1 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
4 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5 machine->attr_no_prologue_p.
6 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
7 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
8 (nds32_expand_epilogue): Consider attr_naked_p.
9 (nds32_expand_epilogue_v3pop): Likewise.
10 (nds32_can_use_return_insn): Likewise.
11 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
12 attr_no_prologue_p fields.
13 * config/nds32/nds32.opt (mret-in-naked-func): New option.
15 2018-05-27 Jakub Jelinek <jakub@redhat.com>
18 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
21 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
23 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
25 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
27 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
29 (*<floatsuffix>floatv2div2sf2): Rename to ...
30 (*float<floatunssuffix>v2div2sf2): ... this.
31 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
32 (float<floatunssuffix>v2div2sf2_mask): ... this.
33 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
34 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
35 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
37 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
39 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
41 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
43 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
45 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
47 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
48 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
49 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
50 gen_ufix_truncv8dfv8si2.
51 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
52 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
53 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
54 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
55 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
56 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
57 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
58 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
60 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
64 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
66 2018-05-25 Jim Wilson <jimw@sifive.com>
68 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
69 * config/riscv/riscv.c (struct machine_function): Add
70 interrupt_handler_p and attribute_checked_p fields.
71 (riscv_attribute_table): Add interrupt.
72 (riscv_interrupt_type_p): New.
73 (riscv_save_reg_p): Save extra regs for interrupt handler.
74 (riscv_use_save_libcall): Return false for interrupt handler.
75 (riscv_first_stack_step): Add forward declaration.
76 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
77 for interrupt handler with large frame. Use it for saved reg list.
78 (riscv_expand_prologue): Move flag_stack_usage_info support to
79 eliminate duplication.
80 (riscv_expand_epilogue): Generate mret for interrupt handler.
81 (riscv_epilogue_uses): New.
82 (riscv_can_use_return_insn): Return false for interrupt handler.
83 (riscv_function_ok_for_sibcall): Likewise.
84 (riscv_set_current_function): Add interrupt handler support.
85 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
86 * config/riscv/riscv.md (UNSPECV_MRET): New.
88 (riscv_frflags, riscv_fsflags): Use tab after opcode.
90 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
92 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
94 PR tree-optimization/85712
95 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
96 this candidate has already been replaced in-situ by a copy.
98 2018-05-25 Jason Merrill <jason@redhat.com>
100 PR c++/80485 - inline function non-zero address.
101 * symtab.c (nonzero_address): Check DECL_COMDAT.
103 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
106 * config/alpha/alpha.md (ashlsi3): New insn pattern.
107 (*ashlsi_se): Rename from *ashldi_se. Define as sign
108 extension of SImode operation. Use const123_operand predicate.
110 (*saddl_se_1): Ditto.
112 (*ssubl_se_1): Ditto.
113 * config/alpha/predicates.md (const123_operand): New predicate.
114 * config/alpha/constraints.md (P): Use IN_RANGE.
116 2018-05-25 Richard Biener <rguenther@suse.de>
118 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
120 (ref_maybe_used_by_stmt_p): Likewise.
121 (stmt_may_clobber_ref_p): Likewise.
122 (stmt_may_clobber_ref_p_1): Likewise.
123 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
125 (ref_maybe_used_by_stmt_p): Likewise.
126 (stmt_may_clobber_ref_p): Likewise.
127 (stmt_may_clobber_ref_p_1): Likewise.
128 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
129 the alias oracle to disambiguate DRs with stmts DR analysis
131 (vect_analyze_data_refs): Do not give up on not analyzable
132 DRs for BB vectorization. Remove code truncating the dataref
135 2018-05-25 Jakub Jelinek <jakub@redhat.com>
138 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
139 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
140 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
142 2018-05-25 Richard Biener <rguenther@suse.de>
144 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
145 function, combining stmt data ref gathering and fatal analysis
147 (vect_analyze_data_refs): Remove now redudnant code and simplify.
148 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
149 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
150 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
151 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
153 2018-05-25 Bin Cheng <bin.cheng@arm.com>
155 PR tree-optimization/85720
156 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
157 SCC if all partitions are builtins.
158 (version_loop_by_alias_check): New parameter. Generate cancelable
159 runtime alias check if all partitions are builtins.
160 (distribute_loop): Update call to above function.
162 2018-05-25 Bin Cheng <bin.cheng@arm.com>
164 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
165 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
166 (parm_default_def_partition_arg): Ditto.
167 (set_parm_default_def_partition): Ditto.
168 (get_parm_default_def_partitions): Ditto and make it static.
169 (get_undefined_value_partitions): Ditto and make it static.
170 (remove_ssa_form): Refactor call to init_var_map here.
171 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
172 computation for loop region.
173 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
174 (register_default_def): Delete.
175 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
176 (parm_default_def_partition_arg): Ditto.
177 (set_parm_default_def_partition): Ditto.
178 (get_parm_default_def_partitions): Ditto and make it static.
179 (get_undefined_value_partitions): Ditto and make it static.
180 (coalesce_with_default, coalesce_with_default): Update comment.
181 (create_coalesce_list_for_region): New func factored out from
182 create_outofssa_var_map.
183 (populate_coalesce_list_for_outofssa): New func factored out from
184 create_outofssa_var_map and coalesce_ssa_name.
185 (create_outofssa_var_map): Delete.
186 (coalesce_ssa_name): Refactor to support live range computation.
187 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
188 (get_parm_default_def_partitions): Delete.
189 (get_undefined_value_partitions): Ditto.
190 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
191 computation for loop region.
192 (new_tree_live_info, loe_visit_block): Ditto.
193 (live_worklist, set_var_live_on_entry): Ditto.
194 (calculate_live_on_exit, verify_live_on_entry): Ditto.
195 * tree-ssa-live.h (struct _var_map): New fields.
196 (init_var_map): Change decl.
197 (region_contains_p): New.
199 2018-05-25 Bin Cheng <bin.cheng@arm.com>
201 * tree-ssa-live.h (live_merge_and_clear): Delete.
203 2018-05-25 Richard Biener <rguenther@suse.de>
206 * tree-dump.c (dequeue_and_dump): Remove access to removed
207 operand 2 of a SWITCH_EXPR.
209 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
211 * doc/sourcebuild.texi (vect_double_cond_arith): Include
212 multiplication and division.
213 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
214 (cond_udiv@var{m}, cond_umod@var{m}): Document.
215 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
216 (cond_udiv_optab, cond_umod_optab): New optabs.
217 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
218 (IFN_COND_RDIV): New internal functions.
219 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
220 TRUNC_MOD_EXPR and RDIV_EXPR.
221 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
222 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
224 (SVE_INT_BINARY): Include mult.
225 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
226 (optab, sve_int_op): Handle mult.
227 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
229 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
230 for SVE_INT_BINARY_SD.
232 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
234 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
235 (optab, sve_int_op): Handle div and udiv.
236 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
237 for SVE_INT_BINARY_SD.
238 (*<optab><mode>3): New insn for the same.
240 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
242 * tree-vect-patterns.c: Include predict.h.
243 (vect_recog_divmod_pattern): Restrict check for division support
244 to when optimizing for size.
246 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
248 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
249 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
250 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
251 (gimple_match_op::set_op): Likewise.
252 (gimple_resimplify4): Declare.
253 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
254 (expr::gen_transform): Likewise.
255 (decision_tree::gen): Generate a simplification routine for 4 operands.
256 * gimple-match-head.c (gimple_simplify): Add an overload for
257 4 operands. In the top-level function, handle up to 4 call
258 arguments and call gimple_resimplify4.
259 (gimple_resimplify4): New function.
260 (build_call_internal): Pass a fourth operand.
261 (maybe_push_to_seq): Likewise.
262 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
263 Fold VEC_COND_EXPRs of an operation and a default value into
264 an IFN_COND_* function if possible.
265 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
267 (SVE_COND_FP_BINARY): Include them.
268 (optab, sve_fp_op): Handle them.
269 (SVE_INT_BINARY_REV): New code iterator.
270 (SVE_COND_FP_BINARY_REV): New int iterator.
271 (commutative): New int attribute.
272 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
274 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
276 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
277 (*cond_<optab><mode>): New patterns for reversed operands.
279 2018-05-25 Richard Biener <rguenther@suse.de>
281 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
282 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
283 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
284 (STMT_VINFO_GROUPED_ACCESS): Adjust.
285 * tree-vect-data-refs.c (everywhere): Adjust users.
286 * tree-vect-loop.c (everywhere): Likewise.
287 * tree-vect-slp.c (everywhere): Likewise.
288 * tree-vect-stmts.c (everywhere): Likewise.
289 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
291 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
293 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
295 (gcc_cv_as_section_exclude): ... this.
296 Try Solaris as #exclude syntax.
297 * configure: Regenerate.
298 * config.in: Regenerate.
299 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
301 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
302 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
304 * varasm.c (default_elf_asm_named_section): Don't check if
305 HAVE_GAS_SECTION_EXCLUDE is defined.
307 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
309 * doc/md.texi: Update the documentation of the cond_* optabs
310 to mention the new final operand. Fix GET_MODE_NUNITS call.
311 Describe the scalar case too.
312 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
313 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
315 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
316 (get_conditional_internal_fn): Update comment.
317 * tree-vect-loop.c (vectorizable_reduction): Pass the original
318 accumulator value as a final argument to conditional functions.
319 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
320 a define_expand and add an "else" operand. Assert for now that
321 the else operand is equal to operand 2. Use SVE_INT_BINARY and
322 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
323 (*cond_<optab><mode>): New patterns.
324 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
325 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
326 (UNSPEC_COND_EOR): Delete.
327 (optab): Remove associated mappings.
328 (SVE_INT_BINARY): New code iterator.
329 (sve_int_op): Remove int attribute and add "minus" to the code
331 (SVE_COND_INT_OP): Delete.
332 (SVE_COND_FP_OP): Rename to...
333 (SVE_COND_FP_BINARY): ...this.
335 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
337 * optabs.c (can_reuse_operands_p): New function.
338 (maybe_legitimize_operands): Try to reuse the results for
341 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
343 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
344 Add {q} suffix to insn mnemonic.
346 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
348 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
349 (msp430_warn_func_return): New.
351 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
353 * fold-const.c (tree_nonzero_bits): New function.
354 * fold-const.h (tree_nonzero_bits): Likewise.
355 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
356 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
358 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
362 * varasm.c (assemble_alias): Check ifunc_resolver only on
365 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
368 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
369 when memory input operand is handled.
371 2018-05-24 Luis Machado <luis.machado@linaro.org>
373 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
375 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
377 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
379 * match.pd: Delay FMA folds until after vectorization.
381 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
384 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
385 address check not strict.
387 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
389 * gimple-match.h (gimple_match_op): New class.
390 (mprts_hook): Replace parameters with a gimple_match_op *.
391 (maybe_build_generic_op): Likewise.
392 (gimple_simplified_result_is_gimple_val): Replace parameters with
393 a const gimple_match_op *.
394 (gimple_simplify): Replace code_helper * and tree * parameters with
395 a gimple_match_op * parameter.
396 (gimple_resimplify1): Replace code_helper *, tree and tree *
397 parameters with a gimple_match_op * parameter.
398 (gimple_resimplify2): Likewise.
399 (gimple_resimplify3): Likewise.
400 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
401 parameters with a gimple_match_op * parameter.
402 * gimple-match-head.c (gimple_simplify): Change prototypes of
403 auto-generated functions to take a gimple_match_op * instead of
404 separate code_helper * and tree * parameters. Make the same
405 change in the top-level overload and update calls to the
406 gimple_resimplify routines. Update calls to the auto-generated
407 functions and to maybe_push_res_to_seq in the publicly-facing
408 operation-specific gimple_simplify overloads.
409 (gimple_match_op::MAX_NUM_OPS): Define.
410 (gimple_resimplify1): Replace rcode and ops with a single res_op
411 parameter. Update call to gimple_simplify.
412 (gimple_resimplify2): Likewise.
413 (gimple_resimplify3): Likewise.
414 (mprts_hook): Replace parameters with a gimple_match_op *.
415 (maybe_build_generic_op): Likewise.
416 (build_call_internal): Replace type, nargs and ops with
418 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
419 with a single gimple_match_op *. Update calls to mprts_hook,
420 build_call_internal and gimple_simplified_result_is_gimple_val.
421 Factor out code that is common to the tree_code and combined_fn cases.
422 * genmatch.c (expr::gen_transform): Replace tem_code and
423 tem_ops with a gimple_match_op called tem_op. Update calls
424 to the gimple_resimplify functions and maybe_push_res_to_seq.
425 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
426 res_ops. Update call to the gimple_resimplify functions.
427 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
428 (decision_tree::gen): Make the functions take a gimple_match_op *
429 called res_op instead of separate res_code and res_ops parameters.
430 Update call accordingly.
431 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
432 and ops with a single res_op parameter. Update calls to
433 maybe_build_generic_op and maybe_push_res_to_seq.
434 (fold_stmt_1): Update calls to gimple_simplify and
435 replace_stmt_with_simplification.
436 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
437 and gimple_simplified_result_is_gimple_val.
438 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
440 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
441 with a gimple_match_op *.
442 (vn_nary_build_or_lookup): Likewise. Update call to
443 vn_nary_build_or_lookup_1.
444 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
445 gimple_match_op *. Update calls to the gimple_resimplify routines
446 and to gimple_simplified_result_is_gimple_val.
447 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
448 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
449 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
450 (visit_nary_op): Likewise.
451 (visit_reference_op_load): Likewise.
453 2018-05-23 Luis Machado <luis.machado@linaro.org>
455 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
456 modifier for printing the step amount.
458 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
461 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
464 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
466 * doc/sourcebuild.texi (Endianness): New subsubsection.
468 2018-05-23 Luis Machado <luis.machado@linaro.org>
470 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
471 <prefetch_dynamic_strides>: New const bool field.
472 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
473 prefetch_dynamic_strides.
474 (exynosm1_prefetch_tune): Likewise.
475 (thunderxt88_prefetch_tune): Likewise.
476 (thunderx_prefetch_tune): Likewise.
477 (thunderx2t99_prefetch_tune): Likewise.
478 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
480 (aarch64_override_options_internal): Update to set
481 PARAM_PREFETCH_DYNAMIC_STRIDES.
482 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
483 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
484 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
485 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
486 prefetch-dynamic-strides setting.
488 2018-05-23 Luis Machado <luis.machado@linaro.org>
490 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
491 <minimum_stride>: New const int field.
492 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
493 minimum_stride field defaulting to -1.
494 (exynosm1_prefetch_tune): Likewise.
495 (thunderxt88_prefetch_tune): Likewise.
496 (thunderx_prefetch_tune): Likewise.
497 (thunderx2t99_prefetch_tune): Likewise.
498 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
499 <default_opt_level>: Set to 3.
500 (aarch64_override_options_internal): Update to set
501 PARAM_PREFETCH_MINIMUM_STRIDE.
502 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
503 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
504 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
505 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
506 stride is constant and is below the minimum stride threshold.
508 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
510 * config/arm/arm-cpus.in (mode26): Delete.
511 (armv4): Delete mode26 reference.
512 * config/arm/arm.c (arm_configure_build_target): Delete use of
515 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
517 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
519 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
520 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
522 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
523 (floatunsdidf2): Ditto.
525 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
527 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
528 (fixuns_trunc<mode>si2_avx512f): Ditto.
529 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
530 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
531 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
533 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
535 PR rtl-optimization/79985
536 * df-scan.c (df_insn_refs_collect): Remove special case for
537 global registers and asm statements.
539 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
541 * extend.texi (Global Register Variables): Rewrite the bullet list.
542 Note that the register is available for allocation. Note that access
543 via inline asm must use constraints. Add note about async-signal
544 handlers. Remove paragraph about automagic register selection.
546 2018-05-23 Richard Biener <rguenther@suse.de>
548 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
549 of fixed offset from memset VN.
551 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
553 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
555 (alloc_cand_and_find_basis): Initialize first_interp field.
556 (slsr_process_mul): Modify first_interp field.
557 (slsr_process_add): Likewise.
558 (slsr_process_cast): Modify first_interp field for each new
560 (slsr_process_copy): Likewise.
561 (dump_candidate): Dump first_interp field.
562 (replace_mult_candidate): Process all interpretations, not just
564 (replace_rhs_if_not_dup): Likewise.
565 (replace_one_candidate): Likewise.
567 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
569 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
571 (aarch64_needs_frame_chain): New function.
572 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
574 2018-05-23 Sudakshina Das <sudi.das@arm.com>
577 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
578 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
579 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
580 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
581 as true for strict-align.
582 (aarch64_can_inline_p): Perform checks even when callee has no
583 attributes to check for strict alignment.
584 * doc/extend.texi (AArch64 Function Attributes): Document
586 * doc/invoke.texi: (AArch64 Options): Likewise.
588 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
590 PR tree-optimization/85853
591 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
592 the handling of the root of the node to...
593 (vect_slp_analyze_node_operations_1): ...this new function,
594 and run the whole thing with the child nodes' def types
595 set according to their SLP node's def type.
597 2018-05-23 Richard Biener <rguenther@suse.de>
600 * tree-data-ref.c (create_runtime_alias_checks): Defer
601 and ignore overflow warnings.
603 2018-05-23 Yury Gribov <tetra2005@gmail.com>
605 PR tree-optimization/85822
606 * tree-vrp.c (is_masked_range_test): Fix handling of negative
609 2018-05-23 Richard Biener <rguenther@suse.de>
611 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
612 memset constants via native_interpret_expr.
614 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
617 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
619 (cgraph_node::create_alias): Likewise.
620 (cgraph_node::get_availability): Check ifunc_resolver instead
621 of looking up ifunc attribute.
622 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
623 * varasm.c (do_assemble_alias): Likewise.
624 (assemble_alias): Likewise.
625 (default_binds_local_p_3): Likewise.
626 * cgraph.h (cgraph_node): Add ifunc_resolver.
627 (cgraph_node::only_called_directly_or_aliased_p): Return false
629 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
631 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
632 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
633 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
634 instead of looking up ifunc attribute.
636 2018-05-22 Luis Machado <luis.machado@linaro.org>
638 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
640 2018-05-22 Martin Sebor <msebor@redhat.com>
643 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
644 only when expasion succeeds.
645 (expand_builtin_strcmp): Same.
646 (expand_builtin_strncmp): Same.
648 2018-05-22 Martin Sebor <msebor@redhat.com>
650 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
652 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
653 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
655 * config/aarch64/aarch64-ldpstp.md: Replace uses of
656 aarch64_mem_pair_operand with memory_operand and delete operand swapping
658 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
659 Add check for legitimate_address.
660 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
661 (aarch64_swap_ldrstr_operands): New.
662 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
665 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
666 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
668 * config/aarch64/aarch64.md: New patterns to generate stp
670 (store_pair_sw, store_pair_dw): New patterns to generate stp for
671 single words and double words.
672 (load_pair_sw, load_pair_dw): Likewise.
673 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
675 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
677 * config/aarch64/aarch64-ldpstp.md: Modify peephole
678 for different mode ldpstp and add peephole for merged zero stores.
680 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
682 (aarch64_gen_store_pair): Rename calls to match new patterns.
683 (aarch64_gen_load_pair): Rename calls to match new patterns.
684 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
685 (load_pair<DREG:mode><DREG2:mode>): ... This.
686 (store_pair<mode>): Rename to...
687 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
688 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
690 (V_INT_EQUIV): Handle SImode.
691 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
694 2018-05-22 Martin Sebor <msebor@redhat.com>
697 * calls.c (maybe_warn_nonstring_arg): Use string length to set
698 or ajust the presumed bound on an operation to avoid unnecessary
701 2018-05-22 Martin Sebor <msebor@redhat.com>
703 PR tree-optimization/85826
704 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
705 assuming that a DECL necesarily has a constant size.
707 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
710 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
712 2018-05-22 Richard Biener <rguenther@suse.de>
714 PR tree-optimization/85834
715 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
716 non-constant and non-zero memset arguments.
718 2018-05-22 Martin Liska <mliska@suse.cz>
721 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
723 2018-05-22 Richard Biener <rguenther@suse.de>
725 PR tree-optimization/85863
726 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
727 comparisons when vectype is specified.
728 (vectorizable_condition): Do not specify vectype for
729 vect_is_simple_cond when SLP vectorizing.
731 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
734 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
735 define __ibm128 as long double.
736 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
737 as a distinct type when IEEE 128-bit support is enabled.
738 (init_float128_ieee): Fix up conversions between IFmode and IEEE
739 128-bit types to use the correct functions.
