law <law@redhat.com>
13078 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13079 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13080 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13081 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13084 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13085 explicitly probe *sp in a noreturn function if there were any callee
13086 register saves or frame pointer is needed.
13088 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13091 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13092 BLKmode for ternary, binary or unary expressions.
13095 * var-tracking.c (delete_vta_debug_insn): New inline function.
13096 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13097 insns from get_insns () to NULL instead of each bb separately.
13098 Use delete_vta_debug_insn. No longer static.
13099 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13100 delete_vta_debug_insns callers.
13101 * rtl.h (delete_vta_debug_insns): Declare.
13102 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13103 instead of variable_tracking_main.
13105 2018-01-03 Martin Sebor <msebor@redhat.com>
13107 PR tree-optimization/83603
13108 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13109 arguments past the endof the argument list in functions declared
13110 without a prototype.
13111 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13112 Avoid checking when arguments are null.
13114 2018-01-03 Martin Sebor <msebor@redhat.com>
13117 * doc/extend.texi (attribute const): Fix a typo.
13118 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13119 issuing -Wsuggest-attribute for void functions.
13121 2018-01-03 Martin Sebor <msebor@redhat.com>
13123 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13124 offset_int::from instead of wide_int::to_shwi.
13125 (maybe_diag_overlap): Remove assertion.
13126 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13127 * gimple-ssa-sprintf.c (format_directive): Same.
13128 (parse_directive): Same.
13129 (sprintf_dom_walker::compute_format_length): Same.
13130 (try_substitute_return_value): Same.
13132 2018-01-03 Jeff Law <law@redhat.com>
13134 PR middle-end/83654
13135 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13136 non-constant residual for zero at runtime and avoid probing in
13137 that case. Reorganize code for trailing problem to mirror handling
13140 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13142 PR tree-optimization/83501
13143 * tree-ssa-strlen.c (get_string_cst): New.
13144 (handle_char_store): Call get_string_cst.
13146 2018-01-03 Martin Liska <mliska@suse.cz>
13148 PR tree-optimization/83593
13149 * tree-ssa-strlen.c: Include tree-cfg.h.
13150 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13151 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13152 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13154 (strlen_dom_walker::before_dom_children): Call
13155 gimple_purge_dead_eh_edges. Dump tranformation with details
13157 (strlen_dom_walker::before_dom_children): Update call by adding
13158 new argument cleanup_eh.
13159 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13161 2018-01-03 Martin Liska <mliska@suse.cz>
13164 * cif-code.def (VARIADIC_THUNK): New enum value.
13165 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13168 2018-01-03 Jan Beulich <jbeulich@suse.com>
13170 * sse.md (mov<mode>_internal): Tighten condition for when to use
13171 vmovdqu<ssescalarsize> for TI and OI modes.
13173 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13175 Update copyright years.
13177 2018-01-03 Martin Liska <mliska@suse.cz>
13180 * ipa-visibility.c (function_and_variable_visibility): Skip
13181 functions with noipa attribure.
13183 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13185 * gcc.c (process_command): Update copyright notice dates.
13186 * gcov-dump.c (print_version): Ditto.
13187 * gcov.c (print_version): Ditto.
13188 * gcov-tool.c (print_version): Ditto.
13189 * gengtype.c (create_file): Ditto.
13190 * doc/cpp.texi: Bump @copying's copyright year.
13191 * doc/cppinternals.texi: Ditto.
13192 * doc/gcc.texi: Ditto.
13193 * doc/gccint.texi: Ditto.
13194 * doc/gcov.texi: Ditto.
13195 * doc/install.texi: Ditto.
13196 * doc/invoke.texi: Ditto.
13198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13200 * vector-builder.h (vector_builder::m_full_nelts): Change from
13201 unsigned int to poly_uint64.
13202 (vector_builder::full_nelts): Update prototype accordingly.
13203 (vector_builder::new_vector): Likewise.
13204 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13205 (vector_builder::operator ==): Likewise.
13206 (vector_builder::finalize): Likewise.
13207 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13208 Take the number of elements as a poly_uint64 rather than an
13210 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13211 from unsigned int to poly_uint64.
13212 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13213 (vec_perm_indices::new_vector): Likewise.
13214 (vec_perm_indices::length): Likewise.
13215 (vec_perm_indices::nelts_per_input): Likewise.
13216 (vec_perm_indices::input_nelts): Likewise.
13217 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13218 number of elements per input as a poly_uint64 rather than an
13219 unsigned int. Use the original encoding for variable-length
13220 vectors, rather than clamping each individual element.
13221 For the second and subsequent elements in each pattern,
13222 clamp the step and base before clamping their sum.
13223 (vec_perm_indices::series_p): Handle polynomial element counts.
13224 (vec_perm_indices::all_in_range_p): Likewise.
13225 (vec_perm_indices_to_tree): Likewise.
13226 (vec_perm_indices_to_rtx): Likewise.
13227 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13228 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13229 (tree_vector_builder::new_binary_operation): Handle polynomial
13230 element counts. Return false if we need to know the number
13231 of elements at compile time.
13232 * fold-const.c (fold_vec_perm): Punt if the number of elements
13233 isn't known at compile time.
13235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13237 * vec-perm-indices.h (vec_perm_builder): Change element type
13238 from HOST_WIDE_INT to poly_int64.
13239 (vec_perm_indices::element_type): Update accordingly.
13240 (vec_perm_indices::clamp): Handle polynomial element_types.
13241 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13242 (vec_perm_indices::all_in_range_p): Likewise.
13243 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13245 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13246 polynomial vec_perm_indices element types.
13247 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13248 * fold-const.c (fold_vec_perm): Likewise.
13249 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13250 * tree-vect-generic.c (lower_vec_perm): Likewise.
13251 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13252 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13253 element type to HOST_WIDE_INT.
13255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13256 Alan Hayward <alan.hayward@arm.com>
13257 David Sherwood <david.sherwood@arm.com>
13259 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13260 rather than an int. Use plus_constant.
13261 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13262 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13265 Alan Hayward <alan.hayward@arm.com>
13266 David Sherwood <david.sherwood@arm.com>
13268 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13269 a HOST_WIDE_INT to a poly_int64.
13271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13272 Alan Hayward <alan.hayward@arm.com>
13273 David Sherwood <david.sherwood@arm.com>
13275 * calls.c (load_register_parameters): Cope with polynomial
13276 mode sizes. Require a constant size for BLKmode parameters
13277 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13278 forces a parameter to be padded at the lsb end in order to
13279 fill a complete number of words, require the parameter size
13280 to be ordered wrt UNITS_PER_WORD.
13282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13283 Alan Hayward <alan.hayward@arm.com>
13284 David Sherwood <david.sherwood@arm.com>
13286 * reload1.c (spill_stack_slot_width): Change element type
13287 from unsigned int to poly_uint64_pod.
13288 (alter_reg): Treat mode sizes as polynomial.
13290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13291 Alan Hayward <alan.hayward@arm.com>
13292 David Sherwood <david.sherwood@arm.com>
13294 * reload.c (complex_word_subreg_p): New function.
13295 (reload_inner_reg_of_subreg, push_reload): Use it.
13297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13298 Alan Hayward <alan.hayward@arm.com>
13299 David Sherwood <david.sherwood@arm.com>
13301 * lra-constraints.c (process_alt_operands): Reject matched
13302 operands whose sizes aren't ordered.
13303 (match_reload): Refer to this check here.
13305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13306 Alan Hayward <alan.hayward@arm.com>
13307 David Sherwood <david.sherwood@arm.com>
13309 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13310 that the mode size is in the set {1, 2, 4, 8, 16}.
13312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13313 Alan Hayward <alan.hayward@arm.com>
13314 David Sherwood <david.sherwood@arm.com>
13316 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13317 Use plus_constant instead of gen_rtx_PLUS.
13319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13320 Alan Hayward <alan.hayward@arm.com>
13321 David Sherwood <david.sherwood@arm.com>
13323 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13324 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13325 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13326 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13327 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13328 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13329 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13330 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13331 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13332 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13334 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13335 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13336 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13337 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13338 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13339 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13340 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13341 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13342 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13343 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13345 * expr.c (emit_move_resolve_push): Treat the input and result
13346 of PUSH_ROUNDING as a poly_int64.
13347 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13348 (emit_push_insn): Likewise.
13349 * lra-eliminations.c (mark_not_eliminable): Likewise.
13350 * recog.c (push_operand): Likewise.
13351 * reload1.c (elimination_effects): Likewise.
13352 * rtlanal.c (nonzero_bits1): Likewise.
13353 * calls.c (store_one_arg): Likewise. Require the padding to be
13354 known at compile time.