740 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
741 convert between 128-bit floating point types that have different
742 modes but the same representation, instead of using gen_lowpart to
744 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
746 (IFKF_reg): New attributes to give the register constraints for
748 (extend<mode>tf2_internal): New insns to mark an explicit
749 conversion between 128-bit floating point types that have a
750 different mode but share the same representation.
752 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
754 PR tree-optimization/85814
755 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
756 a null return from get_strinfo when unsharing the next
757 strinfo in the chain.
759 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
762 * varasm.c (weak_finish): Clean up weak_decls.
764 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
766 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
767 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
768 UNSPEC_UADALP values.
769 * config/aarch64/iterators.md (ABAL): New int iterator.
772 (sur): Add mappings for the above.
773 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
775 (aarch64_<sur>abal<mode>_4): Likewise.
776 (aarch64_<sur>adalp<mode>_3): Likewise.
777 (<sur>sadv16qi): New define_expand.
779 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
781 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
782 (*movdf_internal): Ditto.
783 (*rcpsf2_sse): Ditto.
784 (*rsqrtsf2_sse): Ditto.
785 (*sqrt<mode>2_sse): Ditto.
787 2018-05-21 Tamar Christina <tamar.christina@arm.com>
789 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
791 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
792 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
793 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
794 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
796 * config/aarch64/arm_neon.h: Likewise.
797 * config/aarch64/iterators.md (VQ_I): New.
799 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
801 * config.gcc: Add arc/t-multilib-linux to tmake_file for
803 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
806 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
808 * config/nds32/constraints.md (S): New constraint.
809 * config/nds32/nds32.md (call_internal): Use constraint S.
810 (call_value_internal): Likewise.
811 (sibcall_internal): Likewise.
812 (sibcall_value_internal): Likewise.
814 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
815 Chung-Ju Wu <jasonwucj@gmail.com>
817 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
820 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
821 Chung-Ju Wu <jasonwucj@gmail.com>
823 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
824 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
825 (nds32_rtx_costs_impl): Simplify.
826 (nds32_address_cost_impl): Simplify.
827 (nds32_init_rtx_costs): New function.
828 (nds32_rtx_costs_speed_prefer): Likewise.
829 (nds32_rtx_costs_size_prefer): Likewise.
830 (nds32_address_cost_speed_prefer): Likewise.
831 (nds32_address_cost_speed_fwprop): Likewise.
832 (nds32_address_cost_size_prefer): Likewise.
833 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
834 * config/nds32/nds32.c (nds32_option_override): Use
835 nds32_init_rtx_costs function.
837 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
839 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
840 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
841 (TARGET_PIPELINE_N8): Likewise.
842 (TARGET_PIPELINE_N10): Likewise.
843 (TARGET_PIPELINE_N13): Likewise.
844 (TARGET_PIPELINE_GRAYWOLF): Likewise.
846 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
848 * config/nds32/nds32-fpu.md: Update copyright year.
850 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
852 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
854 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
856 * config/nds32/nds32.c
857 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
858 * config/nds32/nds32.opt (minline-asm-r15): New option.
860 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
862 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
864 * config/nds32/nds32.md (abssi2): New pattern.
866 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
868 * config/i386/i386.md (rex64namesuffix): New mode attribute.
869 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
870 Merge insn pattern from sse_cvtsi2ss<round_name> and
871 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
872 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
873 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
874 using SWI48 mode iterator.
875 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
876 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
877 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
878 pattern from sse_cvttss2si<round_saeonly_name>
879 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
880 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
881 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
882 using SWI48 mode iterator.
883 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
884 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
885 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
886 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
887 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
888 using SWI48 mode iterator.
889 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
890 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
891 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
892 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
893 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
895 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
896 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
897 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
898 pattern from sse_cvttsd2si<round_saeonly_name>
899 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
901 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
903 * config/nds32/nds32-md-auxiliary.c
904 (nds32_valid_smw_lwm_base_p): Refine.
905 (nds32_output_smw_single_word): Refine.
906 (nds32_output_smw_double_word): New.
907 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
909 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
911 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
912 (nds32_output_stack_pop): Refine.
913 (nds32_expand_unaligned_load): Refine.
914 (nds32_expand_unaligned_store): Refine.
916 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
917 Chung-Ju Wu <jasonwucj@gmail.com>
919 * config/nds32/constants.md: Add TP_REGNUM constant.
920 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
921 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
923 * config/nds32/nds32-doubleword.md: Consider flag_pic.
924 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
925 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
926 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
927 and PIC code generation.
928 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
930 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
932 * config/nds32/nds32.md: Support TLS and PIC.
933 * config/nds32/nds32.c: Support TLS and PIC.
934 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
935 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
938 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
940 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
943 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
944 Chung-Ju Wu <jasonwucj@gmail.com>
946 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
947 * config/nds32/nds32-md-auxiliary.c
948 (symbolic_reference_mentioned_p): New.
949 (nds32_legitimize_ict_address): New.
950 (nds32_expand_ict_move): New.
951 (nds32_indirect_call_referenced_p): New.
952 (nds32_symbol_binds_local_p): Delete.
953 (nds32_long_call_p): Modify.
954 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
955 * config/nds32/nds32-protos.h
956 (symbolic_reference_mentioned_p): Declare.
957 (nds32_legitimize_ict_address): Declare.
958 (nds32_expand_ict_move): Declare.
959 (nds32_indirect_call_referenced_p): Declare.
960 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
961 (nds32_relax_group): Use nds32_ict_const_p as condition.
962 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
963 (nds32_asm_file_start): Output ict_model directive in asm code.
964 (nds32_legitimate_address_p): Consider indirect call.
965 (nds32_print_operand): Consider indirect call.
966 (nds32_print_operand_address): Consider indirect call.
967 (nds32_insert_attributes): Handle "indirect_call" attribute.
968 (TARGET_LEGITIMATE_ADDRESS_P): Define.
969 (TARGET_LEGITIMATE_CONSTANT_P): Define.
970 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
971 (TARGET_DELEGITIMIZE_ADDRESS): Define.
972 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
973 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
974 (TARGET_ICT_MODEL_SMALL): Define.
975 (TARGET_ICT_MODEL_LARGE): Define.
976 * config/nds32/nds32.md (movsi): Consider ict model.
977 (call, call_value): Consider ict model.
978 (sibcall, sibcall_value): Consider ict model.
979 * config/nds32/nds32.opt (mict-model): New option.
980 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
983 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
984 Monk Chiang <sh.chiang04@gmail.com>
985 Jim Wilson <jimw@sifive.com>
987 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
988 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
989 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
990 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
991 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
992 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
993 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
994 compute save_libcall_adjustment properly.
995 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
996 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
997 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
998 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
999 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1000 (ABI_SPEC): Handle mabi=ilp32e.
1001 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1002 (RVE): Add RVE mask.
1003 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1004 <-march>: Add rv32e as an example.
1006 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1009 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1010 (intra_create_variable_infos): Handle C++ constructors.
1012 2018-05-18 Martin Liska <mliska@suse.cz>
1014 * passes.def: Remove a redundant pass.
1016 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1019 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1021 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1023 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1025 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1027 (armv2, armv2a, armv3, armv3m): Delete architectures.
1028 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1029 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1030 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1032 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1033 (*mulsidi3adddi): Likewise.
1034 (mulsidi3): Likewise.
1035 (*mulsidi3_nov6): Likewise.
1036 (umulsidi3): Likewise.
1037 (umulsidi3_nov6): Likewise.
1038 (umaddsidi4): Likewise.
1039 (*umulsidi3adddi): Likewise.
1040 (smulsi3_highpart): Likewise.
1041 (*smulsi3_highpart_nov6): Likewise.
1042 (umulsi3_highpart): Likewise.
1043 (*umulsi3_highpart_nov6): Likewise.
1044 * config/arm/arm.h (arm_arch3m): Delete.
1045 * config/arm/arm.c (arm_arch3m): Delete.
1046 (arm_option_override_internal): Update armv3-related comment.
1047 (arm_configure_build_target): Delete use of isa_bit_mode32.
1048 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1049 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1050 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1052 (mulusa3): Likewise.
1053 * config/arm/arm-protos.h (arm_arch3m): Delete.
1054 * config/arm/arm-tables.opt: Regenerate.
1055 * config/arm/arm-tune.md: Likewise.
1056 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1057 deleted architectures.
1059 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1061 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1062 (armv5t, armv5te): New features.
1063 (ARMv5, ARMv5e): Delete fgroups.
1064 (ARMv5t, ARMv5te): Adjust for above changes.
1066 (armv5, armv5e): Delete arches.
1067 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1069 (*call_reg_arm): Likewise.
1070 (*call_value_reg_armv5): Likewise.
1071 (*call_value_reg_arm): Likewise.
1072 (*call_symbol): Likewise.
1073 (*call_value_symbol): Likewise.
1074 (*sibcall_insn): Likewise.
1075 (*sibcall_value_insn): Likewise.
1077 (prefetch): Likewise.
1078 (define_split and define_peephole2 dependent on arm_arch5):
1080 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1082 (TARGET_ARM_QBIT): Likewise.
1083 (TARGET_DSP_MULTIPLY): Likewise.
1084 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1085 (arm_arch5, arm_arch5e): Delete.
1086 (arm_arch5t, arm_arch5te): Declare.
1087 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1088 (arm_arch5t): Declare.
1089 (arm_option_reconfigure_globals): Update for the above.
1090 (arm_options_perform_arch_sanity_checks): Update comment, replace
1091 use of arm_arch5 with arm_arch5t.
1092 (use_return_insn): Likewise.
1093 (arm_emit_call_insn): Likewise.
1094 (output_return_instruction): Likewise.
1095 (arm_final_prescan_insn): Likewise.
1096 (arm_coproc_builtin_available): Likewise.
1097 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1098 arm_arch5e with arm_arch5t and arm_arch5te.
1099 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1100 (arm_arch5t, arm_arch5te): Declare.
1101 * config/arm/arm-tables.opt: Regenerate.
1102 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1103 * config/arm/t-multilib: Likewise.
1104 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1105 instead of arm_arch5.
1106 (*call_reg_thumb1): Likewise.
1107 (*call_value_reg_thumb1_v5): Likewise.
1108 (*call_value_reg_thumb1): Likewise.
1109 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1111 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1113 2018-05-18 Martin Liska <mliska@suse.cz>
1115 PR gcov-profile/84846
1116 * doc/gcov.texi: Document -t option of gcov tool.
1118 2018-05-18 Martin Liska <mliska@suse.cz>
1120 PR gcov-profile/84846
1121 * gcov.c (print_usage): Add new -t option.
1122 (process_args): Handle the option.
1123 (generate_results): Use stdout as output when requested by
1126 2018-05-18 Martin Liska <mliska@suse.cz>
1128 PR gcov-profile/84846
1129 * coverage.c (coverage_init): Write PWD to .gcno file.
1130 * doc/gcov.texi: Document how working directory is printed.
1131 * gcov-dump.c (dump_gcov_file): Print PWD.
1132 * gcov.c (output_intermediate_file): Likewise.
1133 (read_graph_file): Read PWD string.
1134 (output_lines): Print PWD.
1136 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1139 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1140 for retval and return false if all args to phi are zero.
1142 2018-05-18 Richard Biener <rguenther@suse.de>
1144 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1146 (evrp_dom_walker::before_dom_children): Call it.
1148 2018-05-18 Richard Biener <rguenther@suse.de>
1150 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1151 results when processing array refs with variable index.
1153 2018-05-18 Toon Moene <toon@moene.org>
1155 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1156 directly after that of -floop-interchange. Indicate that both
1157 options are enabled by default when specifying -O3.
1159 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1161 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1162 iterator. Delete separate integer-mode vec_set<mode> expander.
1163 (aarch64_simd_vec_setv2di): Delete.
1164 (vec_setv2di): Delete.
1165 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1166 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1167 the "w, r" alternative.
1169 2018-05-18 Martin Liska <mliska@suse.cz>
1171 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1172 * tree-pass.h (make_pass_lower_switch_O0): New function.
1173 * tree-switch-conversion.c (node_has_low_bound): Remove.
1174 (node_has_high_bound): Likewise.
1175 (node_is_bounded): Likewise.
1176 (class pass_lower_switch): Make it a template type and create
1178 (pass_lower_switch::execute): Add template argument.
1179 (make_pass_lower_switch): New function.
1180 (make_pass_lower_switch_O0): New function.
1181 (do_jump_if_equal): Remove.
1182 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1183 all the hard work to tree optimization passes.
1185 2018-05-18 Martin Liska <mliska@suse.cz>
1187 * dbgcnt.c (limit_low): Renamed from limit.
1188 (limit_high): New variable.
1189 (dbg_cnt_is_enabled): Check for upper limit.
1190 (dbg_cnt): Adjust dumping.
1191 (dbg_cnt_set_limit_by_index): Add new argument for high
1193 (dbg_cnt_set_limit_by_name): Likewise.
1194 (dbg_cnt_process_single_pair): Parse new format.
1195 (dbg_cnt_process_opt): Use strtok.
1196 (dbg_cnt_list_all_counters): Remove 'value' and add
1198 * doc/invoke.texi: Document changes.
1200 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1202 * doc/sourcebuild.texi (scalar_all_fma): Document.
1203 * tree.def (FMA_EXPR): Delete.
1204 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1205 * internal-fn.c (ternary_direct): New macro.
1206 (expand_ternary_optab_fn): Likewise.
1207 (direct_ternary_optab_supported_p): Likewise.
1208 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1209 * builtins.c (fold_builtin_fma): Delete.
1210 (fold_builtin_3): Don't call it.
1211 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1212 * expr.c (expand_expr_real_2): Likewise.
1213 * fold-const.c (operand_equal_p): Likewise.
1214 (fold_ternary_loc): Likewise.
1215 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1216 * gimple.c (DEFTREECODE): Likewise.
1217 * gimplify.c (gimplify_expr): Likewise.
1218 * optabs-tree.c (optab_for_tree_code): Likewise.
1219 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1220 * tree-eh.c (operation_could_trap_p): Likewise.
1221 (stmt_could_throw_1_p): Likewise.
1222 * tree-inline.c (estimate_operator_cost): Likewise.
1223 * tree-pretty-print.c (dump_generic_node): Likewise.
1224 (op_code_prio): Likewise.
1225 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1226 * tree-ssa-operands.c (get_expr_operands): Likewise.
1227 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1228 * fold-const-call.h (fold_fma): Delete.
1229 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1230 CFN_FNMA and CFN_FNMS.
1232 * genmatch.c (combined_fn): New enum.
1233 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1234 (commutative_op): New function.
1235 (commutate): Use it. Handle more than 2 operands.
1236 (dt_operand::gen_gimple_expr): Use commutative_op.
1237 (parser::parse_expr): Allow :c to be used with non-binary
1238 operators if the commutative operand is known.
1239 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1240 CFN_FMS, CFN_FNMA and CFN_FNMS.
1241 (backprop::process_assign_use): Remove FMA_EXPR handling.
1242 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1243 (gen_hsa_fma): New function.
1244 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1245 IFN_FNMA and IFN_FNMS.
1246 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1247 * gimple-fold.h (follow_all_ssa_edges): Declare.
1248 * gimple-fold.c (follow_all_ssa_edges): New function.
1249 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1250 gimple_build interface and use follow_all_ssa_edges to fold the result.
1251 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1252 instead of checking for optabs directly.
1253 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1254 rather than FMA_EXPRs.
1255 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1256 call to IFN_FMA instead of an FMA_EXPR.
1258 2018-05-17 Jim Wilson <jimw@sifive.com>
1260 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1261 sign extended promoted subreg, and the range does not have the sign bit
1262 set, then do a sign extend.
1264 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1265 test, check for sign extended subreg and/or constant operands, and
1266 do a sign extend in that case.
1268 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1270 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1271 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1273 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1274 Change logics_shift_reg to logics_shift_imm.
1275 (thunderx2t99_fp_loadpair_basic): Delete.
1276 (thunderx2t99_fp_storepair_basic): Delete.
1277 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1278 (thunderx2t99_asimd_polynomial): Delete.
1279 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1280 and neon_fp_mul_d_scalar_q.
1281 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1282 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1283 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1284 (thunderx2t99_asimd_lut): Add missing tbl types.
1285 (thunderx2t99_asimd_ext): Delete.
1286 (thunderx2t99_asimd_load1_1_mult): Delete.
1287 (thunderx2t99_asimd_load1_2_mult): Delete.
1288 (thunderx2t99_asimd_load1_ldp): New.
1289 (thunderx2t99_asimd_load1): New.
1290 (thunderx2t99_asimd_load2): Add missing *load2* types.
1291 (thunderx2t99_asimd_load3): New.
1292 (thunderx2t99_asimd_load4): New.
1293 (thunderx2t99_asimd_store1_1_mult): Delete.
1294 (thunderx2t99_asimd_store1_2_mult): Delete.
1295 (thunderx2t99_asimd_store2_mult): Delete.
1296 (thunderx2t99_asimd_store2_onelane): Delete.
1297 (thunderx2t99_asimd_store_stp): New.
1298 (thunderx2t99_asimd_store1): New.
1299 (thunderx2t99_asimd_store2): New.
1300 (thunderx2t99_asimd_store3): New.
1301 (thunderx2t99_asimd_store4): New.
1303 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1305 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1306 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1308 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1309 Segher Boessenkool <segher@kernel.crashing.org>
1312 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1315 2018-05-17 Richard Biener <rguenther@suse.de>
1317 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1318 for pruning loop and prune defs feeding only already visited PHIs.
1320 2018-05-17 Richard Biener <rguenther@suse.de>
1322 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1324 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1325 Richard Biener <rguenther@suse.de>
1327 PR tree-optimization/85793
1328 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1329 for VMAT_ELEMENTWISE.
1331 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1333 * internal-fn.h (lookup_internal_fn): Declare
1334 * internal-fn.c (lookup_internal_fn): New function.
1335 * gimple.c (gimple_build_call_from_tree): Handle calls to
1337 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1338 internal function names.
1339 * tree-pretty-print.c (dump_generic_node): Likewise.
1340 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1342 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1344 * gimple-fold.h (gimple_build): Make the function forms take
1345 combined_fn rather than built_in_function.
1346 (gimple_simplify): Likewise.
1347 * gimple-match-head.c (gimple_simplify): Likewise.
1348 * gimple-fold.c (gimple_build): Likewise.
1349 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1350 rather than gimple_build_call_internal.
1351 (get_initial_defs_for_reduction): Likewise.
1352 (vect_create_epilog_for_reduction): Likewise.
1353 (vectorizable_live_operation): Likewise.
1355 2018-05-17 Martin Liska <mliska@suse.cz>
1357 * gimple-ssa-sprintf.c (format_directive): Do not use
1358 space in between 'G_' and '('.
1360 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1363 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1364 even if the mask is not all ones.
1367 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1369 (ix86_gimple_fold_builtin): Likewise.
1372 * config/i386/i386.c: Include tree-vector-builder.h.
1373 (ix86_vector_shift_count): New function.
1374 (ix86_fold_builtin): Fold shift builtins by scalar count.
1375 (ix86_gimple_fold_builtin): Likewise.
1377 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1378 _mm512_setzero): New intrinsics.
1380 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1381 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1383 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1384 code generation for cases where splatting a value is not useful.
1385 * simplify-rtx.c (simplify_ternary_operation): Simplify
1386 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1387 mode to a vec_concat.
1389 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1391 * config.gcc: Support "goldmont-plus".
1392 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1394 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1395 PROCESSOR_GOLDMONT_PLUS.
1396 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1397 (processor_target_table): Add "goldmont-plus".
1398 (PTA_GOLDMONT_PLUS): Define.
1399 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1400 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1401 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1402 (fold_builtin_cpu): Add "goldmont-plus".
1403 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1404 (ix86_option_override_internal): Add "goldmont-plus".
1405 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1406 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1407 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1408 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1410 2018-05-17 Richard Biener <rguenther@suse.de>
1412 PR tree-optimization/85757
1413 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1414 remove defs that only feed that PHI from further processing.
1416 2018-05-16 Jim Wilson <jimw@sifive.com>
1418 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1420 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1422 2018-05-16 Mark Wielaard <mark@klomp.org>
1424 * dwarf2out.c (count_index_strings): New function.
1425 (output_indirect_strings): Call count_index_strings and generate
1426 header for dwarf_version >= 5.
1428 2018-05-16 Mark Wielaard <mark@klomp.org>
1430 * dwarf2out.c (dwarf_FORM): New function.
1431 (set_indirect_string): Use dwarf_FORM.
1432 (reset_indirect_string): Likewise.
1433 (size_of_die): Likewise.
1434 (value_format): Likewise.
1435 (output_die): Likewise.
1436 (add_skeleton_AT_string): Likewise.
1437 (output_macinfo_op): Likewise.
1438 (index_string): Likewise.
1439 (output_index_string_offset): Likewise.
1440 (output_index_string): Likewise.
1441 (count_index_strings): Likewise.
1443 2018-05-16 Carl Love <cel@us.ibm.com>
1445 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1446 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1448 2018-05-16 Martin Jambor <mjambor@suse.cz>
1450 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1451 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1453 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1455 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1456 (fnma<mode>4): Likewise.
1457 (fms<mode>4): Likewise.
1458 (fnms<mode>4): Likewise.
1459 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1460 (aarch64_fnma<mode>4): Likewise.
1461 (aarch64_fms<mode>4): Likewise.
1462 (aarch64_fnms<mode>4): Likewise.
1463 (aarch64_fnmadd<mode>4): Likewise.
1465 2018-05-16 Jason Merrill <jason@redhat.com>
1467 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1469 2018-05-16 Richard Biener <rguenther@suse.de>
1471 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1472 (dump_stmt_cost): Declare.
1473 (add_stmt_cost): Dump cost we add.
1474 (add_stmt_costs): New function.
1475 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1477 (vect_analyze_stmt): Adjust prototype.
1478 (vectorizable_condition): Likewise.
1479 (vectorizable_live_operation): Likewise.
1480 (vectorizable_reduction): Likewise.
1481 (vectorizable_induction): Likewise.
1482 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1483 cost vector to pass to vectorizable_ and record afterwards.
1484 (vect_model_reduction_cost): Take cost vector argument and adjust.
1485 (vect_model_induction_cost): Likewise.
1486 (vectorizable_reduction): Likewise.
1487 (vectorizable_induction): Likewise.
1488 (vectorizable_live_operation): Likewise.
1489 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1490 SLP_TREE_NUMBER_OF_VEC_STMTS.
1491 (vect_analyze_slp_cost_1): Remove.
1492 (vect_analyze_slp_cost): Likewise.
1493 (vect_slp_analyze_node_operations): Take visited args and
1494 a target cost vector. Avoid processing already visited stmt sets.
1495 (vect_slp_analyze_operations): Use a local cost vector to gather
1496 costs and register those of non-discarded instances.
1497 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1498 (vect_schedule_slp_instance): Remove copying of
1499 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1501 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1502 adding cost. Record cost entry location.
1503 (vect_prologue_cost_for_slp_op): Function to compute cost of
1504 a constant or invariant generated for SLP vect in the prologue,
1505 split out from vect_analyze_slp_cost_1.
1506 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1507 (vect_model_promotion_demotion_cost): Likewise.
1508 (vect_model_store_cost): Likewise, make static.
1509 (vect_model_load_cost): Likewise.
1510 (vectorizable_bswap): Add cost vector arg and adjust.
1511 (vectorizable_call): Likewise.
1512 (vectorizable_simd_clone_call): Likewise.
1513 (vectorizable_conversion): Likewise.
1514 (vectorizable_assignment): Likewise.
1515 (vectorizable_shift): Likewise.
1516 (vectorizable_operation): Likewise.
1517 (vectorizable_store): Likewise.
1518 (vectorizable_load): Likewise.
1519 (vectorizable_condition): Likewise.
1520 (vectorizable_comparison): Likewise.
1521 (can_vectorize_live_stmts): Likewise.
1522 (vect_analyze_stmt): Likewise.
1523 (vect_transform_stmt): Adjust calls to vectorizable_*.
1524 * tree-vectorizer.c: Include gimple-pretty-print.h.
1525 (dump_stmt_cost): New function.
1527 2018-05-16 Richard Biener <rguenther@suse.de>
1529 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1530 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1531 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1532 (check_name): New callback.
1533 (dse_classify_store): Track cycles via a visited bitmap of PHI
1534 defs and simplify handling of in-loop and across loop dead stores
1535 and properly fail for loop-variant refs. Handle byte-tracking with
1536 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1539 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1541 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1542 (vect_get_mask_type_for_stmt): Likewise.
1543 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1545 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1546 to determine the statement's vector type and the vector type that
1547 should be used for calculating nunits. Deal with cases in which
1548 the type has to be deferred.
1549 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1550 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1551 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1552 (vect_determine_vf_for_stmt): New functions, split out from...
1553 (vect_determine_vectorization_factor): ...here.
1554 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1555 (vect_get_mask_type_for_stmt): New functions, split out from
1556 vect_determine_vectorization_factor.
1558 2018-05-16 Richard Biener <rguenther@suse.de>
1560 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1561 verify the [VEC_]COND_EXPR embedded comparison.
1563 2018-05-15 Martin Sebor <msebor@redhat.com>
1565 PR tree-optimization/85753
1566 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1567 RECORD_TYPE in addition to ARRAY_TYPE.
1569 2018-05-15 Martin Sebor <msebor@redhat.com>
1572 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1574 2018-05-15 Richard Biener <rguenther@suse.de>
1576 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1577 add by_clobber_p one. Change algorithm to collect all defs
1578 representing uses we need to walk and try reducing them to
1579 a single one before failing.
1580 (dse_dom_walker::dse_optimize_stmt): Adjust.
1582 2018-05-13 Mark Wielaard <mark@klomp.org>
1584 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1585 (size_of_loc_descr): Likewise.
1586 (output_loc_operands): Likewise.
1587 (output_loc_operands_raw): Likewise.
1588 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1589 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1590 (hash_loc_operands): Likewise.
1591 (compare_loc_operands): Likewise.
1593 2018-05-14 Mark Wielaard <mark@klomp.org>
1595 * dwarf2out.c (count_index_addrs): New function.
1596 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1598 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1600 PR tree-optimization/83648
1601 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1602 return value as malloc candidate.
1604 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1607 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1608 as true in call to suggest_attribute.
1610 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1612 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1613 -mreadonly-in-sdata.
1615 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1617 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1619 (aarch64_crypto_aesd_fused): Likewise.
1621 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1623 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1624 (movsi_aarch64): Likewise.
1625 (load_pairsi): Likewise.
1626 (load_pairdi): Likewise.
1627 (store_pairsi): Likewise.
1628 (store_pairdi): Likewise.
1629 (load_pairsf): Likewise.
1630 (load_pairdf): Likewise.
1631 (store_pairsf): Likewise.
1632 (store_pairdf): Likewise.
1633 (zero_extend): Likewise.
1634 (trunc): Swap alternatives.
1635 (fcvt_target): Add '?' to prefer w over r.
1637 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1640 * config/i386/i386.md: Disallow non-commutative arithmetics in
1641 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1642 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1643 in the peephole2 before it.
1645 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1647 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1648 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1649 (ix86_handle_option): Handle -mcldemote.
1650 * config.gcc: New header.
1651 * config/i386/cldemoteintrin.h: New file.
1652 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1653 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1655 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1656 OPTION_MASK_ISA_CLDEMOTE.
1657 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1658 (ix86_valid_target_attribute_inner_p): Ditto.
1659 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1660 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1661 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1662 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1663 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1665 * config/i386/i386.opt: Add -mcldemote.
1666 * config/i386/x86intrin.h: New header.
1667 * doc/invoke.texi: Add -mcldemote.
1669 2018-05-14 Richard Biener <rguenther@suse.de>
1671 * doc/match-and-simplify.texi: Adjust :s documentation.
1673 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1675 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1676 intended memcpy size.
1677 (REORDER_45): Likewise.
1679 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1681 * sort.cc: New file.
1682 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1683 * vec.c (qsort_chk): Use gcc_qsort.
1684 * Makefile.in (OBJS-libcommon): Add sort.o.
1685 (build/sort.o): New target. Use it...
1686 (BUILD_RTL): ... here, and...
1687 (build/gencfn-macros): ... here, and...
1688 (build/genmatch): ... here.
1690 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1691 Chung-Ju Wu <jasonwucj@gmail.com>
1693 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1694 * config/nds32/nds32-graywolf.md: New file.
1695 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1696 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1698 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1699 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1700 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1701 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1702 * config/nds32/pipelines.md: Include n15 settings.
1704 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1705 Chung-Ju Wu <jasonwucj@gmail.com>
1707 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1708 * config/nds32/nds32-n13.md: New file.
1709 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1710 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1712 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1713 * config/nds32/nds32.md (pipeline_model): Add n13.
1714 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1715 * config/nds32/pipelines.md: Include n13 settings.
1717 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1718 Chung-Ju Wu <jasonwucj@gmail.com>
1720 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1721 * config/nds32/nds32-n10.md: New file.
1722 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1723 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1725 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1726 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1727 * config/nds32/nds32.md (pipeline_model): Add n10.
1728 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1729 * config/nds32/pipelines.md: Include n10 settings.
1731 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1732 Kito Cheng <kito.cheng@gmail.com>
1733 Chung-Ju Wu <jasonwucj@gmail.com>
1735 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1736 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1737 Add enum values for DSP extension instructions.
1738 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1740 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1741 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1743 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1744 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1745 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1746 * config/nds32/nds32-intrinsic.md: Likewise.
1747 * config/nds32/nds32_intrinsic.h: Likewise.
1748 * config/nds32/nds32-md-auxiliary.c: Likewise.
1749 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1750 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1751 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1752 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1753 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1754 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1755 TYPE_DMAC in switch statement.
1756 * config/nds32/nds32.c: New checking and implementation for DSP
1757 extension instructions.
1758 * config/nds32/nds32.h: Likewise.
1759 * config/nds32/nds32.md: Likewise.
1760 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1761 * config/nds32/predicates.md: Implement new predicates for DSP
1764 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1766 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1767 Reformat alternatives and attributes so it is easier to identify
1768 which constraints/attributes go with which instruction.
1769 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1770 (mov<mode>_softfloat32, FMOVE64): Likewise.
1771 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1772 (mov<mode>_softfloat64, FMOVE64): Likewise.
1774 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1776 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1778 (Basic PowerPC Built-in Functions): The new name of the
1779 subsection previously known as "PowerPC Built-in Functions".
1780 (Basic PowerPC Built-in Functions Available on all Configurations):
1782 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1783 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1784 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1785 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1787 2018-05-11 Martin Jambor <mjambor@suse.cz>
1790 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1793 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1796 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1798 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1800 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1801 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1802 (ix86_handle_option): Handle -mwaitpkg.
1803 * config.gcc: New header.
1804 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1805 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1806 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1808 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1809 OPTION_MASK_ISA_WAITPKG.
1810 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1811 (ix86_option_override_internal): Add PTA_WAITPKG.
1812 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1813 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1814 IX86_BUILTIN_TPAUSE.
1815 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1816 __builtin_ia32_umwait and __builtin_ia32_tpause.
1817 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1818 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1819 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1820 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1821 UNSPECV_TPAUSE): New.
1822 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1823 * config/i386/i386.opt: Add -mwaitpkg.
1824 * config/i386/waitpkgintrin.h: New file.
1825 * config/i386/x86intrin.h: New header.
1826 * doc/invoke.texi: Add -mwaitpkg.
1828 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1831 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1833 (cortex-m0): Use armv6s-m isa.
1834 (cortex-m0plus): Likewise.
1835 (cortex-m1): Likewise.
1836 (cortex-m0.small-multiply): Likewise.
1837 (cortex-m0plus.small-multiply): Likewise.
1838 (cortex-m1.small-multiply): Likewise.
1840 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1841 Jakub Jelinek <jakub@redhat.com>
1843 PR tree-optimization/85692
1844 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1845 source permute as well.
1847 2018-05-11 Martin Liska <mliska@suse.cz>
1850 * doc/extend.texi: Document LLVM style format for no_sanitize
1853 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1855 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1856 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1857 (mode_supports_vsx_dform_quad): Likewise.
1858 (mode_supports_vmx_dform): Move these functions to be next to the
1859 other mode_supports functions.
1860 (mode_supports_dq_form): Likewise.
1861 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1862 mode_supports_dq_form.
1863 (reg_offset_addressing_ok_p): Likewise.
1864 (offsettable_ok_by_alignment): Likewise.
1865 (rs6000_legitimate_offset_address_p): Likewise.
1866 (legitimate_lo_sum_address_p): Likewise.
1867 (rs6000_legitimize_address): Likewise.
1868 (rs6000_legitimize_reload_address): Likewise.
1869 (rs6000_secondary_reload_inner): Likewise.
1870 (rs6000_preferred_reload_class): Likewise.
1871 (rs6000_output_move_128bit): Likewise.
1873 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1875 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1876 Generate SImode target register for null target.
1877 <case IX86_BUILTIN_XGETBV>: Ditto.
1878 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1879 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1881 2018-05-10 Carl Love <cel@us.ibm.com>
1883 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1884 dcbtt and dcbtstt if operands[2] is 0.
1886 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1889 * config/i386/sse.md (usadv64qi): New expander.
1891 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1893 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1894 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1895 -maltivec=be support.
1896 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1897 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1898 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1899 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1900 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1901 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1902 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1903 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1904 altivec_vsumsws): Adjust.
1905 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1906 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1907 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1909 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1910 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1911 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1912 (altivec_lve<VI_char>x): Delete expand.
1913 (*altivec_lve<VI_char>x_internal): Rename to...
1914 (altivec_lve<VI_char>x): ... this.
1915 (altivec_lvxl_<mode>): Delete expand.
1916 (*altivec_lvxl_<mode>_internal): Rename to ...
1917 (altivec_lvxl_<mode>): ... this.
1918 (altivec_stvxl_<mode>): Delete expand.
1919 (*altivec_stvxl_<mode>_internal): Rename to ...
1920 (altivec_stvxl_<mode>): ... this.
1921 (altivec_stve<VI_char>x): Delete expand.
1922 (*altivec_stve<VI_char>x_internal): Rename to ...
1923 (altivec_stve<VI_char>x): ... this.
1924 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1925 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1926 reduc_plus_scal_<mode>): Adjust.
1927 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1929 (rs6000_cpu_cpp_builtins): Adjust.
1930 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1931 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1932 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1933 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1934 -maltivec=be support.
1935 (rs6000_split_vec_extract_var): Adjust.
1936 (rs6000_split_v4si_init): Adjust.
1937 (swap_selector_for_mode): Delete.
1938 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1939 altivec_expand_stvex_be): Delete.
1940 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1941 -maltivec=be support.
1942 (rs6000_gimple_fold_builtin): Ditto.
1943 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1945 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1946 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1947 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1948 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1949 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1950 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1951 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1952 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1953 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1954 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1955 anonymous split): Adjust.
1956 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1957 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1959 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1961 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1962 when --with-gxx-include-dir is also specified.
1963 * configure: Regenerate.
1965 2018-05-09 Jim Wilson <jimw@sifive.com>
1968 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1969 * config/riscv/t-withmultilib: New.
1970 * config/riscv/withmultilib.h: New.
1971 * doc/install.texi: Document RISC-V --with-multilib-list support.
1973 2018-05-09 Richard Biener <rguenther@suse.de>
1975 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1977 (vect_bb_vectorization_profitable_p): Adjust. Compute
1978 actual scalar cost using the cost vector and the add_stmt_cost
1981 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1983 PR rtl-optimization/85645
1984 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1985 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1987 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1989 PR rtl-optimization/85645
1990 * shrink-wrap.c (spread_components): Return a boolean saying if
1991 anything was changed.
1992 (try_shrink_wrapping_separate): Iterate spread_components until
1993 nothing changes anymore.
1995 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1997 PR rtl-optimization/85645
1998 * regrename.c (build_def_use): Also kill the chains that include the
1999 destination of a REG_CFA_REGISTER note.
2001 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2003 PR rtl-optimization/85645
2004 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2005 insn that has a REG_CFA_REGISTER note.
2007 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2009 * cfgexpand.c (expand_clobber): New function.
2010 (expand_gimple_stmt_1): Use it.
2011 * tree-vect-stmts.c (vect_clobber_variable): New function,
2013 (vectorizable_simd_clone_call): ...here.
2014 (vectorizable_store): Emit a clobber either side of an
2015 IFN_STORE_LANES sequence.
2016 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2018 2018-05-09 Tom de Vries <tom@codesourcery.com>
2021 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2022 (define_insn "trap_if_false"): Add exit after trap.
2024 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2026 PR rtl-optimization/85638
2027 * bb-reorder.c: Include common/common-target.h.
2028 (create_forwarder_block): New function extracted from...
2029 (fix_up_crossing_landing_pad): ...here. Rename into...
2030 (dw2_fix_up_crossing_landing_pad): ...this.
2031 (sjlj_fix_up_crossing_landing_pad): New function.
2032 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2033 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2034 from both partitions and exit the loop after one iteration.
2036 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2039 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2041 (Basic PowerPC Built-in Functions): The new name of the
2042 subsection previously known as "PowerPC Built-in Functions".
2043 (Basic PowerPC Built-in Functions Available on all Configurations):
2045 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2047 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2049 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2051 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2054 2018-05-08 Jim Wilson <jimw@sifive.com>
2056 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2057 (LD_EMUL_SUFFIX): New.
2058 (LINK_SPEC): Use it.
2060 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2062 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2064 (Basic PowerPC Built-in Functions): The new name of the
2065 subsection previously known as "PowerPC Built-in Functions".
2066 (Basic PowerPC Built-in Functions Available on all Configurations):
2068 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2070 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2072 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2074 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2077 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2080 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2081 after cmpelim optimization.
2083 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2085 * config.gcc: Support "goldmont".
2086 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2087 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2089 * config/i386/i386.c (m_GOLDMONT): Define.
2090 (processor_target_table): Add "goldmont".
2091 (PTA_GOLDMONT): Define.
2092 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2093 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2094 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2095 (fold_builtin_cpu): Add "goldmont".
2096 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2097 (ix86_option_override_internal): Add "goldmont".
2098 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2099 (processor_type): Add PROCESSOR_GOLDMONT.
2100 * config/i386/i386.md: Add CPU "glm".
2101 * config/i386/glm.md: New file.
2102 * config/i386/x86-tune.def: Add m_GOLDMONT.
2103 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2105 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2108 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2110 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2111 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2112 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2115 * config/i386/i386.c (ix86_fold_builtin): Handle
2116 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2119 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2120 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2122 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2125 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2126 (check_arch): Likewise.
2127 (check_fpu): Return the result rather than printing it.
2128 (end arch): Fix operator precedence.
2129 (end cpu): Likewise.
2130 (END): Print the result from check_fpu.
2132 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2133 Alan Hayward <alan.hayward@arm.com>
2134 David Sherwood <david.sherwood@arm.com>
2136 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2137 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2138 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2139 (*fcmuo<mode>_and): New patterns.
2141 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2143 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2144 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2145 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2146 (cmp_op, sve_imm_con): New code attributes.
2147 (SVE_COND_INT_CMP, imm_con): Delete.
2148 (cmp_op): Remove above unspecs from int attribute.
2149 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2151 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2152 comparison-specific unspecs.
2153 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2154 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2155 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2156 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2157 (*vec_fcm<cmp_op><mode>): Rename to...
2158 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2159 (*vec_fcmuo<mode>): Rename to...
2160 (*fcmuo<mode>): ...this and adjust likewise.
2161 (*pred_fcm<cmp_op><mode>): New pattern.
2162 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2163 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2165 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2167 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2168 (aarch64_emit_sve_predicated_cond): New function.
2169 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2170 (aarch64_emit_unspec_cond_or): Replace with...
2171 (aarch64_emit_sve_or_conds): ...this new function. Use
2172 aarch64_emit_sve_ptrue_op for the individual comparisons and
2173 aarch64_emit_binop to OR them together.
2174 (aarch64_emit_inverted_unspec_cond): Replace with...
2175 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2176 aarch64_emit_sve_ptrue_op for the comparison and
2177 aarch64_emit_unop to invert the result.
2178 (aarch64_expand_sve_vec_cmp_float): Update after the above
2179 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2181 2018-05-07 Nathan Sidwell <nathan@acm.org>
2183 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2184 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2185 (Backwards Compatibility): Likewise.
2187 2018-05-07 Luis Machado <luis.machado@linaro.org>
2191 2018-05-07 Luis Machado <luis.machado@linaro.org>
2193 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2194 <prefetch_dynamic_strides>: New const bool field.
2195 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2196 prefetch_dynamic_strides.
2197 (exynosm1_prefetch_tune): Likewise.
2198 (thunderxt88_prefetch_tune): Likewise.
2199 (thunderx_prefetch_tune): Likewise.