13356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13357 Alan Hayward <alan.hayward@arm.com>
13358 David Sherwood <david.sherwood@arm.com>
13360 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13361 Use plus_constant instead of gen_rtx_PLUS.
13363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13364 Alan Hayward <alan.hayward@arm.com>
13365 David Sherwood <david.sherwood@arm.com>
13367 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13368 rather than an int.
13370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13371 Alan Hayward <alan.hayward@arm.com>
13372 David Sherwood <david.sherwood@arm.com>
13374 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13375 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13376 via stack temporaries. Treat the mode size as polynomial too.
13378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13379 Alan Hayward <alan.hayward@arm.com>
13380 David Sherwood <david.sherwood@arm.com>
13382 * expr.c (expand_expr_real_2): When handling conversions involving
13383 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13384 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13385 as a poly_uint64 too.
13387 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13388 Alan Hayward <alan.hayward@arm.com>
13389 David Sherwood <david.sherwood@arm.com>
13391 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13394 Alan Hayward <alan.hayward@arm.com>
13395 David Sherwood <david.sherwood@arm.com>
13397 * combine.c (can_change_dest_mode): Handle polynomial
13398 REGMODE_NATURAL_SIZE.
13399 * expmed.c (store_bit_field_1): Likewise.
13400 * expr.c (store_constructor): Likewise.
13401 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13402 and polynomial REGMODE_NATURAL_SIZE.
13403 (gen_lowpart_common): Likewise.
13404 * reginfo.c (record_subregs_of_mode): Likewise.
13405 * rtlanal.c (read_modify_subreg_p): Likewise.
13407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13408 Alan Hayward <alan.hayward@arm.com>
13409 David Sherwood <david.sherwood@arm.com>
13411 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13412 numbers of elements.
13414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13415 Alan Hayward <alan.hayward@arm.com>
13416 David Sherwood <david.sherwood@arm.com>
13418 * match.pd: Cope with polynomial numbers of vector elements.
13420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13421 Alan Hayward <alan.hayward@arm.com>
13422 David Sherwood <david.sherwood@arm.com>
13424 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13425 in a POINTER_PLUS_EXPR.
13427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13428 Alan Hayward <alan.hayward@arm.com>
13429 David Sherwood <david.sherwood@arm.com>
13431 * omp-simd-clone.c (simd_clone_subparts): New function.
13432 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13433 (ipa_simd_modify_function_body): Likewise.
13435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13436 Alan Hayward <alan.hayward@arm.com>
13437 David Sherwood <david.sherwood@arm.com>
13439 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13440 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13441 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13442 (expand_vector_condition, vector_element): Likewise.
13443 (subparts_gt): New function.
13444 (get_compute_type): Use subparts_gt.
13445 (count_type_subparts): Delete.
13446 (expand_vector_operations_1): Use subparts_gt instead of
13447 count_type_subparts.
13449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13450 Alan Hayward <alan.hayward@arm.com>
13451 David Sherwood <david.sherwood@arm.com>
13453 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13454 (vect_compile_time_alias): ...this new function. Do the calculation
13455 on poly_ints rather than trees.
13456 (vect_prune_runtime_alias_test_list): Update call accordingly.
13458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13459 Alan Hayward <alan.hayward@arm.com>
13460 David Sherwood <david.sherwood@arm.com>
13462 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13464 (vect_schedule_slp_instance): Likewise.
13466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13467 Alan Hayward <alan.hayward@arm.com>
13468 David Sherwood <david.sherwood@arm.com>
13470 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13471 constant and extern definitions for variable-length vectors.
13472 (vect_get_constant_vectors): Note that the number of units
13473 is known to be constant.
13475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13476 Alan Hayward <alan.hayward@arm.com>
13477 David Sherwood <david.sherwood@arm.com>
13479 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13480 of units as polynomial. Choose between WIDE and NARROW based
13483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13484 Alan Hayward <alan.hayward@arm.com>
13485 David Sherwood <david.sherwood@arm.com>
13487 * tree-vect-stmts.c (simd_clone_subparts): New function.
13488 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13495 vectors as polynomial. Use build_index_vector for
13496 IFN_GOMP_SIMD_LANE.
13498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13499 Alan Hayward <alan.hayward@arm.com>
13500 David Sherwood <david.sherwood@arm.com>
13502 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13503 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13504 for variable-length vectors.