2200 (thunderx2t99_prefetch_tune): Likewise.
2201 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2202 (aarch64_override_options_internal): Update to set
2203 PARAM_PREFETCH_DYNAMIC_STRIDES.
2204 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2205 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2206 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2207 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2208 prefetch-dynamic-strides setting.
2210 2018-05-07 Luis Machado <luis.machado@linaro.org>
2212 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2213 <minimum_stride>: New const int field.
2214 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2215 minimum_stride field.
2216 (exynosm1_prefetch_tune): Likewise.
2217 (thunderxt88_prefetch_tune): Likewise.
2218 (thunderx_prefetch_tune): Likewise.
2219 (thunderx2t99_prefetch_tune): Likewise.
2220 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2221 (aarch64_override_options_internal): Update to set
2222 PARAM_PREFETCH_MINIMUM_STRIDE.
2223 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2224 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2225 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2226 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2227 stride is constant and is below the minimum stride threshold.
2229 2018-05-07 Luis Machado <luis.machado@linaro.org>
2231 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2234 2018-05-07 Luis Machado <luis.machado@linaro.org>
2236 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2237 <prefetch_dynamic_strides>: New const bool field.
2238 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2239 prefetch_dynamic_strides.
2240 (exynosm1_prefetch_tune): Likewise.
2241 (thunderxt88_prefetch_tune): Likewise.
2242 (thunderx_prefetch_tune): Likewise.
2243 (thunderx2t99_prefetch_tune): Likewise.
2244 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2245 (aarch64_override_options_internal): Update to set
2246 PARAM_PREFETCH_DYNAMIC_STRIDES.
2247 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2248 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2249 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2250 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2251 prefetch-dynamic-strides setting.
2253 2018-05-07 Luis Machado <luis.machado@linaro.org>
2255 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2256 <minimum_stride>: New const int field.
2257 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2258 minimum_stride field.
2259 (exynosm1_prefetch_tune): Likewise.
2260 (thunderxt88_prefetch_tune): Likewise.
2261 (thunderx_prefetch_tune): Likewise.
2262 (thunderx2t99_prefetch_tune): Likewise.
2263 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2264 (aarch64_override_options_internal): Update to set
2265 PARAM_PREFETCH_MINIMUM_STRIDE.
2266 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2267 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2268 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2269 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2270 stride is constant and is below the minimum stride threshold.
2272 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2275 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2276 the type is addressable. Don't force op into register if it has
2279 2018-05-05 Roland McGrath <mcgrathr@google.com>
2282 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2283 any section for which we don't know a specific type it should have,
2284 regardless of name. Previously this was done only for the exact
2285 names ".init_array", ".fini_array", and ".preinit_array".
2286 (default_elf_asm_named_section): Add comment about
2287 relationship with default_section_type_flags and SECTION_NOTYPE.
2288 (get_section): Don't consider it a type conflict if one side has
2289 SECTION_NOTYPE and the other doesn't, as long as neither has the
2290 SECTION_BSS et al used in the default_section_type_flags logic.
2292 2018-05-05 Tom de Vries <tom@codesourcery.com>
2295 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2296 (workaround_barsyncs): New function.
2297 (nvptx_reorg): Use workaround_barsyncs.
2298 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2299 (define_expand "nvptx_membar_cta"): New define_expand.
2300 (define_insn "*nvptx_membar_cta"): New insn.
2302 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2304 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2305 To improve optimization opportunities.
2306 * builtin-types.def: The new needed builtin types for the above.
2308 2018-05-04 Richard Biener <rguenther@suse.de>
2310 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2311 * gimple-ssa-store-merging.c
2312 (imm_store_chain_info::output_merged_store): Remove redundant create,
2313 release split_store vector contents on failure.
2314 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2315 scalar stmt vector on cache hit.
2317 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2319 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2321 * config.gcc (powerpc-xilinx-eabi*): Remove.
2322 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2324 (fusion_addis_mem_combo_load): Ditto.
2325 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2327 (rs6000_cpu_cpp_builtins): Ditto.
2328 * config/rs6000/rs6000-linux.c
2329 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2330 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2331 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2333 (rs6000_setup_reg_addr_masks): Ditto.
2334 (rs6000_init_hard_regno_mode_ok): Ditto.
2335 (rs6000_option_override_internal): Ditto.
2336 (legitimate_lo_sum_address_p): Ditto.
2337 (rs6000_legitimize_address): Ditto.
2338 (rs6000_legitimize_reload_address): Ditto.
2339 (rs6000_legitimate_address_p): Ditto.
2340 (abi_v4_pass_in_fpr): Ditto.
2341 (setup_incoming_varargs): Ditto.
2342 (rs6000_gimplify_va_arg): Ditto.
2343 (rs6000_split_multireg_move): Ditto.
2344 (rs6000_savres_strategy): Ditto.
2345 (rs6000_emit_prologue_components): Ditto.
2346 (rs6000_emit_epilogue_components): Ditto.
2347 (rs6000_emit_prologue): Ditto.
2348 (rs6000_emit_epilogue): Ditto.
2349 (rs6000_elf_file_end): Ditto.
2350 (rs6000_function_value): Ditto.
2351 (rs6000_libcall_value): Ditto.
2352 * config/rs6000/rs6000.h: Ditto.
2353 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2354 (TARGET_MINMAX): ... this. New.
2355 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2356 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2357 (*movsi_internal1_single): Delete.
2358 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2359 mfpu=, mxilinx-fpu): Delete.
2360 * config/rs6000/singlefp.h: Delete.
2361 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2362 * config/rs6000/t-rs6000: Ditto.
2363 * config/rs6000/t-xilinx: Delete.
2364 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2365 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2366 (VStype_simple): Delete.
2367 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2368 * config/rs6000/xfpu.h: Delete.
2369 * config/rs6000/xfpu.md: Delete.
2370 * config/rs6000/xilinx.h: Delete.
2371 * config/rs6000/xilinx.opt: Delete.
2372 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2373 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2375 2018-05-04 Tom de Vries <tom@codesourcery.com>
2378 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2381 2018-05-04 Richard Biener <rguenther@suse.de>
2384 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2385 (expand_complex_div_wide): Likewise.
2386 (expand_complex_operations_1): Likewise.
2387 (expand_complex_libcall): Preserve EH info of the original stmt.
2388 (tree_lower_complex): Handle removed blocks.
2389 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2390 on complex multiplication and division libcall builtins.
2392 2018-05-04 Richard Biener <rguenther@suse.de>
2395 * fold-const.c (negate_expr_p): Restrict negation of operand
2396 zero of a division to when we know that can happen without
2398 (fold_negate_expr_1): Likewise.
2400 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2403 * real.h (real_nextafter): Declare.
2404 * real.c (real_nextafter): New function.
2405 * fold-const-call.c (fold_const_nextafter): New function.
2406 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2407 CASE_CFN_NEXTTOWARD.
2408 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2409 even when arg1_mode is different from arg0_mode.
2411 2018-05-03 Nathan Sidwell <nathan@acm.org>
2413 * doc/extend.texi (Deprecated Features): Remove
2415 (Backwards Compatibility): Likewise.
2416 * doc/invoke.texi (C++ Language Options): Likewise.
2417 (C++ Dialect Options): Likewise.
2419 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2422 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2423 _mm512_mask_mullox_epi64): New intrinsics.
2425 2018-05-03 Tom de Vries <tom@codesourcery.com>
2428 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2429 dump files): Add offload-tree.
2431 2018-05-03 Richard Biener <rguenther@suse.de>
2433 PR tree-optimization/85615
2434 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2435 to loops not nested in BBs loop father to avoid creating multi-entry
2438 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2440 PR tree-optimization/70291
2441 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2442 arguments. Change return type to tree. Emit libcall as a new
2443 statement rather than replacing existing one when inplace_p is true.
2444 (expand_complex_multiplication_components): New function.
2445 (expand_complex_multiplication): Expand floating-point complex
2446 multiplication using the above.
2447 (expand_complex_division): Rename inner_type parameter to type.
2448 Update expand_complex_libcall call-site.
2449 (expand_complex_operations_1): Update expand_complex_multiplication
2450 and expand_complex_division call-sites.
2452 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2455 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2456 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2457 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2458 the highest significant bit of the shift count mask is clear. In
2459 check whether and[sq]i3 is needed verify that all significant bits
2460 of the shift count other than the highest are set.
2462 2018-05-02 Tom de Vries <tom@codesourcery.com>
2465 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2466 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2467 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2468 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2469 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2470 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2471 __builtin_goacc_parlevel_size.
2473 2018-05-02 Richard Biener <rguenther@suse.de>
2475 PR tree-optimization/85597
2476 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2477 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2480 2018-05-02 Tom de Vries <tom@codesourcery.com>
2483 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2484 dump files): Add ltrans-tree.
2486 2018-05-02 Tom de Vries <tom@codesourcery.com>
2489 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2490 dump files): Add wpa-ipa.
2492 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2494 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2495 powerpc*-*-linux*paired* target.
2496 * config/rs6000/750cl.h: Delete.
2497 * config/rs6000/paired.h: Delete.
2498 * config/rs6000/paired.md: Delete.
2499 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2501 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2502 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2503 comment. Remove paired float support.
2504 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2505 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2507 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2508 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2510 * config/rs6000/rs6000.c: Remove paired float support.
2511 (paired_expand_vector_init, paired_expand_vector_move,
2512 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2513 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2514 paired_expand_builtin, paired_expand_predicate_builtin,
2515 paired_init_builtins): Delete.
2516 * config/rs6000/rs6000.h: Remove paired float support.
2517 * config/rs6000/rs6000.md: Remove paired float support.
2518 (move_from_CR_ov_bit): Delete.
2519 * config/rs6000/rs6000.opt (mpaired): Delete.
2520 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2521 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2523 2018-05-02 Richard Biener <rguenther@suse.de>
2526 * gimplify.c (gimplify_save_expr): When in SSA form allow
2527 SAVE_EXPRs to compute to SSA vars.
2529 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2532 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2533 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2534 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2535 clobber operands[2], instead use a new pseudo. Formatting fixes.
2537 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2539 PR tree-optimization/85586
2540 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2541 exit early for statements in the same group if the accesses are
2544 2018-05-02 Tom de Vries <tom@codesourcery.com>
2547 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2550 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2552 PR tree-optimization/85143
2553 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2555 2018-05-01 Tom de Vries <tom@codesourcery.com>
2558 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2559 not found" error message.
2561 2018-05-01 Tom de Vries <tom@codesourcery.com>
2564 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2565 * vec.c (test_ordered_remove_if): New function.
2566 (vec_c_tests): Call test_ordered_remove_if.
2567 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2568 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2569 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2570 VEC_ORDERED_REMOVE_IF.
2572 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2574 PR tree-optimization/82665
2575 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2576 pointer subtraction where arguments come from a memchr call.
2578 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2580 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2581 --push-state --as-needed and --pop-state instead of --as-needed and
2582 --no-as-needed if ld supports it.
2583 * configure: Regenerated.
2586 * doc/install.texi2html: Replace _002d with - and _002a with * in
2587 generated html files using sed.
2589 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2592 * gcc-rich-location.c (blank_line_before_p): New function.
2593 (use_new_line): New function.
2594 (gcc_rich_location::add_fixit_insert_formatted): New function.
2595 * gcc-rich-location.h
2596 (gcc_rich_location::add_fixit_insert_formatted): New function.
2598 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2600 * selftest.c (assert_streq): Rename "expected" and "actual" to
2601 "val1" and "val2". Extend NULL-handling to cover both inputs
2602 symmetrically, while still requiring both to be non-NULL for a pass.
2603 * selftest.h (assert_streq): Rename "expected" and "actual" to
2605 (ASSERT_EQ): Likewise.
2606 (ASSERT_EQ_AT): Likewise.
2607 (ASSERT_KNOWN_EQ): Likewise.
2608 (ASSERT_KNOWN_EQ_AT): Likewise.
2609 (ASSERT_NE): Likewise.
2610 (ASSERT_MAYBE_NE): Likewise.
2611 (ASSERT_MAYBE_NE_AT): Likewise.
2612 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2613 the assertion to pass.
2614 (ASSERT_STREQ_AT): Likewise.
2616 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2618 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2619 interaction with -pie.
2621 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2623 * selftest.h: Fix alphabetization of per-source-file selftest
2626 2018-04-30 Jason Merrill <jason@redhat.com>
2628 PR c++/61982 - dead stores to destroyed objects.
2629 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2632 2018-04-30 Jason Merrill <jason@redhat.com>
2634 * tree.c (build_clobber): New.
2635 * tree.h: Declare it.
2636 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2638 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2640 * diagnostic-show-locus.c (layout::layout): Update for
2641 location_get_source_line returning a char_span.
2642 (struct char_span): Move to input.h.
2643 (struct correction): Update for fields in char_span becoming
2645 (struct source_line): Update for location_get_source_line
2646 returning a char_span.
2647 (layout::print_line): Likewise.
2648 * edit-context.c (edited_file::print_content): Likewise.
2649 (edited_file::print_diff_hunk): Likewise.
2650 (edited_file::print_run_of_changed_lines): Likewise.
2651 (edited_file::get_num_lines): Likewise.
2652 (edited_line::edited_line): Likewise.
2653 * final.c (asm_show_source): Likewise.
2654 * input.c (location_get_source_line): Convert return type
2655 from const char * to char_span, losing the final "line_len"
2657 (dump_location_info): Update for the above.
2658 (get_substring_ranges_for_loc): Likewise. Use a char_span
2659 when handling the literal within the line.
2660 (test_reading_source_line): Update for location_get_source_line
2661 returning a char_span.
2662 * input.h (class char_span): Move here from
2663 diagnostic-show-locus.c, converting from a struct to a class.
2664 Make data members private.
2665 (char_span::operator bool): New.
2666 (char_span::length): New.
2667 (char_span::get_buffer): New.
2668 (char_span::operator[]): New.
2669 (char_span::subspan): Make const.
2670 (char_span::xstrdup): New.
2671 (location_get_source_line): Convert return type from const char *
2672 to char_span, losing the final "line_size" param.
2674 2018-04-30 Jan Hubicka <jh@suse.cz>
2676 * lto-wrapper.c (ltrans_priorities): New static var.
2677 (cmp_priority): New.
2678 (run_gcc): Read priorities and if doing parallel build order
2679 the Makefile by them.
2681 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2683 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2685 2018-04-30 Richard Biener <rguenther@suse.de>
2687 * tree-cfg.c (verify_address): Remove base argument, add
2688 flag whether to check TREE_ADDRESSABLE and do that.
2689 (verify_expr): Remove.
2690 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2691 (verify_gimple_assign_single): Likewise.
2692 (verify_gimple_switch): Likewise.
2693 (verify_expr_location_1): Dereference tp once. Add (disabled)
2694 piece from verify_expr.
2695 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2697 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2699 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2701 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2703 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2704 (small_data_pattern): Likewise.
2705 (arc_rewrite_small_data): Likewise.
2706 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2707 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2708 (get_symbol_alignment): New function.
2709 (legitimate_small_data_address_p): Likewise.
2710 (legitimate_scaled_address): Update, call
2711 legitimate_small_data_address_p.
2712 (output_sdata): New static variable.
2713 (arc_print_operand): Update how we handle small data operands.
2714 (arc_print_operand_address): Likewise.
2715 (arc_legitimate_address_p): Update, use
2716 legitimate_small_data_address_p.
2717 (arc_rewrite_small_data_p): Remove.
2718 (arc_rewrite_small_data_1): Likewise.
2719 (arc_rewrite_small_data): Likewise.
2720 (small_data_pattern): Likewise.
2721 (compact_sda_memory_operand): Update to use
2722 legitimate_small_data_address_p and get_symbol_alignment.
2723 (prepare_move_operands): Don't rewite sdata pattern.
2724 (prepare_extend_operands): Remove.
2725 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2727 (zero_extendqisi2): Likewise.
2728 (zero_extendhisi2): Likewise.
2729 (extendqihi2): Likewise.
2730 (extendqisi2): Likewise.
2731 (extendhisi2): Likewise.
2735 * config/arc/constraints.md (Usd): Change it to memory constraint.
2737 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2739 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2740 as source of std instructions.
2741 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2742 allow 6-bit constants as source for store instructions.
2743 (movdi_insn): Update instruction pattern to allow 6-bit constants
2744 as source for store instructions.
2746 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2748 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2750 2018-04-30 Nathan Sidwell <nathan@acm.org>
2751 Sandra Loosemore <sandra@codesourcery.com>
2753 * dumpfile.c (dump_open): Allow '-' for stdout.
2754 * doc/invoke.texi (Developer Options): Document dump filename
2755 determination early. Document stdin/stdout selection.
2757 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2759 Microblaze Target: PIC data text relative
2761 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2762 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2764 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2765 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2766 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2767 New addressing mode for data-text relative position indepenedent code.
2768 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2769 'ADDRESS_SYMBOLIC_TXT_REL'.
2770 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2771 (microblaze_legitimate_pic_operand): Exclude function calls from
2772 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2773 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2775 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2776 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2777 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2778 for 'address + offset'.
2779 (microblaze_expand_prologue): Add new function prologue call for
2781 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2782 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2783 table in case of TARGET_PIC_DATA_TEXT_REL.
2784 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2785 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2786 Add new macros 'UNSPEC_TEXT',
2787 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2788 + exclude function calls from 'UNSPEC_PLT' in case of data text
2790 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2791 new target hook for generating address diff vector tables in case of
2793 * doc/tm.texi : Regenerate.
2794 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2795 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2796 of addr diff vector generation.
2797 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2798 target hook definition.
2799 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2800 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2801 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2802 option for data text relative.
2804 2018-04-30 Richard Biener <rguenther@suse.de>
2806 * tree-chrec.h (evolution_function_is_constant_p): Remove
2808 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2810 2018-04-30 Richard Biener <rguenther@suse.de>
2813 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2815 2018-04-30 Richard Biener <rguenther@suse.de>
2817 PR tree-optimization/28364
2818 PR tree-optimization/85275
2819 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2820 copying first exit test.
2822 2018-04-28 Mark Wielaard <mark@klomp.org>
2824 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2826 (dwarf_AT): Handle DW_AT_addr_base.
2827 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2829 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2832 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2833 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2834 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2835 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2837 2018-04-28 Richard Biener <rguenther@suse.de>
2839 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2840 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2841 to reflect use. Only add interesting stmts.
2843 2018-04-27 Martin Jambor <mjambor@suse.cz>
2846 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2847 the jump function allows for passing through aggregate values.
2849 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2851 * input.h (in_system_header_at): Convert from macro to inline
2853 (from_macro_expansion_at): Likewise.
2854 (from_macro_definition_at): Likewise.
2856 2018-04-27 Jeff Law <law@redhat.com>
2858 * config.gcc: Mark tile* targets as deprecated/obsolete.
2860 2018-04-27 Richard Biener <rguenther@suse.de>
2862 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2865 2018-04-27 Richard Biener <rguenther@suse.de>
2867 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2869 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2871 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2872 with Yd constraint. Set "preferred_for_speed" attribute from
2873 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2875 (*movdi_internal): Ditto.
2876 (movti_interunit splitters): Remove
2877 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2878 (movdi_interunit splitters): Ditto.
2879 * config/i386/constraints.md (Ye): Remove.
2880 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2882 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2885 * config/aarch64/constraints.md (Usg): Limit to 31.
2888 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2890 PR tree-optimization/85529
2891 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2892 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2893 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2894 zero extension or masking of the MSB bit.
2895 (optimize_range_tests): Add FIRST_BB argument, pass it through
2896 to optimize_range_tests_var_bound.
2897 (maybe_optimize_range_tests, reassociate_bb): Adjust
2898 optimize_range_tests callers.
2900 2018-04-26 Richard Biener <rguenther@suse.de>
2901 Jakub Jelinek <jakub@redhat.com>
2903 * cgraph.h (symbol_table): Just declare debug method here.
2904 * symtab.c (symbol_table::debug): Define.
2906 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2908 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2910 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2912 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2913 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2914 (*movdi_internal): Substitute Yi and Yj constraint with x
2915 and Ym and Yn constraint with y constraint. Update "isa"
2916 attribute and set "preferred_for_speed" attribute from
2917 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2918 (*movsi_internal): Ditto.
2919 (*movdf_internal): Ditto.
2920 (*movsf_internal): Ditto.
2921 (*zero_extendsidi2): Ditto.
2922 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2923 (sse2_loadld): Ditto.
2924 (*vec_extract<ssevecmodelower>_0): Ditto.
2925 (*vec_extractv4si_0_zext_sse4): Ditto.
2926 (vec_concatv2di): Ditto.
2927 (*vec_dup<mode>): Ditto.
2928 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2929 * config/i386/constraints.md (Yi): Remove.
2934 2018-04-26 Nathan Sidwell <nathan@acm.org>
2936 * dumpfile.c (dump_open): New.
2937 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2938 (dump_finish): Detect stdio/stderr by value not name.
2940 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2942 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2944 2018-04-26 Tom de Vries <tom@codesourcery.com>
2947 * config/nvptx/nvptx.c (verify_neutering_jumps)
2948 (verify_neutering_labels): New function
2949 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2951 2018-04-26 Tom de Vries <tom@codesourcery.com>
2954 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2955 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2958 2018-04-26 Richard Biener <rguenther@suse.de>
2959 Tom de Vries <tom@codesourcery.com>
2962 * lto-streamer-out.c (output_function): Fixup loops if required to match
2963 discovery done in the reader.
2965 2018-04-26 Richard Biener <rguenther@suse.de>
2967 PR tree-optimization/85116
2968 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2969 have a loop exit from the single latch predecessor. Remove
2970 case of header with just condition.
2971 (ch_base::copy_headers): Exclude infinite loops from any
2973 (pass_ch::execute): Record exits.
2975 2018-04-26 Richard Biener <rguenther@suse.de>
2977 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2978 prologue cost vector and pass it to vect_get_load_cost.
2979 (vect_get_peeling_costs_all_drs): Likewise.
2980 (vect_peeling_hash_get_lowest_cost): Likewise.
2981 (vect_enhance_data_refs_alignment): Likewise.
2983 2018-04-26 Richard Biener <rguenther@suse.de>
2986 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2987 checking of integer<->pointer conversions.
2988 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2989 sign-/zero-extending pointer types.
2990 (expand_omp_for_static_chunk): Likewise.
2992 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2993 Jean Lee <xiaoyur347@gmail.com>
2995 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2996 (TARGET_ASAN_SHADOW_OFFSET): Define.
2997 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2998 true for -fsanitize=address.
3000 2018-04-25 Mark Wielaard <mark@klomp.org>
3002 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3005 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3007 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3008 than "alu", remove explicit "memory" and "imm_disp" attributes.
3009 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3012 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3013 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3014 gen_lowpart_no_emit.
3016 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3019 * config/i386/i386.c (ix86_expand_builtin): Change memory
3020 operand to XI, extend p0 to Pmode.
3021 * config/i386/i386.md: Change unspec volatile and operand
3022 1 mode to XI, change operand 0 mode to P.
3024 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3026 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3027 GET_MODE_MASK before any checking.
3028 (nds32_can_use_bset_p): Likewise.
3029 (nds32_can_use_btgl_p): Likewise.
3031 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3033 * config/nds32/nds32-doubleword.md: New define_split pattern for
3034 illegal register number.
3036 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3038 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3040 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3042 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3044 2018-04-25 Richard Biener <rguenther@suse.de>
3046 * lto-streamer.h (LTO_major_version): Bump to 8.
3048 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3050 * BASE-VER: Set to 9.0.0.
3052 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3054 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3055 in __abskf2 and __powikf2.
3057 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3060 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3061 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3062 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3063 Use the above on operand 2. Reindent.
3064 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3066 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3069 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3071 * config/i386/i386.opt (mcet): Removed.
3072 * doc/install.texi: Remove -mcet documentation.
3073 * doc/invoke.texi: Likewise.
3075 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3078 * doc/install.texi: Remove -mcet from bootstrap-cet.
3080 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3083 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3084 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3088 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3089 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3090 containing a CONST_VECTOR.
3092 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3094 * doc/install.texi: Update newlib dependency for nvptx.
3096 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3099 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3100 instead of INTVAL when shifting x left.
3102 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3104 PR tree-optimization/85478
3105 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3106 vect_grouped_store_supported for single element vectors.
3108 2018-04-24 Richard Biener <rguenther@suse.de>
3111 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3112 load cost increase to the case of non-constant step.
3114 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3117 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3118 destination if any_malformed_asm.
3120 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3123 * expr.c (store_field): In the bitfield case, if the value comes from
3124 a function call and is returned in registers by means of a PARALLEL,
3125 do not change the mode of the temporary unless BLKmode and VOIDmode.
3127 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3129 PR rtl-optimization/85423
3130 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3131 dependencies to debug insns when the previous insn is non-debug.
3133 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3135 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3136 enums into a single definition.
3137 (fls): Fix predicates and printing.
3140 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3142 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3143 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3144 and short u6 immediate.
3145 (check_if_valid_sleep_operand): Remove.
3146 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3148 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3150 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3151 flag_always_save_lp condition.
3152 * config/nds32/nds32.opt (malways-save-lp): New option.
3154 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3156 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3157 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3158 * config/nds32/nds32.h
3159 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3160 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3162 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3164 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3165 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3167 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3168 Chung-Ju Wu <jasonwucj@gmail.com>
3170 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3172 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3173 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3175 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3177 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3179 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3181 * config/nds32/nds32-protos.h (nds32_data_alignment,
3182 nds32_local_alignment): Declare.
3183 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3184 nds32_local_alignment): New functions.
3185 (TARGET_CONSTANT_ALIGNMENT): Define.
3186 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3188 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3190 * config/nds32/nds32.c
3191 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3192 (TARGET_MODES_TIEABLE_P): Likewise.
3194 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3196 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3199 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3200 Chung-Ju Wu <jasonwucj@gmail.com>
3202 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3203 for unaligned access.
3204 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3206 * config/nds32/nds32-intrinsic.md: Likewise.
3207 * config/nds32/nds32_intrinsic.h: Likewise.
3208 * config/nds32/nds32.h (nds32_builtins): Likewise.
3209 * config/nds32/nds32.opt (munaligned-access): New option.
3210 * config/nds32/nds32.c (nds32_asm_file_start): Display
3211 flag_unaligned_access status.
3213 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3215 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3216 -mno-relax is present.
3217 * config/riscv/linux.h (LINK_SPEC): Ditto.
3219 2018-04-20 Martin Sebor <msebor@redhat.com>
3222 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3224 (gimple_fold_builtin_stxcpy_chk): Same.
3225 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3227 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3230 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3231 __powikf2 when long double is IEEE 128-bit.
3233 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3235 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3236 step to make sure stack always aligned.
3238 2018-04-20 Carl Love <cel@us.ibm.com>
3241 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3242 size check for arg0.
3244 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3245 Tom de Vries <tom@codesourcery.com>
3248 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3249 Emit insns for calls too.
3250 (nvptx_find_par): Always look for worker-level predecessor insn.
3251 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3253 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3254 (nvptx_process_pars): Propagate frames for calls.
3256 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3259 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3261 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3262 (ix86_handle_option): Don't handle OPT_mibt.
3263 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3265 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3267 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3268 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3269 (ix86_target_macros): Define __CET__ with flag_cf_protection
3270 for -fcf-protection.
3271 * config/i386/i386.c (isa2_opts): Remove -mibt.
3272 * config/i386/i386.h (TARGET_IBT): Removed.
3273 (TARGET_IBT_P): Likewise.
3274 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3275 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3276 * config/i386/i386.opt (mcet): Update help message.
3279 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3280 -mcet as an alias for -mshstk.
3282 2018-04-20 Richard Biener <rguenther@suse.de>
3285 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3286 complexity by forcing a single use of the multiply operand.
3288 2018-04-20 Martin Jambor <mjambor@suse.cz>
3291 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3292 recursion dependency to only apply to non-clones.
3294 2018-04-20 Martin Jambor <mjambor@suse.cz>
3297 * ipa-cp.c (create_specialized_node): Check that clones of
3298 self-recursive edges exist during IPA-CP.
3300 2018-04-19 Toon Moene <toon@moene.org>
3302 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3305 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3307 PR tree-optimization/85467
3308 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3309 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3310 VECTOR_CST element to type.
3312 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3315 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3316 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3317 (builtin_longjmp): Likewise.
3318 (save_stack_nonlocal): New pattern.
3319 (restore_stack_nonlocal): Likewise.
3321 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3324 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3325 Replace ASM_OUTPUT_LABEL with fprintf.
3327 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3330 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3331 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3332 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3333 define __IBT__ and __SHSTK__ for -fcf-protection.
3334 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3336 (ix86_trampoline_init): Likewise.
3337 (x86_output_mi_thunk): Likewise.
3338 (ix86_notrack_prefixed_insn_p): Likewise.
3339 (ix86_option_override_internal): Don't disallow -fcf-protection.
3340 * config/i386/i386.md (rdssp<mode>): Also enable for
3342 (incssp<mode>): Likewise.
3343 (nop_endbr): Likewise.
3344 * config/i386/i386.opt (mcet): Change help message to built-in
3348 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3349 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3350 enable CET built-in functions.
3352 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3354 * common/config/i386/i386-common.c
3355 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3356 OPTION_MASK_ISA_MOVDIRI_UNSET,
3357 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3358 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3359 * config.gcc (movdirintrin.h): New header.
3360 * config/i386/cpuid.h (bit_MOVDIRI,
3361 bit_MOVDIR64B): New bits.
3362 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3364 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3365 (VOID, PVOID, PCVOID)): New function types.
3366 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3367 __builtin_ia32_directstoreu_u64,
3368 __builtin_ia32_movdir64b): New builtins.
3369 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3370 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3372 (ix86_valid_target_attribute_inner_p): Ditto.
3373 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3374 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3375 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3376 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3377 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3378 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3379 (movdiri<mode>, movdir64b_<mode>): New.
3380 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3381 * config/i386/immintrin.h: Include movdirintrin.h.
3382 * config/i386/movdirintrin.h: New file.
3383 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3385 2018-04-19 Richard Biener <rguenther@suse.de>
3388 * cfg.c (clear_bb_flags): When loop state says we have
3389 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3391 2018-04-19 Richard Biener <rguenther@suse.de>
3393 PR tree-optimization/84737
3394 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3395 copying restrict info.
3396 (vect_setup_realignment): Use it.
3397 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3398 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3399 the first DR to all generated stores.
3400 (vectorizable_load): Likewise for loads.
3402 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3404 PR tree-optimization/85446
3405 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3406 the integral and pointer types to have the same precision.
3408 * doc/install.texi: Document --disable-cet being the default and
3411 2018-04-18 Martin Liska <mliska@suse.cz>
3413 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3416 2018-04-18 Martin Liska <mliska@suse.cz>
3419 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3422 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3423 arguments if they are comparable.
3425 2018-04-18 Martin Liska <mliska@suse.cz>
3428 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3431 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3434 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3437 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3438 ENDBR after calling __morestack.
3440 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3443 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3444 by using gcc_base_ver to generate a gcc_driver_version, and use
3445 it when generating GCC_DRIVER_NAME.
3446 * configure: Regenerate.
3448 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3451 * config.gcc: Obsolete powerpc*-*-*spe*.
3453 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3456 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3457 (stabstr_D): Change type of unum from unsigned int to
3458 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3461 2018-04-17 Jim Wilson <jimw@sifive.com>
3464 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3465 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3466 Set arg_pointer_offset after using pretend_args_size.
3468 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3470 PR rtl-optimization/85431
3471 * dse.c (record_store): Ignore zero width stores.
3474 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3475 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3476 __builtin_stack_restore rather than after it.
3477 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3478 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3479 argument instead of virtual_dynamic_stack_rtx.
3481 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3483 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3485 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3486 Add note to error message to explain internal mapping of overloaded
3487 built-in function name to non-overloaded built-in function name.
3488 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3491 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3494 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3495 where the inputs overlap with the output.
3497 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3499 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3500 (=v, v) alternative and explicit "memory" attribute.
3501 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3502 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3504 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3505 "sselog1" type instead of "sselog".
3506 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3507 "sselog". Remove explicit "memory" attribute.
3508 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3509 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3511 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3512 "sselog1" type instead of "sselog". Remove explicit "memory"
3514 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3515 use "sselog1" type instead of "sselog". Remove explicit "memory"
3517 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3518 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3520 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3521 "sselog1" type instead of "sselog". Remove explicit "memory"
3523 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3524 use "sselog1" type instead of "sselog". Remove explicit "memory"
3528 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3531 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3534 2018-04-17 Martin Jambor <mjambor@suse.cz>
3537 * ipa-cp.c (create_specialized_node): Call
3538 expand_all_artificial_thunks if necessary.
3540 2018-04-17 Martin Liska <mliska@suse.cz>
3543 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3544 in message, remote space in between '_G' and '('.
3546 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3549 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3550 avx512f_vmcmp<mode>3<round_saeonly_name>,
3551 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3552 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3553 avx512f_rndscale<mode><round_saeonly_name>,
3554 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3555 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3556 Use %<iptr>2 instead of %2 for -masm=intel.
3557 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3558 avx512f_vcvttss2usi<round_saeonly_name>,
3559 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3561 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3562 avx512f_vcvttsd2usi<round_saeonly_name>,
3563 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3564 Use %q1 instead of %1 for -masm=intel.
3565 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3566 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3567 of %3 for -masm=intel.
3568 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3570 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3572 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3574 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3575 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3577 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3579 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3580 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3581 %g1 and one with %0 and %1.
3582 (avx512er_vmrcp28<mode><round_saeonly_name>,
3583 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3585 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3586 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3587 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3588 of %0 and %{%4%} for -masm=intel.
3589 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3590 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3591 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3592 order of %0 and %{%5%}%{z%} for -masm=intel.
3594 2018-04-17 Jan Hubicka <jh@suse.cz>
3597 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3599 2018-04-17 Martin Liska <mliska@suse.cz>
3602 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3603 for target_clone error message. Make default implementation
3604 clone to be a local declaration.
3605 (separate_attrs): Add new argument and check for an empty
3607 (expand_target_clones): Handle it.
3608 (ipa_target_clone): Make redirection just for target_clones
3611 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3612 Tom de Vries <tom@codesourcery.com>
3615 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3616 tiled basic blocks without omp continue statements.
3618 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3621 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3622 vec_extract expression as having side effects to make sure it gets
3625 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3628 * config/i386/i386.c (get_builtin_code_for_version): Check
3631 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3634 * gcc/config.gcc: Support "skylake".
3635 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3637 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3638 (processor_target_table): Add "skylake".
3639 (ix86_option_override_internal): Add "skylake".
3640 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3641 PROCESSOR_CANNONLAKE.
3642 (get_builtin_code_for_version): Fix priority for
3643 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3644 PROCESSOR_SKYLAKE-AVX512.
3645 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3646 (processor_type): Add PROCESSOR_SKYLAKE.
3648 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3649 Jason Merrill <jason@redhat.com>
3652 * convert.c (convert_to_integer_1): Use direct recursion for
3653 enumeral types and types with a precision less than the number
3654 of bits in their mode.
3656 2018-04-16 Julia Koval <julia.koval@intel.com>
3659 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3660 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3662 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3665 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3666 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3667 and -mno-direct-move.
3669 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3672 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3673 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3674 Ensure negative shifts result in {0}.
3676 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3678 PR rtl-optimization/79916
3679 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3680 regs (if any) to define how to gnerate SD moves when LRA is in
3683 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3685 PR rtl-optimization/85393
3686 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3687 * except.c (expand_dw2_landing_pad_for_region): Make static.
3688 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3689 a label and unconditional jump to old_bb, rather than
3690 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3693 PR rtl-optimization/85376
3694 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3695 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3696 instead of a specific value.
3698 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3699 Bin Cheng <bin.cheng@arm.com>
3701 PR tree-optimization/82965
3702 PR tree-optimization/83991
3703 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3704 by_profile_only parameter.
3705 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3706 information if the loop was predicted to iterate too many times.
3707 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3709 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3712 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3715 2018-04-13 Martin Liska <mliska@suse.cz>
3716 Jakub Jelinek <jakub@redhat.com>
3719 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3720 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3721 * builtins.c (expand_builtin_memory_copy_args): Use
3722 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3723 handle dest_addr == pc_rtx.
3725 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3728 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3729 asked to not generate direct moves.
3730 (fix_trunc<mode>si2_stfiwx): Similar.
3731 (fix_trunc<mode>si2_internal): Similar.
3733 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3736 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3737 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3738 lookup if dest in some wider mode is known to be const0_rtx and
3739 if so, record permanent equivalence for it to be ZERO_EXTEND of
3740 the narrower mode destination.
3742 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3744 * lto-streamer-out.c (output_function): Revert 259346.
3745 * omp-expand.c (expand_oacc_for): Likewise.
3747 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3749 PR rtl-optimization/85354
3750 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3751 * sel-sched.c (sel_global_init): ... here.
3753 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3756 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3757 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3758 mode for PE-COFF targets.
3759 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3760 (i386_pe_asm_lto_end): Likewise.
3761 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3762 (TARGET_ASM_LTO_END): Likewise.
3763 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3764 (i386_pe_asm_lto_start): New function.
3765 (i386_pe_asm_lto_end): Likewise.
3767 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3768 Richard Biener <rguenther@suse.de>
3771 * lto-streamer-out.c (output_function): Fix CFG loop state before
3773 * omp-expand.c (expand_oacc_for): Handle calls to internal
3774 functions like regular functions.
3776 2018-04-12 Richard Biener <rguenther@suse.de>
3779 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3780 for the early LTO debug to properly generate references to it
3781 during DIE emission. Do not re-use that for the skeleton for
3783 (dwarf2out_early_finish): Likewise.
3785 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3788 * config/i386/sse.md
3789 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3790 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3791 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3792 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3793 and output is a reg, avoid creating invalid lowpart subreg, but
3794 instead split into a 512-bit move. Don't split if not AVX512VL,
3795 input is xmm16+ reg and output is a mem.
3796 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3797 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3798 xmm16+ reg and output is a mem.
3800 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3802 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3803 also for flag_dwarf2_cfi_asm.
3805 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3807 PR rtl-optimization/85342
3808 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3809 a bool scalar var inside of the loop instead. Don't try to update
3810 recog_data.operand after failed apply_change_group.
3812 2018-04-12 Tom de Vries <tom@codesourcery.com>
3815 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3816 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3817 array with flexible array member as array without given dimension.
3818 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3819 argument for undefined param to true.
3821 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3824 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3825 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3826 from PowerPC section.
3827 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3828 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3829 help text that is too long.
3830 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3831 help text that is too long.
3832 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3833 help text that is too long.
3835 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3837 * config/alpha/alpha.md (stack_probe_internal): Rename
3838 from "probe_stack". Update all callers.
3840 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3842 PR rtl-optimization/84566
3843 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3844 sched_macro_fuse_insns.
3846 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3849 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3850 (compute_block_dependences): ... from here.
3852 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3854 PR tree-optimization/85331
3855 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3856 from int to HOST_WIDE_INT.
3858 2018-04-11 Martin Jambor <mjambor@suse.cz>
3861 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3862 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3863 not the same as the source val.
3864 (cgraph_edge_brings_value_p): New parameter.
3865 (gather_edges_for_value): Pass destination value to
3866 cgraph_edge_brings_value_p.
3867 (perhaps_add_new_callers): Likewise.
3868 (get_info_about_necessary_edges): Likewise and exclude values brought
3869 only by self-recursive edges.
3870 (create_specialized_node): Redirect only clones of self-calling edges.
3871 (+self_recursive_pass_through_p): New function.
3872 (find_more_scalar_values_for_callers_subset): Use it.
3873 (find_aggregate_values_for_callers_subset): Likewise.
3874 (known_aggs_to_agg_replacement_list): Removed.
3875 (decide_whether_version_node): Re-calculate known constants for all
3876 remaining context clones.
3878 2018-04-11 Richard Biener <rguenther@suse.de>
3881 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3882 from early DWARF output.
3883 (dwarf2out_early_finish): Output line info unconditionally into
3884 early DWARF and add reference to it.
3886 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3889 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3890 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3891 other than V2DFmode using iptr mode attribute.
3892 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3894 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3896 PR rtl-optimization/84659
3897 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3899 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3902 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3904 (output_loc_list): Pass address of a dummy size variable even in the
3905 locview handling loop.
3906 (index_location_lists): Add comment on why skip_loc_list_entry can't
3909 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3912 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3915 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3918 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3919 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3920 and -mstring-compare-inline-limit.
3922 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3925 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3926 for stack clash protection in a register whenever we need it to be in
3929 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3931 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3932 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3934 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3937 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3939 (mlong-double-): Ditto.
3940 * config/rs6000/sysv4.opt (msdata=): Ditto.
3941 (mtls-size=): Ditto.