13505 (vectorizable_mask_load_store): Treat the number of units as
13506 polynomial, asserting that it is constant if the condition has
13507 already been enforced.
13508 (vectorizable_store, vectorizable_load): Likewise.
13510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13511 Alan Hayward <alan.hayward@arm.com>
13512 David Sherwood <david.sherwood@arm.com>
13514 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13515 of units as polynomial. Punt if we can't tell at compile time
13516 which vector contains the final result.
13518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13519 Alan Hayward <alan.hayward@arm.com>
13520 David Sherwood <david.sherwood@arm.com>
13522 * tree-vect-loop.c (vectorizable_induction): Treat the number
13523 of units as polynomial. Punt on SLP inductions. Use an integer
13524 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13525 cast of such a series for variable-length floating-point
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.h (build_index_vector): Declare.
13533 * tree.c (build_index_vector): New function.
13534 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13535 of units as polynomial, forcibly converting it to a constant if
13536 vectorizable_reduction has already enforced the condition.
13537 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13538 to create a {1,2,3,...} vector.
13539 (vectorizable_reduction): Treat the number of units as polynomial.
13540 Choose vectype_in based on the largest scalar element size rather
13541 than the smallest number of units. Enforce the restrictions
13544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13545 Alan Hayward <alan.hayward@arm.com>
13546 David Sherwood <david.sherwood@arm.com>
13548 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13549 number of units as polynomial.
13551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13552 Alan Hayward <alan.hayward@arm.com>
13553 David Sherwood <david.sherwood@arm.com>
13555 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13556 * target.def (autovectorize_vector_sizes): Return the vector sizes
13557 by pointer, using vector_sizes rather than a bitmask.
13558 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13559 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13560 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13562 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13563 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13564 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13565 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13566 * omp-general.c (omp_max_vf): Likewise.
13567 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13568 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13569 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13570 * tree-vect-slp.c (vect_slp_bb): Likewise.
13571 * doc/tm.texi: Regenerate.
13572 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13574 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13575 the vector size as a poly_uint64 rather than an unsigned int.
13576 (current_vector_size): Change from an unsigned int to a poly_uint64.
13577 (get_vectype_for_scalar_type): Update accordingly.
13578 * tree.h (build_truth_vector_type): Take the size and number of
13579 units as a poly_uint64 rather than an unsigned int.
13580 (build_vector_type): Add a temporary overload that takes
13581 the number of units as a poly_uint64 rather than an unsigned int.
13582 * tree.c (make_vector_type): Likewise.
13583 (build_truth_vector_type): Take the number of units as a poly_uint64
13584 rather than an unsigned int.
13586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13587 Alan Hayward <alan.hayward@arm.com>
13588 David Sherwood <david.sherwood@arm.com>
13590 * target.def (get_mask_mode): Take the number of units and length
13591 as poly_uint64s rather than unsigned ints.
13592 * targhooks.h (default_get_mask_mode): Update accordingly.
13593 * targhooks.c (default_get_mask_mode): Likewise.
13594 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13595 * doc/tm.texi: Regenerate.
13597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13598 Alan Hayward <alan.hayward@arm.com>
13599 David Sherwood <david.sherwood@arm.com>
13601 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13602 * omp-general.c (omp_max_vf): Likewise.
13603 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13604 (expand_omp_simd): Handle polynomial safelen.
13605 * omp-low.c (omplow_simd_context): Add a default constructor.
13606 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13607 (lower_rec_simd_input_clauses): Update accordingly.
13608 (lower_rec_input_clauses): Likewise.
13610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13611 Alan Hayward <alan.hayward@arm.com>
13612 David Sherwood <david.sherwood@arm.com>
13614 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13615 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13616 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13617 (vect_analyze_slp_cost): Likewise.
13618 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13619 (vect_model_load_cost): Likewise.
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 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13626 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13627 from an unsigned int * to a poly_uint64_pod *.
13628 (calculate_unrolling_factor): New function.
13629 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13632 Alan Hayward <alan.hayward@arm.com>
13633 David Sherwood <david.sherwood@arm.com>
13635 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13636 from an unsigned int to a poly_uint64.
13637 (_loop_vec_info::slp_unrolling_factor): Likewise.
13638 (_loop_vec_info::vectorization_factor): Change from an int
13640 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13641 (vect_get_num_vectors): New function.