3943 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3945 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3946 erroneous entries for
3947 "vector int vec_ldl (int, long int *)", and
3948 "vector unsigned int vec_ldl (int, unsigned long int *)".
3949 Add comments and entries for
3950 "vector bool char vec_ldl (int, bool char *)",
3951 "vector bool short vec_ldl (int, bool short *)",
3952 "vector bool int vec_ldl (int, bool int *)",
3953 "vector bool long long vec_ldl (int, bool long long *)",
3954 "vector pixel vec_ldl (int, pixel *)",
3955 "vector long long vec_ldl (int, long long *)",
3956 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3957 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3958 type tree bool_long_long_type_node and correct definition of
3959 bool_V2DI_type_node to make reference to this new type tree.
3960 (rs6000_mangle_type): Replace erroneous reference to
3961 bool_long_type_node with bool_long_long_type_node.
3962 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3963 comments to emphasize sign distinctions for char and int types and
3964 replace RS6000_BTI_bool_long constant with
3965 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3966 use of RS6000_BTI_pixel.
3967 (bool_long_type_node): Remove this macro definition.
3968 (bool_long_long_type_node): New macro definition
3970 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3972 PR rtl-optimization/85300
3973 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3974 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3975 simplify_unary_operation fails.
3977 2018-04-10 Martin Liska <mliska@suse.cz>
3979 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3980 cgraph_edge and ipa_ref.
3982 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3986 * config/i386/sse.md
3987 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3988 computation of the VEC_MERGE selector from mask.
3989 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3990 Fix decoding of the VEC_MERGE selector into mask.
3992 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3994 PR tree-optimization/85286
3995 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3997 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3999 * final.c (final_1): Set insn_last_address as well as
4000 insn_current_address.
4002 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4005 * explow.c (emit_stack_probe): Call validize_mem on memory location
4006 before passing it to gen_probe_stack. Create address operand and
4007 legitimize it for the probe_stack_address case.
4009 2018-04-09 Jan Hubicka <jh@suse.cz>
4012 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4013 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4014 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4016 (free_lang_data): Rebuild type inheritance graph.
4018 2018-04-09 Martin Sebor <msebor@redhat.com>
4020 * invoke.texi (-finline-small-functions): Mention other optimization
4022 (-findirect-inlining, -fpartial-inlining): Same.
4023 (-finline-functions-called-once): Same.
4024 (-freorder-blocks-and-partition): Same.
4026 2018-04-09 Jan Hubicka <jh@suse.cz>
4029 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4030 jumps; choose last target that matches the criteria (i.e.
4031 no partition changes for non-crossing jumps).
4032 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4033 support for redirecting crossing jumps to non-crossing.
4035 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4037 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4038 also for naked functions.
4040 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4042 * config/arc/arc.md (add_shift): New pattern.
4043 (add_shift2): Likewise.
4044 (sub_shift): Likewise.
4045 (sub_shift_cmp0_noout): Likewise.
4046 (compare_si_ashiftsi): Likewise.
4047 (xbfu_cmp0_noout): New combine pattern.
4048 (xbfu_cmp0"): Likewise.
4049 (movsi_set_cc_insn): Place the predicable variant first.
4050 (commutative_binary_cmp0_noout): Remove clobber.
4051 (commutative_binary_cmp0): New pattern.
4052 (noncommutative_binary_cmp0): Likewise.
4053 (noncommutative_binary_cmp0_noout): Likewise.
4054 (noncommutative_binary_comparison_result_used): Removed.
4055 (rsub_cmp0): New pattern.
4056 (rsub_cmp0_noout): Likewise.
4057 (extzvsi): Changed, keep only meaningful variants.
4058 (SQH, SEZ): New iterators.
4059 (SQH_postfix): New mode attribute.
4060 (SEZ_prefix): New code attribute.
4061 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4062 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4063 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4065 (noncommutative_operator): Check the availability of barrel
4068 2018-04-09 Richard Biener <rguenther@suse.de>
4070 PR tree-optimization/85284
4071 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4072 Only use the niter constraining form of simple_iv when the exit
4075 2018-04-09 Tom de Vries <tom@codesourcery.com>
4078 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4079 (define_expand "*memory_barrier"): New define_expand.
4080 (define_insn "memory_barrier"): New insn.
4082 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4084 PR rtl-optimization/80463
4085 PR rtl-optimization/83972
4086 PR rtl-optimization/83480
4088 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4089 correct producer for the insn.
4090 (tidy_control_flow): Fixup seqnos in case of debug insns.
4092 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4094 PR rtl-optimization/83913
4096 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4097 different sched-times when merging exprs.
4099 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4101 PR rtl-optimization/83962
4103 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4104 tidy_fallthru_edge and tidy_control_flow.
4106 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4108 PR rtl-optimization/83530
4110 * sel-sched.c (force_next_insn): New global variable.
4111 (remove_insn_for_debug): When force_next_insn is true, also leave only
4112 next insn in the ready list.
4113 (sel_sched_region): When the region wasn't scheduled, make another pass
4114 over it with force_next_insn set to 1.
4116 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4118 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4120 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4121 for interrupt control.
4122 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4123 functions for interrupt control.
4124 * config/nds32/nds32-intrinsic.md: Likewise.
4125 * config/nds32/nds32_intrinsic.h: Likewise.
4126 * config/nds32/nds32.h (nds32_builtins): Likewise.
4128 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4130 * config/nds32/nds32.c (nds32_init_machine_status,
4131 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4132 strict_aligned_p field.
4133 (nds32_expand_to_rtl_hook): New function.
4134 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4135 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4137 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4138 Chung-Ju Wu <jasonwucj@gmail.com>
4140 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4141 * config/nds32/nds32-n7.md: New file.
4142 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4143 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4145 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4146 * config/nds32/nds32.md (pipeline_model): Add n7.
4147 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4148 * config/nds32/pipelines.md: Include n7 settings.
4150 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4151 Chung-Ju Wu <jasonwucj@gmail.com>
4153 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4154 * config/nds32/nds32-e8.md: New file.
4155 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4156 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4158 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4159 * config/nds32/nds32.md (pipeline_model): Add e8.
4160 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4161 * config/nds32/pipelines.md: Include e8 settings.
4163 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4164 Chung-Ju Wu <jasonwucj@gmail.com>
4166 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4167 * config/nds32/nds32-n8.md: New file.
4168 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4169 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4171 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4172 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4173 * config/nds32/nds32.md (pipeline_model): Add n8.
4174 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4175 * config/nds32/pipelines.md: Include n8 settings.
4177 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4178 Chung-Ju Wu <jasonwucj@gmail.com>
4180 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4181 * config/nds32/nds32-n9-2r1w.md: New file.
4182 * config/nds32/nds32-n9-3r2w.md: New file.
4183 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4184 nds32_register_ports): New or modify for cpu n9.
4185 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4187 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4188 * config/nds32/nds32-utils.c: New file.
4189 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4190 TARGET_MUL_SLOW): Define.
4191 * config/nds32/nds32.md (pipeline_model): New attribute.
4192 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4193 New options that support cpu n9.
4194 * config/nds32/pipelines.md: Include n9 settings.
4195 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4197 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4199 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4200 information if necessary.
4201 (output_cond_branch_compare_zero): Likewise.
4202 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4203 (nds32_target_alignment): Refine for alignment.
4204 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4205 (FUNCTION_BOUNDARY): Modify.
4206 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4208 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4210 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4212 * config/nds32/constants.md (unspec_volatile_element): Add values for
4213 TLB operation and data prefetch.
4214 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4215 functions for TLB operation and data prefetch.
4216 * config/nds32/nds32-intrinsic.md: Likewise.
4217 * config/nds32/nds32_intrinsic.h: Likewise.
4218 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4219 (nds32_print_operand): Likewise.
4220 * config/nds32/nds32.h (nds32_builtins): Likewise.
4222 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4223 Andrew Pinski <pinsika@gcc.gnu.org>
4226 * match.pd: Use constant_boolean_node of correct type instead of
4227 boolean_true_node or boolean_false_node for simplifying
4228 pointer comparisons to zero.
4230 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4232 PR tree-optimization/80021
4233 * tree.c (verify_type_variant): Make error call in verify_variant_match
4234 translatable and remove final full stop.
4236 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4238 * config/nds32/constants.md (unspec_volatile_element): Add
4239 UNSPEC_VOLATILE_EH_RETURN.
4240 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4241 nds32_output_stack_pop): Support dwarf exception handling process.
4242 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4243 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4244 exception handling process.
4245 (nds32_compute_stack_frame): Likewise.
4246 (nds32_return_addr_rtx): Likewise.
4247 (nds32_initial_elimination_offset): Likewise.
4248 (nds32_expand_prologue): Likewise.
4249 (nds32_expand_epilogue): Likewise.
4250 (nds32_dynamic_chain_address): New function.
4251 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4253 (DYNAMIC_CHAIN_ADDRESS): Define.
4254 (EH_RETURN_DATA_REGNO): Define.
4255 (EH_RETURN_STACKADJ_RTX): Define.
4256 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4257 patterns for dwarf exception handling.
4259 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4261 * config/nds32/nds32.h: Clean up obsolete macros.
4263 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4265 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4266 Add enum values for particular instructions.
4267 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4268 particular intrinsic functions.
4269 * config/nds32/nds32-intrinsic.md: Likewise.
4270 * config/nds32/nds32_intrinsic.h: Likewise.
4271 * config/nds32/nds32.h (nds32_builtins): Likewise.
4272 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4273 (btst, ave): New patterns for particular instructions.
4275 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4277 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4278 Add enum values for atomic load/store and memory sync.
4279 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4281 * config/nds32/nds32-intrinsic.md: Likewise.
4282 * config/nds32/nds32_intrinsic.h: Likewise.
4283 * config/nds32/nds32.h (nds32_builtins): Likewise.
4285 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4287 PR tree-optimization/85257
4288 * fold-const.c (native_encode_vector): If not all elts could fit
4289 and off is -1, return 0 rather than offset.
4290 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4291 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4292 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4293 adjust buffer in native_interpret_expr call.
4295 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4297 * config/nds32/constants.md (unspec_volatile_element): Add cache
4298 control enum values.
4299 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4300 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4301 * config/nds32/nds32.c (nds32_cctl_names): New.
4302 (nds32_print_operand): Handle cache control register names.
4303 * config/nds32/nds32.h (nds32_builtins): New enum values.
4304 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4306 * config/nds32/nds32.md (type): Add mmu.
4307 * config/nds32/pipelines.md (simple_insn): Add mmu.
4309 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4311 * config/nds32/nds32.md (type): Remove call.
4312 * config/nds32/pipelines.md (simple_insn): Likewise.
4314 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4316 * config/nds32/constants.md (unspec_volatile_element): Add
4317 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4318 UNSPEC_VOLATILE_FMFCFG.
4319 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4320 description for fmfcfg and fmfcsr.
4321 (bdesc_1arg): Add fmtcsr.
4322 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4323 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4324 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4325 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4326 unspec_fmfcfg): New patterns.
4327 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4328 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4329 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4330 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4331 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4332 __nds32__fmfcfg): Define.
4334 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4336 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4337 intrinsic register names.
4338 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4339 intrinsic register enum values and macros.
4341 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4343 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4344 for load/store addressing form.
4345 (nds32_print_operand_address): Likewise.
4347 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4350 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4351 based on LABEL_REF. Remove useless assertion.
4352 (pic_address_needs_scratch): Fix formatting.
4353 (sparc_legitimize_pic_address): Minor tweaks.
4354 (sparc_delegitimize_address): Adjust assertion accordingly.
4355 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4356 into symbolic_operand.
4357 (movsi_high_pic_label_ref): Likewise.
4358 (movsi_lo_sum_pic_label_ref): Likewise.
4359 (movdi_pic_label_ref): Likewise.
4360 (movdi_high_pic_label_ref): Likewise.
4361 (movdi_lo_sum_pic_label_ref): Likewise.
4363 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4365 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4366 custom LIB_SPEC setup.
4368 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4369 Kito Cheng <kito.cheng@gmail.com>
4371 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4372 * config/riscv/freebsd.h: New.
4374 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4376 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4377 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4380 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4381 Kito Cheng <kito.cheng@gmail.com>
4383 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4384 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4385 * config/nds32/nds32-protos.h (nds32_output_call,
4386 nds32_output_return): Declare.
4387 * config/nds32/nds32.md: Refine all the call and return patterns.
4389 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4392 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4393 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4395 PR rtl-optimization/84872
4396 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4397 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4400 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4402 * expr.c (copy_blkmode_to_reg): Revert 254862.
4403 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4405 2018-04-06 Richard Biener <rguenther@suse.de>
4408 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4409 after seeing a component reference with an adjacent field. Treat
4410 refs to arrays at struct end of external decls similar to
4411 refs to unconstrained commons.
4413 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4416 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4417 look through SAVE_EXPRs with non-side-effects argument. Adjust
4419 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4422 2018-04-06 Richard Biener <rguenther@suse.de>
4425 * alias.c (find_base_term): New wrapper around find_base_term
4426 unwinding CSELIB_VAL_PTR changes.
4427 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4430 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4432 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4434 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4435 constant definitions.
4436 ("nop"): lr 0,0 -> nopr r0
4437 ("nop_lr0", "nop_lr1"): New insn definitions.
4439 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4441 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4442 NDS32_V3PUSH_AVAILABLE_P macro.
4444 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4445 Chung-Ju Wu <jasonwucj@gmail.com>
4447 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4448 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4449 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4450 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4451 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4452 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4453 * config/nds32/constraints.md: New constraints and checking for hard
4454 float configuration.
4455 * config/nds32/iterators.md: New mode iterator and attribute for hard
4456 float configuration.
4457 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4459 * config/nds32/nds32-fpu.md: New file.
4460 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4461 deal with hard float code generation.
4462 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4464 (abi_type, float_reg_number): New enum type.
4465 * config/nds32/nds32-predicates.c: New predicates for hard float.
4466 * config/nds32/nds32-protos.h: Declare functions for hard float.
4467 * config/nds32/nds32.c: Implementation for hard float configuration.
4468 * config/nds32/nds32.h: Definitions for hard float configuration.
4469 * config/nds32/nds32.md: Include hard float machine description and
4470 modify patterns for hard float configuration.
4471 * config/nds32/nds32.opt: New options for hard float configuration.
4472 * config/nds32/predicates.md: New predicates for hard float
4475 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4477 * common/config/nds32/nds32-common.c
4478 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4480 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4483 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4484 CONSTRUCTOR_ELT (ctor, ...)->value.
4486 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4489 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4491 2018-04-05 Tom de Vries <tom@codesourcery.com>
4494 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4497 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4498 Kito Cheng <kito.cheng@gmail.com>
4500 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4501 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4502 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4503 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4505 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4506 Kito Cheng <kito.cheng@gmail.com>
4508 * config/nds32/constraints.md (Ufe): New memory constraint.
4509 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4510 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4511 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4513 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4514 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4516 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4518 * config/nds32/nds32.md: Use optimize_size in the condition for
4519 alu-shift instructions.
4521 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4523 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4525 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4527 * config/nds32/nds32.md (negsi2): Refine pattern.
4529 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4530 Chung-Ju Wu <jasonwucj@gmail.com>
4532 * config/nds32/iterators.md (shift_rotate): New code iterator.
4533 (shift): New code attribute.
4534 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4535 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4536 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4537 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4538 bit-wise operations.
4539 (andsi3, *andsi3): Ditto.
4540 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4541 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4542 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4543 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4544 nds32_ior_operand, nds32_xor_operand): New predicates.
4546 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4548 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4549 (addsi3, subsi3): ... this.
4551 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4553 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4555 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4557 * config/nds32/nds32.md: Adjust indention.
4559 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4561 * config/nds32/nds32.md (feature): New attribute.
4563 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4565 * config/nds32/nds32.md (subtype): New attribute.
4567 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4570 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4571 expansion to perform a bitwise AND of the argument followed by a
4572 boolean negation of the result.
4574 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4576 PR rtl-optimization/84878
4577 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4578 the basic block. Assert the use reference is not artificial and that
4579 it has an associated insn.
4581 2018-04-04 Michael Matz <matz@suse.de>
4583 * builtins.c (compute_objsize): Pass correct operand
4584 to array_at_struct_end_p.
4586 2018-04-04 Richard Biener <rguenther@suse.de>
4589 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4590 from contexts for DINFO_LEVEL_TERSE and below.
4592 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4594 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4595 resiter_operand condition.
4596 * config/nds32/nds32.md (*move<mode>): Ditto.
4598 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4599 Monk Chiang <sh.chiang04@gmail.com>
4601 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4603 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4605 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4607 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4608 Kito Cheng <kito.cheng@gmail.com>
4610 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4611 nds32_cond_code_str, output_cond_branch,
4612 output_cond_branch_compare_zero, nds32_expand_cbranch,
4613 nds32_expand_cstore, nds32_expand_movcc,
4614 nds32_output_cbranchsi4_equality_zero,
4615 nds32_output_cbranchsi4_equality_reg,
4616 nds32_output_cbranchsi4_equality_reg_or_const_int,
4617 nds32_output_cbranchsi4_greater_less_zero: New functions.
4618 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4619 nds32_expand_cstore, nds32_expand_movcc,
4620 nds32_output_cbranchsi4_equality_zero,
4621 nds32_output_cbranchsi4_equality_reg,
4622 nds32_output_cbranchsi4_equality_reg_or_const_int,
4623 nds32_output_cbranchsi4_greater_less_zero): Declare.
4624 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4625 nds32_rimm11s_operand): New predicates.
4626 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4627 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4630 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4632 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4633 * config/nds32/nds32.md: Ditto.
4634 * config/nds32/pipelines.md: Ditto.
4636 2018-04-04 Richard Biener <rguenther@suse.de>
4638 PR tree-optimization/85168
4639 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4640 propagating abnormals.
4642 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4644 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4646 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4647 Kito Cheng <kito.cheng@gmail.com>
4649 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4650 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4651 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4652 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4653 * config/nds32/nds32.md (sibcall_internal): New.
4654 (sibcall_register): Remove.
4655 (sibcall_immediate): Remove.
4656 (sibcall_value_internal): New.
4657 (sibcall_value_register): Remove.
4658 (sibcall_value_immediate): Remove.
4659 * config/nds32/predicates.md (nds32_general_register_operand): New.
4660 (nds32_call_address_operand): New.
4662 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4664 PR rtl-optimization/85167
4665 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4666 bb_defs if *split_p, instead preinitialize it to NULL.
4668 PR tree-optimization/85156
4669 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4670 evaluating the argument multiple times.
4672 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4674 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4676 (_mm_cvtpd_ps): Likewise.
4677 (_mm_cvttpd_epi32): Likewise.
4678 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4679 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4680 vector, pixel, and bool following altivec.h include.
4682 2018-04-03 Martin Sebor <msebor@redhat.com>
4684 * doc/extend.texi (Common Function Attributes): Clarify.
4685 (const attribute): Likewise.
4686 (pure attribute): Likewise.
4688 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4691 * config/i386/i386.c (ix86_expand_vector_set): Use
4692 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4694 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4696 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4697 instructions when changing rounding bits to preserve precision bits
4698 in the x87 control word.
4700 2018-04-03 Martin Liska <mliska@suse.cz>
4702 PR tree-optimization/82491
4703 * rtl.h (strip_offset_and_add): Replace += suboffset with
4704 poly_uint64 () + suboffset.
4706 2018-03-29 Martin Liska <mliska@suse.cz>
4707 Martin Jambor <mjambor@suse.cz>
4710 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4711 param_type is not an integral or pointer type.
4713 2018-04-03 Richard Biener <rguenther@suse.de>
4715 * sese.h (recompute_all_dominators): Remove.
4717 2018-04-02 Martin Sebor <msebor@redhat.com>
4719 * doc/invoke.texi (-Wrestrict): Fix typos.
4721 2018-04-02 Jim Wilson <jimw@sifive.com>
4723 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4724 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4725 (<optab>di3, <optab>si3_extend): Likewise.
4726 (<optab>si3_mask, <optab>si3_mask_1): New.
4727 (<optab>di3_mask, <optab>di3_mask_1): New.
4728 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4729 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4730 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4732 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4734 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4737 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4739 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4740 (nds32_canonicalize_comparison): New function.
4742 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4743 Kito Cheng <kito.cheng@gmail.com>
4744 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4746 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4747 * config/nds32/constants.md (unspec_volatile_element): Add
4748 UNSPEC_VOLATILE_RELAX_GROUP.
4749 * config/nds32/nds32-relax-opt.c: New file.
4750 * config/nds32/nds32-predicates.c
4751 (nds32_symbol_load_store_p): New function.
4752 * config/nds32/nds32-protos.h
4753 (nds32_symbol_load_store_p): Declare function.
4754 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4755 * config/nds32/nds32.c
4756 (nds32_register_pass): New function to register pass.
4757 (nds32_register_passes): New function to register passes.
4758 * config/nds32/nds32.md (relax_group): New pattern.
4759 * config/nds32/nds32.opt (mrelax-hint): New option.
4760 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4762 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4764 * config/nds32/t-nds32: Modify files dependency.
4766 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4768 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4769 (PROFILE_HOOK): Define its implementation.
4771 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4773 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4774 type and 32-bit size.
4776 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4779 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4780 (V_128_256): New mode iterator.
4781 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4782 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4783 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4785 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4786 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4788 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4791 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4792 NaN inputs correctly.
4794 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4797 * config/rs6000/vsx.md (??r): New mode attribute.
4798 (*vsx_mov<mode>_64bit): Use it.
4799 (*vsx_mov<mode>_32bit): Likewise.
4801 2018-03-30 Martin Sebor <msebor@redhat.com>
4803 PR tree-optimization/84818
4804 * builtins.c (check_access): Use warning_n.
4806 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4809 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4811 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4814 2018-03-30 Julia Koval <julia.koval@intel.com>
4817 * x86-tune.def (movx, partial_reg_dependency): Enable for
4820 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4823 * lra-constraints.c (process_alt_operands): Move setting
4824 this_alternative_matches below.
4826 2018-03-29 Martin Liska <mliska@suse.cz>
4829 * doc/invoke.texi: Document how LTO works with debug info.
4830 Describe auto-load support of binutils. Mention 'x86-64'
4831 as valid option value of -march option.
4833 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4835 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4838 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4839 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4840 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4843 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4846 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4847 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4848 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4849 for RS6000_BTM_POWERPC64.
4850 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4851 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4852 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4857 2018-03-28 Carl Love <cel@us.ibm.com>
4860 2017-09-27 Carl Love <cel@us.ibm.com>
4862 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4863 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4864 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4867 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4869 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4870 instead of __vector bool.
4871 (_mm_max_pu8): Likewise.
4872 (_mm_min_pi16): Likewise.
4874 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4877 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4878 (DIVWEUO): Likewise.
4880 (DIVDEUO): Likewise.
4881 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4882 DIVWEUO and DIVDEUO.
4883 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4884 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4885 (div_extend): Likewise.
4886 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4888 (__builtin_divweuo): Likewise.
4889 (__builtin_divdeo): Likewise.
4890 (__builtin_divdeuo): Likewise.
4892 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4895 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4896 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4898 PR tree-optimization/82004
4899 * gimple-match-head.c (optimize_pow_to_exp): New function.
4900 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4901 Don't fold to exp if optimize_pow_to_exp is false.
4903 2018-03-28 Martin Liska <mliska@suse.cz>
4906 * calls.c (initialize_argument_information): Fix trailing space.
4907 * common.opt: Fix typo and provide better explanation for
4908 -fsanitize-coverage option.
4909 * config/i386/i386.opt: Fix typo.
4911 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4912 Martin Liska <mliska@suse.cz>
4915 * gimplify.c (asan_poison_variable): Don't do the check for
4916 gimplify_omp_ctxp here.
4917 (gimplify_decl_expr): Do it here.
4918 (gimplify_target_expr): Likewise.
4920 2018-03-28 Martin Liska <mliska@suse.cz>
4923 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4924 chkp_type_bounds_count if MPX is not enabled.
4926 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4928 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4930 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4933 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4934 function to create the function decl for complex long double
4935 multiply and divide for -mabi=ieeelongdouble.
4936 (init_float128_ieee): Call it.
4938 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4941 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4942 -fcf-protection=branch -mibt.
4943 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4945 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4948 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4950 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4953 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4954 extern array declarations.
4956 2018-03-27 Richard Biener <rguenther@suse.de>
4959 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4960 explicit single_use checks.
4962 2018-03-27 Richard Biener <rguenther@suse.de>
4964 PR tree-optimization/85082
4965 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4968 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4970 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4971 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4972 Turn on fasynchronous-unwind-tables and funwind-tables.
4974 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4977 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4978 (*bmi_blsr_<mode>_ccz): Ditto.
4980 2018-03-26 Tom de Vries <tom@codesourcery.com>
4982 PR tree-optimization/85063
4983 * omp-general.c (offloading_function_p): New function. Factor out
4985 * omp-offload.c (pass_omp_target_link::gate): ... here.
4986 * omp-general.h (offloading_function_p): Declare.
4987 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4988 with attribute omp declare target for offloading functions.
4990 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4992 PR tree-optimization/84005
4993 * tree-data-ref.h (get_base_for_alignment): Declare.
4994 * tree-data-ref.c (get_base_for_alignment_1): New function.
4995 (get_base_for_alignment): Likewise.
4996 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4997 get_base_for_alignment to find a suitable base object, instead
4998 of always using drb->base_address.
5000 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5003 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5004 known size by default.
5006 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5009 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5010 and non BLKmode operands.
5012 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5015 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5016 Clean up attributes.
5018 2018-03-23 Richard Biener <rguenther@suse.de>
5021 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5022 we are going to emit early debug for LTO.
5024 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5027 * function.c (match_asm_constraints_1): Don't optimize if input
5028 doesn't satisfy general_operand predicate for output's mode.
5031 * alias.c (write_dependence_p): Don't require for x_canonicalized
5032 non-VOIDmode if x has VOIDmode.
5035 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5036 just don't try to optimize it rather than assert it never happens.
5038 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5040 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5041 macro expansions for definition of ST_INTERNAL_<mode> and
5042 LD_INTERNAL_<mode> builtins.
5043 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5045 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5047 (altivec_expand_st_builtin): Likewise.
5048 (altivec_expand_builtin): Remove calls to deleted functions.
5049 (rs6000_address_for_altivec): Delete this function.
5050 * config/rs6000/vector.md: Remove expands for
5051 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5053 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5056 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5057 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5058 re-computing once computed.
5059 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5060 (arm_init_machine_status): Initialize
5061 machine->static_chain_stack_bytes.
5063 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5066 * doc/extend.texi: Add four new prototypes for vec_ld.
5067 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5068 definitions for more logical presentation.
5069 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5070 entries for V1TI variants of __builtin_altivec_ld builtin.
5071 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5072 handling of V1TI variant of LVX icode pattern.
5073 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5074 (rs6000_gimple_fold_builtin): Likewise.
5075 (altivec_init_builtins): Add code to define
5076 __builtin_altivec_lvx_v1ti function.
5078 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5081 * function.c (match_asm_constraints_1): Don't do the optimization
5082 if input isn't a REG, SUBREG, MEM or constant.
5084 2018-03-22 Tom de Vries <tom@codesourcery.com>
5086 PR tree-optimization/84956
5087 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5088 bb_has_abnormal_pred.
5090 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5093 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5094 DECL_INITIAL (decl) to decl at the end.
5095 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5098 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5100 * doc/extend.texi (__builtin_tgmath): Document when complex
5101 integer types are treated as _Complex _Float64.
5103 2018-03-21 Tom de Vries <tom@codesourcery.com>
5105 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5107 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5109 PR tree-optimization/84960
5110 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5111 if it is ENTRY block, move them into single succ of ENTRY in that case.
5113 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5115 PR tree-optimization/84811
5116 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5117 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5118 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5119 (known_subrange_p): Update accordingly. Cast each value involved
5120 in the size comparison, rather than casting the result of the
5123 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5125 PR tree-optimization/84982
5126 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5127 by flipping the least significant bit rather than all bits from
5128 bitpos to bitpos + bitsize - 1.
5130 2018-03-21 Nathan Sidwell <nathan@acm.org>
5132 * doc/extend.texi (Deprecated Features): Remove mention of
5133 long-deleted deprecations.
5135 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5138 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5139 * configure: Regenerate.
5141 2018-03-21 Tom de Vries <tom@codesourcery.com>
5143 PR tree-optimization/83126
5144 * tree-parloops.c (num_phis): New function.
5145 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5147 2018-03-21 Nathan Sidwell <nathan@acm.org>
5149 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5150 mention anon-struct/union members and trailing attributes.
5152 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5154 PR tree-optimization/84969
5155 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5156 builtin memset partitions if they set different rhs values.
5158 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5160 PR rtl-optimization/84989
5161 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5162 VEC_DUPLICATE with scalar result mode.
5164 2018-03-21 Martin Liska <mliska@suse.cz>
5167 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5168 not intended return statement.
5170 2018-03-21 Martin Liska <mliska@suse.cz>
5173 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5174 (chkp_find_bound_slots_1): Limit number of iterations.
5176 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5179 * Minor grammar fixes for x86 options.
5181 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5184 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5185 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5187 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5190 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5191 (altivec_lvx_<mode>_1op): Likewise.
5192 (altivec_stvx_<mode>_2op): Likewise.
5193 (altivec_stvx_<mode>_1op): Likewise.
5194 (altivec_lvx_<VM2:mode>): New define_expand.
5195 (altivec_stvx_<VM2:mode>): Likewise.
5196 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5197 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5198 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5199 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5200 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5201 (rs6000_gen_lvx): Likewise.
5202 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5203 (altivec_expand_stv_builtin): Likewise.
5204 (altivec_expand_builtin): Likewise.
5205 * config/rs6000/vector.md: Likewise.
5207 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5210 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5213 2018-03-20 Richard Biener <rguenther@suse.de>
5216 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5217 sign-conversions as zero, fall back to standard scalar_stmt
5220 2018-03-20 Martin Liska <mliska@suse.cz>
5223 * predict.c (rebuild_frequencies): Handle case when we have
5224 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5226 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5229 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5230 flag_section_anchors.
5231 * varasm.c (use_blocks_for_decl_p): Remove hack for
5232 dw2_force_const_mem.
5235 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5237 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5238 be created, use lowpart_subreg of operands[0] rather than operands[0]
5240 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5241 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5242 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5243 and n constraint instead of aarch64_shift_imm_di and Usd.
5244 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5245 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5247 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5250 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5251 to favor GPR over NEON registers.
5252 (<shift>di3_neon): Likewise.
5254 2018-03-20 Tom de Vries <tom@codesourcery.com>
5257 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5258 (nvptx_process_pars): Emit bar.sync asap and alap.
5260 2018-03-20 Tom de Vries <tom@codesourcery.com>
5263 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5264 seen_label if seen_label is already set.
5266 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5269 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5270 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5271 Use 1U instead of 1. Formatting fixes.
5274 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5275 instead of TREE_TYPE (s1) for the return value.
5277 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5279 PR tree-optimization/84946
5280 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5281 bitsize + bitsize in poly_uint64 rather than poly_int64.
5284 * dwarf2asm.c: Include fold-const.c.
5285 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5286 of decl rather than decl itself.
5288 PR rtl-optimization/84643
5289 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5291 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5294 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5295 calling assemble_variable.
5297 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5300 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5301 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5303 2018-03-19 Jim Wilson <jimw@sifive.com>
5306 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5307 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5308 (riscv_first_stack_step): Likewise.
5309 (riscv_option_override): Use STACK_BOUNDARY instead of
5311 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5313 (BIGGEST_ALIGNMENT): Set to 128.
5314 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5315 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5318 2018-03-19 Richard Biener <rguenther@suse.de>
5320 PR tree-optimization/84933
5321 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5322 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5324 2018-03-19 Richard Biener <rguenther@suse.de>
5326 PR tree-optimization/84859
5327 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5328 (cond_if_else_store_replacement): Perform sinking operation on
5329 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5330 Generalize what a BB with a single eligible store is.
5332 2018-03-19 Richard Biener <rguenther@suse.de>
5334 PR tree-optimization/84929
5335 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5336 chrec_is_positive against non-chrec arg.
5338 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5341 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5343 2018-03-18 Martin Liska <mliska@suse.cz>
5345 PR rtl-optimization/84635
5346 * regrename.c (build_def_use): Use matches_mode only when
5349 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5351 PR tree-optimization/84913
5352 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5353 vectorize chains of COND_EXPRs.
5355 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5357 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5359 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5361 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5363 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5365 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5367 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5368 Kito Cheng <kito.cheng@gmail.com>
5370 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5371 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5372 (nds32_adjust_reg_alloc_order): New function.
5373 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5375 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5377 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5378 nds32_print_operand, nds32_print_operand_address): Use
5379 HOST_WIDE_INT_PRINT_DEC instead.
5381 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5383 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5385 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5388 * config/i386/i386.c (initial_ix86_tune_features,
5389 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5391 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5392 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5393 rather than 1u << ix86_tune. Formatting fix.
5394 (ix86_option_override_internal): Change ix86_arch_mask from
5395 unsigned int to unsigned HOST_WIDE_INT, initialize to
5396 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5397 (ix86_function_specific_restore): Likewise.
5399 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5402 * postreload.c (reload_combine_recognize_pattern): Perform
5403 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5404 truncate_int_for_mode the result for the destination's mode.
5407 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5408 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5410 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5413 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5416 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5417 regno_allocno_class_array and sorted_pseudos.
5418 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5419 insns where regno is used.
5421 2018-03-16 Martin Liska <mliska@suse.cz>
5424 * multiple_target.c (create_dispatcher_calls): Redirect
5425 reference in the symbol table.
5427 2018-03-16 Martin Liska <mliska@suse.cz>
5430 * multiple_target.c (create_dispatcher_calls): Redirect also
5433 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5437 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5438 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5439 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5441 2018-03-16 Julia Koval <julia.koval@intel.com>
5443 * doc/invoke.texi (Skylake Server): Add CLWB.
5444 Cannonlake): Remove CLWB.
5446 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5448 PR tree-optimization/84841
5449 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5451 (FLOAT_ONE_CONST_TYPE): Define.
5452 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5453 (sort_by_operand_rank): Put entries with higher constant_type last
5454 rather than first to match comments.
5456 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5458 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5461 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5464 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5465 check and use address of target always.
5467 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5470 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5471 (indirect_thunk_bnd_needed): Likewise.
5472 (indirect_thunks_used): Likewise.
5473 (indirect_thunks_bnd_used): Likewise.
5474 (indirect_return_needed): New.
5475 (indirect_return_bnd_needed): Likewise.
5476 (output_indirect_thunk_function): Add a bool argument for
5478 (output_indirect_thunk_function): Don't generate alias for
5479 function return thunk.
5480 (ix86_code_end): Call output_indirect_thunk_function to generate
5481 function return thunks.
5482 (ix86_output_function_return): Set indirect_return_bnd_needed
5483 and indirect_return_needed instead of indirect_thunk_bnd_needed
5484 and indirect_thunk_needed.
5486 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5488 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5489 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5490 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5492 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5493 Paul Hua <paul.hua.gm@gmail.com>
5496 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5498 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5500 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5501 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5504 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5507 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5508 instead of GET_MODE_SIZE when comparing Units.
5510 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5513 * varasm.c (hash_section): Return an unchangeble hash value
5514 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5515 Return !aarch64_can_use_per_function_literal_pools_p ().
5517 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5520 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5521 rather than address of cmode as last argument to prepare_cmp_insn.
5523 2018-03-15 Julia Koval <julia.koval@intel.com>
5525 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5526 F_AVX512VNNI, F_AVX512BITALG): New.
5528 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5531 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5532 insn for floating-point loads and stores.
5534 2018-03-14 Carl Love <cel@us.ibm.com>
5536 * config/rs6000/rs6000-c.c: Add macro definitions for
5537 ALTIVEC_BUILTIN_VEC_PERMXOR.
5538 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5539 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5540 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5541 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5543 * config/doc/extend.texi: Add prototypes for vec_permxor.
5545 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5548 * diagnostic-show-locus.c (class layout_point): Convert m_line
5549 from int to linenum_type.
5550 (line_span::comparator): Use linenum "compare" function when
5551 comparing line numbers.
5552 (test_line_span): New function.
5553 (layout_range::contains_point): Convert param "row" from int to
5555 (layout_range::intersects_line_p): Likewise.
5556 (layout::will_show_line_p): Likewise.
5557 (layout::print_source_line): Likewise.
5558 (layout::should_print_annotation_line_p): Likewise.
5559 (layout::print_annotation_line): Likewise.
5560 (layout::print_leading_fixits): Likewise.
5561 (layout::annotation_line_showed_range_p): Likewise.
5562 (struct line_corrections): Likewise for field m_row.
5563 (line_corrections::line_corrections): Likewise for param "row".
5564 (layout::print_trailing_fixits): Likewise.
5565 (layout::get_state_at_point): Likewise.
5566 (layout::get_x_bound_for_row): Likewise.
5567 (layout::print_line): Likewise.
5568 (diagnostic_show_locus): Likewise for locals "last_line" and
5570 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5571 * input.c (selftest::test_linenum_comparisons): New function.
5572 (selftest::input_c_tests): Call it.
5573 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5574 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5575 * selftest.h (ASSERT_GT): New macro.
5576 (ASSERT_GT_AT): New macro.
5577 (ASSERT_LT): New macro.
5578 (ASSERT_LT_AT): New macro.
5580 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5582 PR rtl-optimization/84780
5583 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5585 2018-03-14 Martin Liska <mliska@suse.cz>
5587 * tree.c (record_node_allocation_statistics): Use
5588 get_stats_node_kind.
5589 (get_stats_node_kind): New function extracted from
5590 record_node_allocation_statistics.
5591 (free_node): Use get_stats_node_kind.
5593 2018-03-14 Richard Biener <rguenther@suse.de>
5595 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5596 that the value-set of ANTIC_IN doesn't grow.
5599 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5601 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5602 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5604 2018-03-14 Julia Koval <julia.koval@intel.com>
5606 * config.gcc (icelake-client, icelake-server): New.
5608 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5609 (initial_ix86_arch_features): Ditto.
5610 (PTA_SKYLAKE): Add SGX.
5611 (PTA_ICELAKE): Remove.
5612 (PTA_ICELAKE_CLIENT): New.
5613 (PTA_ICELAKE_SERVER): New.
5614 (ix86_option_override_internal): Split up icelake on icelake client and
5616 (get_builtin_code_for_version): Ditto.
5617 (fold_builtin_cpu): Ditto.
5618 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5619 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5620 * config/i386/i386.h (processor_type): Ditto.
5621 * doc/invoke.texi: Ditto.
5623 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5626 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5627 INTEGER_CST offset, add it together with bitpos / 8 and
5628 sign extend based on POINTER_SIZE.
5632 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5635 * config/i386/constraints.md (Yc): New register constraint.
5636 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5637 Use Yc constraint for alternative 2 of operand 0. Remove
5638 preferred_for_speed attribute.
5640 2018-03-14 Richard Biener <rguenther@suse.de>
5642 PR tree-optimization/84830
5643 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5644 with the old one to avoid oscillations.
5646 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5649 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5651 (assign_by_spills): Return a flag of reload assignment failure.
5652 Do not process the reload assignment failures. Do not spill other
5653 reload pseudos if they has the same reg class. Update n if
5655 (lra_assign): Add a return arg. Set up from the result of
5656 assign_by_spills call.
5657 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5658 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5659 usage_insns if it is not NULL.
5660 (spill_hard_reg_in_range): New function.
5661 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5662 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5663 function prototypes.
5664 (lra_assign): Change prototype.
5665 * lra.c (lra): Add code to deal with fails by splitting hard reg
5668 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5670 * config/riscv/riscv.opt (mrelax): New option.
5671 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5672 "norelax" when riscv_mrelax is disabled.
5673 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5675 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5678 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5679 reassociation for int modes.
5681 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5683 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5684 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5686 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5687 * config/aarch64/aarch64-sve.md
5688 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5689 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5690 (*extend<mode><Vwide>2): Rename to...
5691 (aarch64_sve_extend<mode><Vwide>2): ...this.
5692 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5693 renaming the old pattern to...
5694 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5696 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5697 define_expand, renaming the old pattern to...
5698 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5699 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5700 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5701 account when deciding which SVE instruction the optab should use.
5702 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5704 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5706 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5707 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5708 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5709 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5710 (tlsdesc_small_<mode>): Turn a define_expand and use
5711 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5712 (tlsdesc_small_advsimd_<mode>): ...this.
5713 (tlsdesc_small_sve_<mode>): New pattern.
5715 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5717 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5718 (UNSPEC_UMUL_HIGHPART): New constants.
5719 (MUL_HIGHPART): New int iteraor.
5720 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5721 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5723 (*<su>mul<mode>3_highpart): New define_insn.