13642 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13643 (vect_get_num_copies): Use vect_get_num_vectors.
13644 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13645 to an unsigned int *.
13646 (vect_analyze_data_refs): Change min_vf from an int * to a
13648 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13649 than an unsigned HOST_WIDE_INT.
13650 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13651 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13652 to an unsigned int *.
13653 (vect_analyze_data_ref_dependences): Likewise.
13654 (vect_compute_data_ref_alignment): Handle polynomial vf.
13655 (vect_enhance_data_refs_alignment): Likewise.
13656 (vect_prune_runtime_alias_test_list): Likewise.
13657 (vect_shift_permute_load_chain): Likewise.
13658 (vect_supportable_dr_alignment): Likewise.
13659 (dependence_distance_ge_vf): Take the vectorization factor as a
13660 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13661 (vect_analyze_data_refs): Change min_vf from an int * to a
13663 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13664 vfm1 as a poly_uint64 rather than an int. Make the same change
13665 for the returned bound_scalar.
13666 (vect_gen_vector_loop_niters): Handle polynomial vf.
13667 (vect_do_peeling): Likewise. Update call to
13668 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13669 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13671 * tree-vect-loop.c (vect_determine_vectorization_factor)
13672 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13673 (vect_get_known_peeling_cost): Likewise.
13674 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13675 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13676 (vect_transform_loop): Likewise. Use the lowest possible VF when
13677 updating the upper bounds of the loop.
13678 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13679 rather than an int.
13680 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13681 polynomial unroll factors.
13682 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13683 (vect_make_slp_decision): Likewise.
13684 (vect_supported_load_permutation_p): Likewise, and polynomial
13686 (vect_analyze_slp_cost): Handle polynomial vf.
13687 (vect_slp_analyze_node_operations): Likewise.
13688 (vect_slp_analyze_bb_1): Likewise.
13689 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13690 than an unsigned HOST_WIDE_INT.
13691 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13692 (vectorizable_load): Handle polynomial vf.
13693 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13695 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13698 Alan Hayward <alan.hayward@arm.com>
13699 David Sherwood <david.sherwood@arm.com>
13701 * match.pd: Handle bit operations involving three constants
13702 and try to fold one pair.
13704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13706 * tree-vect-loop-manip.c: Include gimple-fold.h.
13707 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13708 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13709 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13710 Add a path that uses a step of VF instead of 1, but disable it
13712 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13713 and niters_no_overflow parameters. Update calls to
13714 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13715 Create a new SSA name if the latter choses to use a ste other
13716 than zero, and return it via niters_vector_mult_vf_var.
13717 * tree-vect-loop.c (vect_transform_loop): Update calls to
13718 vect_do_peeling, vect_gen_vector_loop_niters and
13719 slpeel_make_loop_iterate_ntimes.
13720 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13721 (vect_gen_vector_loop_niters): Update declarations after above changes.
13723 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13725 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13726 128-bit round to integer instructions.
13727 (ceil<mode>2): Likewise.
13728 (btrunc<mode>2): Likewise.
13729 (round<mode>2): Likewise.
13731 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13733 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13734 unaligned VSX load/store on P8/P9.
13735 (expand_block_clear): Allow the use of unaligned VSX
13736 load/store on P8/P9.
13738 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13740 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13742 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13743 swap associated with both a load and a store.
13745 2018-01-02 Andrew Waterman <andrew@sifive.com>
13747 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13748 * config/riscv/riscv.md (clear_cache): Use it.
13750 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13752 * web.c: Remove out-of-date comment.
13754 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13756 * expr.c (fixup_args_size_notes): Check that any existing
13757 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13758 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13759 (emit_single_push_insn): ...here.
13761 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13763 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13764 (const_vector_encoded_nelts): New function.
13765 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13766 (const_vector_int_elt, const_vector_elt): Declare.
13767 * emit-rtl.c (const_vector_int_elt_1): New function.
13768 (const_vector_elt): Likewise.
13769 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13770 of CONST_VECTOR_ELT.
13772 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13774 * expr.c: Include rtx-vector-builder.h.
13775 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13776 directly on the tree encoding.
13777 (const_vector_from_tree): Likewise.
13778 * optabs.c: Include rtx-vector-builder.h.
13779 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13780 sequence of "u" values.
13781 * vec-perm-indices.c: Include rtx-vector-builder.h.
13782 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13783 directly on the vec_perm_indices encoding.
13785 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13787 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13788 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13789 * rtx-vector-builder.h: New file.
13790 * rtx-vector-builder.c: Likewise.
13791 * rtl.h (rtx_def::u2): Add a const_vector field.
13792 (CONST_VECTOR_NPATTERNS): New macro.
13793 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13794 (CONST_VECTOR_DUPLICATE_P): Likewise.
13795 (CONST_VECTOR_STEPPED_P): Likewise.
13796 (CONST_VECTOR_ENCODED_ELT): Likewise.
13797 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13798 (unwrap_const_vec_duplicate): Likewise.
13799 (const_vec_series_p): Check for a non-duplicated vector encoding.
13800 Say that the function only returns true for integer vectors.
13801 * emit-rtl.c: Include rtx-vector-builder.h.
13802 (gen_const_vec_duplicate_1): Delete.
13803 (gen_const_vector): Call gen_const_vec_duplicate instead of
13804 gen_const_vec_duplicate_1.
13805 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13806 (gen_const_vec_duplicate): Use rtx_vector_builder.
13807 (gen_const_vec_series): Likewise.
13808 (gen_rtx_CONST_VECTOR): Likewise.
13809 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13810 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13811 Build a new vector rather than modifying a CONST_VECTOR in-place.
13812 (handle_special_swappables): Update call accordingly.
13813 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13814 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13815 Build a new vector rather than modifying a CONST_VECTOR in-place.
13816 (handle_special_swappables): Update call accordingly.
13818 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13820 * simplify-rtx.c (simplify_const_binary_operation): Use
13821 CONST_VECTOR_ELT instead of XVECEXP.
13823 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13825 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13826 the selector elements to be different from the data elements
13827 if the selector is a VECTOR_CST.
13828 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13829 ssizetype for the selector.
13831 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13833 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13834 before testing each element individually.
13835 * tree-vect-generic.c (lower_vec_perm): Likewise.
13837 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13839 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13840 * selftest-run-tests.c (selftest::run_tests): Call it.
13841 * vector-builder.h (vector_builder::operator ==): New function.
13842 (vector_builder::operator !=): Likewise.
13843 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13844 (vec_perm_indices::all_from_input_p): New function.
13845 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13846 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13847 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13848 instead of reading the VECTOR_CST directly. Detect whether both
13849 vector inputs are the same before constructing the vec_perm_indices,
13850 and update the number of inputs argument accordingly. Use the
13851 utility functions added above. Only construct sel2 if we need to.
13853 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13855 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13856 the broadcast of the low byte.
13857 (expand_mult_highpart): Use an explicit encoding for the permutes.
13858 * optabs-query.c (can_mult_highpart_p): Likewise.
13859 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13860 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13861 (vectorizable_bswap): Likewise.
13862 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13863 explicit encoding for the power-of-2 permutes.
13864 (vect_permute_store_chain): Likewise.
13865 (vect_grouped_load_supported): Likewise.
13866 (vect_permute_load_chain): Likewise.
13868 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13870 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13871 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13872 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13873 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13874 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13875 (vect_gen_perm_mask_any): Likewise.
13877 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13879 * int-vector-builder.h: New file.
13880 * vec-perm-indices.h: Include int-vector-builder.h.
13881 (vec_perm_indices): Redefine as an int_vector_builder.
13882 (auto_vec_perm_indices): Delete.
13883 (vec_perm_builder): Redefine as a stand-alone class.
13884 (vec_perm_indices::vec_perm_indices): New function.
13885 (vec_perm_indices::clamp): Likewise.
13886 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13887 (vec_perm_indices::new_vector): New function.
13888 (vec_perm_indices::new_expanded_vector): Update for new
13889 vec_perm_indices class.
13890 (vec_perm_indices::rotate_inputs): New function.
13891 (vec_perm_indices::all_in_range_p): Operate directly on the
13892 encoded form, without computing elided elements.
13893 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13894 encoding. Update for new vec_perm_indices class.
13895 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13896 the given vec_perm_builder.
13897 (expand_vec_perm_var): Update vec_perm_builder constructor.
13898 (expand_mult_highpart): Use vec_perm_builder instead of
13899 auto_vec_perm_indices.
13900 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13901 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13902 or double series encoding as appropriate.
13903 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13904 vec_perm_indices instead of auto_vec_perm_indices.
13905 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13906 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13907 (vect_permute_store_chain): Likewise.
13908 (vect_grouped_load_supported): Likewise.
13909 (vect_permute_load_chain): Likewise.
13910 (vect_shift_permute_load_chain): Likewise.
13911 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13912 (vect_transform_slp_perm_load): Likewise.
13913 (vect_schedule_slp_instance): Likewise.
13914 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13915 (vectorizable_mask_load_store): Likewise.
13916 (vectorizable_bswap): Likewise.
13917 (vectorizable_store): Likewise.
13918 (vectorizable_load): Likewise.
13919 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13920 vec_perm_indices instead of auto_vec_perm_indices. Use
13921 tree_to_vec_perm_builder to read the vector from a tree.
13922 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13923 vec_perm_builder instead of a vec_perm_indices.
13924 (have_whole_vector_shift): Use vec_perm_builder and
13925 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13926 truncation to calc_vec_perm_mask_for_shift.
13927 (vect_create_epilog_for_reduction): Likewise.
13928 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13929 from auto_vec_perm_indices to vec_perm_indices.
13930 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13931 instead of changing individual elements.
13932 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13933 the vector in d.perm.
13934 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13935 from auto_vec_perm_indices to vec_perm_indices.
13936 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13937 instead of changing individual elements.
13938 (arm_vectorize_vec_perm_const): Use new_vector to install
13939 the vector in d.perm.
13940 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13941 Update vec_perm_builder constructor.
13942 (rs6000_expand_interleave): Likewise.
13943 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13944 (rs6000_expand_interleave): Likewise.
13946 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13948 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13949 to qimode could truncate the indices.
13950 * optabs.c (expand_vec_perm_var): Likewise.
13952 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13954 * Makefile.in (OBJS): Add vec-perm-indices.o.
13955 * vec-perm-indices.h: New file.
13956 * vec-perm-indices.c: Likewise.
13957 * target.h (vec_perm_indices): Replace with a forward class
13959 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13960 * optabs.h: Include vec-perm-indices.h.
13961 (expand_vec_perm): Delete.
13962 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13963 (expand_vec_perm_const): Declare.
13964 * target.def (vec_perm_const_ok): Replace with...
13965 (vec_perm_const): ...this new hook.
13966 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13967 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13968 * doc/tm.texi: Regenerate.
13969 * optabs.def (vec_perm_const): Delete.
13970 * doc/md.texi (vec_perm_const): Likewise.
13971 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13972 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13973 expand_vec_perm for constant permutation vectors. Assert that
13974 the mode of variable permutation vectors is the integer equivalent
13975 of the mode that is being permuted.
13976 * optabs-query.h (selector_fits_mode_p): Declare.
13977 * optabs-query.c: Include vec-perm-indices.h.
13978 (selector_fits_mode_p): New function.
13979 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13980 is defined, instead of checking whether the vec_perm_const_optab
13981 exists. Use targetm.vectorize.vec_perm_const instead of
13982 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13983 fit in the vector mode before using a variable permute.
13984 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13985 vec_perm_indices instead of an rtx.
13986 (expand_vec_perm): Replace with...
13987 (expand_vec_perm_const): ...this new function. Take the selector
13988 as a vec_perm_indices rather than an rtx. Also take the mode of
13989 the selector. Update call to shift_amt_for_vec_perm_mask.
13990 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13991 Use vec_perm_indices::new_expanded_vector to expand the original
13992 selector into bytes. Check whether the indices fit in the vector
13993 mode before using a variable permute.
13994 (expand_vec_perm_var): Make global.
13995 (expand_mult_highpart): Use expand_vec_perm_const.
13996 * fold-const.c: Includes vec-perm-indices.h.
13997 * tree-ssa-forwprop.c: Likewise.
13998 * tree-vect-data-refs.c: Likewise.
13999 * tree-vect-generic.c: Likewise.
14000 * tree-vect-loop.c: Likewise.
14001 * tree-vect-slp.c: Likewise.
14002 * tree-vect-stmts.c: Likewise.
14003 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14005 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14006 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14007 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14008 (aarch64_vectorize_vec_perm_const): ...this new function.
14009 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14010 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14011 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14012 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14013 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14014 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14015 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14017 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14018 check for NEON modes.
14019 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14020 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14021 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14022 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14024 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14025 the old VEC_PERM_CONST conditions.
14026 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14027 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14028 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14029 (ia64_vectorize_vec_perm_const_ok): Merge into...
14030 (ia64_vectorize_vec_perm_const): ...this new function.
14031 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14032 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14033 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14034 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14035 * config/mips/mips.c (mips_expand_vec_perm_const)
14036 (mips_vectorize_vec_perm_const_ok): Merge into...
14037 (mips_vectorize_vec_perm_const): ...this new function.
14038 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14039 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14040 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14041 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14042 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14043 (rs6000_expand_vec_perm_const): Delete.
14044 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14046 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14047 (altivec_expand_vec_perm_const_le): Take each operand individually.
14048 Operate on constant selectors rather than rtxes.
14049 (altivec_expand_vec_perm_const): Likewise. Update call to
14050 altivec_expand_vec_perm_const_le.
14051 (rs6000_expand_vec_perm_const): Delete.
14052 (rs6000_vectorize_vec_perm_const_ok): Delete.
14053 (rs6000_vectorize_vec_perm_const): New function.
14054 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14055 an element count and rtx array.
14056 (rs6000_expand_extract_even): Update call accordingly.
14057 (rs6000_expand_interleave): Likewise.
14058 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14059 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14060 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14061 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14062 (rs6000_expand_vec_perm_const): Delete.
14063 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14064 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14065 (altivec_expand_vec_perm_const_le): Take each operand individually.
14066 Operate on constant selectors rather than rtxes.
14067 (altivec_expand_vec_perm_const): Likewise. Update call to
14068 altivec_expand_vec_perm_const_le.
14069 (rs6000_expand_vec_perm_const): Delete.
14070 (rs6000_vectorize_vec_perm_const_ok): Delete.
14071 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14072 reference to the SPE evmerge intructions.
14073 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14074 an element count and rtx array.
14075 (rs6000_expand_extract_even): Update call accordingly.
14076 (rs6000_expand_interleave): Likewise.
14077 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14078 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14080 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14082 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14084 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14085 vector mode and that that mode matches the mode of the data
14087 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14088 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14089 directly using expand_vec_perm_1 when forcing selectors into
14091 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14093 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14095 * optabs-query.h (can_vec_perm_p): Delete.
14096 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14097 * optabs-query.c (can_vec_perm_p): Split into...
14098 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14099 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14100 particular selector is valid.
14101 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14102 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14103 (vect_grouped_load_supported): Likewise.
14104 (vect_shift_permute_load_chain): Likewise.
14105 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14106 (vect_transform_slp_perm_load): Likewise.
14107 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14108 (vectorizable_bswap): Likewise.
14109 (vect_gen_perm_mask_checked): Likewise.
14110 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14111 implementations of variable permutation vectors into account
14112 when deciding which selector to use.
14113 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14114 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14115 with a false third argument.
14116 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14117 to test whether the constant selector is valid and can_vec_perm_var_p
14118 to test whether a variable selector is valid.
14120 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14122 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14123 * optabs-query.c (can_vec_perm_p): Likewise.
14124 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14125 instead of vec_perm_indices.
14126 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14127 (vect_gen_perm_mask_checked): Likewise,
14128 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14129 (vect_gen_perm_mask_checked): Likewise,
14131 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14133 * optabs-query.h (qimode_for_vec_perm): Declare.
14134 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14135 (qimode_for_vec_perm): ...this new function.
14136 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14138 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14140 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14141 does not have a conditional at the top.
14143 2018-01-02 Richard Biener <rguenther@suse.de>
14145 * ipa-inline.c (big_speedup_p): Fix expression.
14147 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14150 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14153 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14157 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14158 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14159 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14160 cond_taken_branch_cost 3->4.
14162 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14164 PR tree-optimization/83581
14165 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14166 TODO_cleanup_cfg if any changes have been made.
14168 PR middle-end/83608
14169 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14170 convert_modes if target mode has the right side, but different mode
14173 PR middle-end/83609
14174 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14175 last argument when extracting from CONCAT. If either from_real or
14176 from_imag is NULL, use expansion through memory. If result is not
14177 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14178 the parts directly to inner mode, if even that fails, use expansion
14181 PR middle-end/83623
14182 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14183 check for bswap in mode rather than HImode and use that in expand_unop
14186 Copyright (C) 2018 Free Software Foundation, Inc.
14188 Copying and distribution of this file, with or without modification,
14189 are permitted in any medium without royalty provided the copyright
14190 notice and this notice are preserved.