5725 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5728 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5731 2018-03-13 Martin Liska <mliska@suse.cz>
5734 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5736 (sem_item_optimizer::~sem_item_optimizer): Release it.
5737 (sem_item_optimizer::merge_classes): Register variable aliases.
5738 (sem_item_optimizer::fixup_pt_set): New function.
5739 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5740 * ipa-icf.h: Declare new variables and functions.
5742 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5745 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5746 integer_pow2p@2 and test integer_pow2p in condition.
5747 (A < 0 ? C : 0): Similarly for @1.
5750 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5751 characters starting at p contain '\0' character, don't look beyond
5755 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5756 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5759 * reg-stack.c (change_stack): Change update_end var from int to
5760 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5761 also call set_block_for_insn on the newly added insns and rescan.
5764 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5765 on the last operand.
5768 * tree.c (stabilize_reference_1): Return save_expr (e) for
5769 STATEMENT_LIST even if it doesn't have side-effects.
5771 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5773 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5775 2018-03-12 Renlin Li <renlin.li@arm.com>
5777 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5778 aarch64_output_scalar_simd_mov_immediate.
5780 2018-03-12 Martin Sebor <msebor@redhat.com>
5782 PR tree-optimization/83456
5783 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5784 for perfectly overlapping calls to memcpy.
5785 (gimple_fold_builtin_memory_chk): Same.
5786 (gimple_fold_builtin_strcpy): Handle no-warning.
5787 (gimple_fold_builtin_stxcpy_chk): Same.
5788 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5790 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5792 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5793 parameter. Use it for SFmode.
5794 (rs6000_function_arg_advance_1): Adjust.
5795 (rs6000_function_arg): Adjust.
5796 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5798 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5800 PR rtl-optimization/84169
5801 PR rtl-optimization/84780
5802 * combine.c (can_combine_p): Check for a 2-insn combination whether
5803 the destination register is used between the two insns, too.
5805 2018-03-12 Richard Biener <rguenther@suse.de>
5807 PR tree-optimization/84803
5808 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5809 for refs DR analysis didn't process.
5811 2018-03-12 Richard Biener <rguenther@suse.de>
5813 PR tree-optimization/84777
5814 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5815 force-vectorize loops ignore whether we are optimizing for size.
5817 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5819 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5820 (TARGET_MD_ASM_ADJUST): Define.
5822 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5823 Kito Cheng <kito.cheng@gmail.com>
5824 Chung-Ju Wu <jasonwucj@gmail.com>
5826 * config/nds32/nds32.c (nds32_compute_stack_frame,
5827 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5828 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5829 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5830 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5831 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5832 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5833 * config/nds32/nds32.md (prologue, epilogue): Use macro
5834 NDS32_V3PUSH_AVAILABLE_P to do checking.
5836 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5839 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5840 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5841 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5842 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5843 addition of most attributes on !orig_type_die or the attribute not
5844 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5846 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5847 Chung-Ju Wu <jasonwucj@gmail.com>
5849 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5851 * config/nds32/nds32.opt (mvh): New option.
5853 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5854 Chung-Ju Wu <jasonwucj@gmail.com>
5856 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5858 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5859 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5862 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5863 Chung-Ju Wu <jasonwucj@gmail.com>
5865 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5867 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5868 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5870 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5871 Kito Cheng <kito.cheng@gmail.com>
5872 Chung-Ju Wu <jasonwucj@gmail.com>
5874 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5875 UNSPEC_FFMISM and UNSPEC_FLMISM.
5876 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5877 for ffb, ffmism and flmism.
5878 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5879 (unspec_ffmism): Ditto.
5880 (unspec_flmism): Ditto.
5881 (nds32_expand_builtin_impl): Check if string extension is available.
5882 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5883 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5885 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5888 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5891 * lra-assigns.c (assign_by_spills): Return a flag of reload
5892 assignment failure. Do not process the reload assignment
5893 failures. Do not spill other reload pseudos if they has the same
5895 (lra_assign): Add a return arg. Set up from the result of
5896 assign_by_spills call.
5897 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5898 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5899 usage_insns if it is not NULL.
5900 (spill_hard_reg_in_range): New function.
5901 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5902 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5903 function prototypes.
5904 (lra_assign): Change prototype.
5905 * lra.c (lra): Add code to deal with fails by splitting hard reg
5908 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5911 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5913 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5916 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5917 (dw_val_node): Add val_symbolic_view.
5918 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5919 (symview_upper_bound): New.
5920 (new_line_info_table): Initialize symviews_since_reset.
5921 (dwarf2out_source_line): Count symviews_since_reset and set
5922 symview_upper_bound.
5923 (dw_val_equal_p): Handle symview.
5924 (add_AT_symview): New.
5925 (print_dw_val): Handle symview.
5926 (attr_checksum, attr_checksum_ordered): Likewise.
5927 (same_dw_val_p, size_of_die): Likewise.
5928 (value_format, output_die): Likewise.
5929 (add_high_low_attributes): Use add_AT_symview for entry_view.
5930 (dwarf2out_finish): Reset symview_upper_bound, clear
5933 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5936 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5937 Add strict argument and use it.
5938 (rs6000_split_multireg_move): Update for new strict argument.
5939 (mem_operand_gpr): Disallow all non-offsettable addresses.
5940 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5942 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5945 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5946 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5947 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5950 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5951 decl, use remap_type if we want to use the type.
5953 2018-03-09 Martin Sebor <msebor@redhat.com>
5955 PR tree-optimization/84526
5956 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5958 (builtin_access::generic_overlap): Be prepared to handle non-array
5961 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5963 PR rtl-optimization/84682
5964 * lra-constraints.c (process_address_1): Check is_address flag
5965 for address constraints.
5966 (process_alt_operands): Likewise.
5967 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5968 preprocess_constraints.
5969 * recog.h (preprocess_constraints): Add oploc parameter.
5971 * recog.c (preprocess_constraints): Test address_operand for
5972 CT_ADDRESS constraints.
5974 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5977 * lra-assigns.c (assign_by_spills): Return a flag of reload
5978 assignment failure. Do not process the reload assignment
5979 failures. Do not spill other reload pseudos if they has the same
5981 (lra_assign): Add a return arg. Set up from the result of
5982 assign_by_spills call.
5983 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5984 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5985 usage_insns if it is not NULL.
5986 (spill_hard_reg_in_range): New function.
5987 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5988 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5989 function prototypes.
5990 (lra_assign): Change prototype.
5991 * lra.c (lra): Add code to deal with fails by splitting hard reg
5994 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5997 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5998 Accept complain bool parameter. Only emit errors if it is true.
5999 (arm_parse_cpu_option_name): Likewise.
6000 (arm_target_thumb_only): Adjust callers of the above.
6001 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6002 prototype to take a default true bool parameter.
6003 (arm_parse_arch_option_name): Likewise.
6005 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6006 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6010 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6011 * configure: Regenerate.
6012 * configure.ac ("linker --version-script option"): New.
6013 ("linker soname option"): New.
6015 2018-03-09 Richard Biener <rguenther@suse.de>
6017 PR tree-optimization/84775
6018 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6019 immediate uses of predicate stmts and mark them modified.
6022 PR tree-optimization/84178
6023 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6025 (version_loop_for_if_conversion): Delay update_ssa call.
6026 (tree_if_conversion): Delay update_ssa until after predicate
6029 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6032 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6033 when the function accesses prior frames.
6035 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6038 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6039 gen_llsym, otherwise call maybe_gen_llsym.
6042 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6043 has ',' character inside of it.
6045 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6048 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6049 as clobbering CC_REGNUM.
6051 2018-03-08 Richard Biener <rguenther@suse.de>
6054 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6055 (follow_copies_to_constant): Do not follow SSA names registered
6058 2018-03-08 Richard Biener <rguenther@suse.de>
6060 PR tree-optimization/84178
6061 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6063 (version_loop_for_if_conversion): Delay update_ssa call.
6064 (tree_if_conversion): Delay update_ssa until after predicate
6067 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6069 PR tree-optimization/84178
6070 * tree-if-conv.c (release_bb_predicate): Remove the
6071 the assertion that the stmts have NULL use_ops.
6072 Discard the statements, asserting that they haven't
6073 yet been added to a BB.
6075 2018-03-08 Richard Biener <rguenther@suse.de>
6077 PR tree-optimization/84746
6078 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6079 (phi_translate): Pass in destination ANTIC_OUT set.
6080 (phi_translate_1): Likewise. For a simplified result lookup
6081 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6082 (phi_translate_set): Adjust.
6083 (do_pre_regular_insertion): Likewise.
6084 (do_pre_partial_partial_insertion): Likewise.
6086 2018-03-08 Martin Liska <mliska@suse.cz>
6088 PR gcov-profile/84735
6089 * doc/gcov.texi: Document usage of profile files.
6090 * gcov-io.h: Document changes in the format.
6092 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6096 * dwarf2out.c (struct dw_line_info_table): Update comments for
6098 (FORCE_RESET_NEXT_VIEW): New.
6099 (FORCE_RESETTING_VIEW_P): New.
6100 (RESETTING_VIEW_P): Check for -1 too.
6101 (ZERO_VIEW_P): Likewise.
6102 (new_line_info_table): Force-reset next view.
6103 (dwarf2out_begin_function): Likewise.
6104 (dwarf2out_source_line): Simplify zero_view_p initialization.
6105 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6106 view directly. Omit view when omitting .loc at line 0.
6108 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6110 PR tree-optimization/84740
6111 * tree-switch-conversion.c (process_switch): Call build_constructors
6112 only if info.phi_count is non-zero.
6114 PR tree-optimization/84739
6115 * tree-tailcall.c (find_tail_calls): Check call arguments against
6116 DECL_ARGUMENTS (current_function_decl) rather than
6117 DECL_ARGUMENTS (func) when checking for tail recursion.
6119 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6121 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6122 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6123 Volker Reichelt's entry and add entries for people that perform
6124 GCC fuzzy testing and report numerous bugs.
6126 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6129 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6130 readonly data in sdata, if that is disabled.
6131 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6132 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6133 -mreadonly-in-sdata option.
6135 2018-03-07 Martin Sebor <msebor@redhat.com>
6137 PR tree-optimization/84468
6138 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6139 basic block when looking for nul assignment.
6141 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6144 * except.h (output_function_exception_table): Adjust prototype.
6145 * except.c (output_function_exception_table): Remove FNNAME parameter
6146 and add SECTION parameter. Ouput one part of the table at a time.
6147 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6148 the first part of the exception table and emit unwind directives.
6149 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6150 (i386_pe_seh_cold_init): Likewise.
6151 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6152 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6153 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6154 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6155 * config/i386/winnt.c: Include except.h.
6156 (struct seh_frame_state): Add reg_offset, after_prologue and
6157 in_cold_section fields.
6158 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6159 (i386_pe_seh_cold_init): New function.
6160 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6161 to seh->in_cold_section.
6162 (seh_emit_push): Record the offset of the push.
6163 (seh_emit_save): Record the offet of the save.
6164 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6165 Test seh->after_prologue to disregard the epilogue.
6166 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6167 (i386_pe_end_cold_function): New function.
6169 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6172 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6173 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6176 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6178 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6179 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6182 * multiple_target.c: Include tree-inline.h and intl.h.
6183 (expand_target_clones): Diagnose and fail if node->definition and
6184 !tree_versionable_function_p (node->decl).
6186 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6188 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6190 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6191 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6192 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6194 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6197 * combine.c (try_combine): Use reg_or_subregno instead of handling
6198 just paradoxical SUBREGs and REGs.
6200 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6202 * config/arc/arc.c (arc_finalize_pic): Remove function.
6203 (arc_must_save_register): We use single base PIC register, remove
6204 checks to save/restore the PIC register.
6205 (arc_expand_prologue): Likewise.
6206 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6208 (arc_verify_short): Likewise.
6209 (arc_attr_type): Likewise.
6210 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6211 (walk_stores): Likewise.
6212 (arc_address_cost): Make it static.
6213 (arc_verify_short): Likewise.
6214 (branch_dest): Likewise.
6215 (arc_attr_type): Likewise.
6216 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6217 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6218 (arc_final_prescan_insn): Remove inserting the nops due to
6219 hardware hazards. It is done in reorg step.
6220 (insn_length_variant_t): Remove.
6221 (insn_length_parameters_t): Likewise.
6222 (arc_insn_length_parameters): Likewise.
6223 (arc_get_insn_variants): Likewise.
6224 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6226 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6229 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6232 PR tree-optimization/84687
6233 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6235 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6237 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6239 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6240 Rename to ppc_speculation_barrier.
6241 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6242 __builtin_ppc_speculation_barrier.
6244 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6247 * combine.c (combine_simplify_rtx): Don't try to simplify if
6248 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6251 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6253 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6254 to 32 bytes when compiling for POWER9.
6256 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6259 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6260 regparm >= 3 with no arg reg available also for calls with
6261 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6264 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6266 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6268 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6271 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6273 2018-03-05 Richard Biener <rguenther@suse.de>
6275 PR tree-optimization/84486
6276 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6277 When inserting a __builtin_assume_aligned call set the LHS
6278 SSA name alignment info accordingly.
6280 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6282 PR tree-optimization/84114
6283 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6284 Avoid reassociation of FLOAT_MODE addition.
6286 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6288 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6289 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6290 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6291 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6292 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6293 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6294 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6296 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6297 __builtin_ia32_wbinvd): New builtins.
6298 (SPECIAL_ARGS2): New.
6299 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6300 (SPECIAL_ARGS2): New.
6301 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6302 (ix86_valid_target_attribute_inner_p): Ditto.
6303 (ix86_init_mmx_sse_builtins): Add special_args2.
6304 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6305 TARGET_WBNOINVD_P): New.
6306 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6307 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6308 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6309 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6310 * config/i386/pconfigintrin.h: New file.
6311 * config/i386/wbnoinvdintrin.h: Ditto.
6312 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6313 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6315 2018-03-05 Richard Biener <rguenther@suse.de>
6317 PR tree-optimization/84670
6318 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6320 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6321 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6322 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6323 doesn't grow if all successors (recursively) were visited at least
6326 2018-03-05 Richard Biener <rguenther@suse.de>
6328 PR tree-optimization/84650
6329 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6330 if executed in the loop pipeline.
6332 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6334 * doc/configfiles.texi (Configuration Files): Move info about
6335 conditionalizing $target-protos.h to...
6336 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6337 differs from $target-protos.h.
6339 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6340 Chung-Ju Wu <jasonwucj@gmail.com>
6342 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6343 * config/nds32/nds32-multiple.md (setmemsi): Define.
6344 * config/nds32/nds32-memory-manipulation.c
6345 (nds32_gen_dup_4_byte_to_word_value): New.
6346 (emit_setmem_word_loop): New.
6347 (emit_setmem_byte_loop): New.
6348 (nds32_expand_setmem_loop): New.
6349 (nds32_expand_setmem_loop_v3m): New.
6350 (nds32_expand_setmem_unroll): New.
6351 (nds32_expand_setmem): New.
6353 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6354 Chung-Ju Wu <jasonwucj@gmail.com>
6356 * config/nds32/nds32-memory-manipulation.c
6357 (nds32_emit_load_store): New.
6358 (nds32_emit_post_inc_load_store): New.
6359 (nds32_emit_mem_move): New.
6360 (nds32_emit_mem_move_block): New.
6361 (nds32_expand_movmemsi_loop_unknown_size): New.
6362 (nds32_expand_movmemsi_loop_known_size): New.
6363 (nds32_expand_movmemsi_loop): New.
6364 (nds32_expand_movmemsi_unroll): New.
6365 (nds32_expand_movmemqi): Rename ...
6366 (nds32_expand_movmemsi): ... to this.
6367 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6368 (movmemsi): ... to this.
6369 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6370 (nds32_expand_movmemsi): ... to this.
6372 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6373 Monk Chiang <sh.chiang04@gmail.com>
6374 Chung-Ju Wu <jasonwucj@gmail.com>
6376 * config/nds32/nds32-protos.h
6377 (nds32_expand_load_multiple): New arguments.
6378 (nds32_expand_store_multiple): Ditto.
6379 (nds32_valid_multiple_load_store): Rename ...
6380 (nds32_valid_multiple_load_store_p): ... to this.
6381 * config/nds32/nds32-memory-manipulation.c
6382 (nds32_expand_load_multiple): Refine implementation.
6383 (nds32_expand_store_multiple): Ditto.
6384 * config/nds32/nds32-multiple.md
6385 (load_multiple): Update nds32_expand_load_multiple interface.
6386 (store_multiple): Update nds32_expand_store_multiple interface.
6387 * config/nds32/nds32-predicates.c
6388 (nds32_valid_multiple_load_store): Rename ...
6389 (nds32_valid_multiple_load_store_p): ... to this and refine
6391 * config/nds32/predicates.md
6392 (nds32_load_multiple_and_update_address_operation): New predicate.
6393 (nds32_store_multiple_and_update_address_operation): New predicate.
6395 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6396 Chung-Ju Wu <jasonwucj@gmail.com>
6398 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6399 (combo): New attribute.
6400 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6402 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6404 * config/nds32/nds32.opt: Change -mcmodel= default value.
6406 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6407 Monk Chiang <sh.chiang04@gmail.com>
6408 Chung-Ju Wu <jasonwucj@gmail.com>
6410 * config/nds32/constants.md (unspec_element): New enum.
6411 * config/nds32/constraints.md (Umw): New constraint.
6412 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6413 * config/nds32/nds32-intrinsic.md: Likewise.
6414 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6415 (nds32_valid_smw_lwm_base_p): New.
6416 (nds32_output_smw_single_word): New.
6417 (nds32_output_lmw_single_word): New.
6418 (nds32_expand_unaligned_load): New.
6419 (nds32_expand_unaligned_store): New.
6420 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6421 (nds32_output_smw_single_word): Declare.
6422 (nds32_output_lmw_single_word): Declare.
6423 (nds32_expand_unaligned_load): Declare.
6424 (nds32_expand_unaligned_store): Declare.
6425 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6426 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6427 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6428 NDS32_BUILTIN_UASTORE_DW.
6429 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6432 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6433 Kito Cheng <kito.cheng@gmail.com>
6434 Chung-Ju Wu <jasonwucj@gmail.com>
6436 * config/nds32/nds32-intrinsic.c
6437 (nds32_expand_builtin_null_ftype_reg): Delete.
6438 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6439 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6440 (nds32_read_argument): New.
6441 (nds32_legitimize_target): Ditto.
6442 (nds32_legitimize_argument): Ditto.
6443 (nds32_check_constant_argument): Ditto.
6444 (nds32_expand_unop_builtin): Ditto.
6445 (nds32_expand_unopimm_builtin): Ditto.
6446 (nds32_expand_binop_builtin): Ditto.
6447 (nds32_builtin_decl_impl): Ditto.
6448 (builtin_description): Ditto.
6449 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6450 (nds32_init_builtins_impl): Ditto.
6451 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6452 (nds32_builtin_decl): New.
6453 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6454 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6456 2018-03-02 Jeff Law <law@redhat.com>
6458 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6459 (redundant_insn, fill_simple_delay_slots): Likewise.
6460 (fill_slots_from_thread): Likewise.
6461 * resource.c (mark_referenced_resources): Likewise.
6462 (mark_set_resources, find_dead_or_set_registers): Likewise.
6464 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6466 * substring-locations.h (format_warning_va): Formatting fix for
6468 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6470 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6471 * substring-locations.c: Include intl.h.
6472 (format_warning_va): Turned into small wrapper around
6473 format_warning_n_va, renamed to ...
6474 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6475 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6477 (format_warning_at_substring_n): New function.
6478 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6479 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6480 format_warning_at_substring with just a shorter name instead of
6481 const function pointer.
6482 (fmtwarn_n): New function.
6483 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6484 appropriate, get rid of all the fmtstr temporaries, move conditionals
6485 with G_() wrapped string literals directly into fmtwarn arguments,
6486 cast dir.len to (int), formatting fixes.
6488 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6490 * doc/invoke.texi: Remove "Cilk Plus" references.
6492 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6493 Richard Biener <rguenther@suse.de>
6496 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6497 for error or warning attributes if CALL_FROM_THUNK_P is set.
6500 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6503 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6504 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6507 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6508 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6510 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6511 instead of -1U in last predictors element's probability member.
6513 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6516 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6517 arguments if they are comparable.
6519 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6521 PR tree-optimization/84634
6522 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6523 masks and masked_loop_p with a single loop_masks, making sure it's
6524 null for bb vectorization.
6526 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6528 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6529 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6530 loop->force_vectorize to check whether there is no alias.
6532 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6535 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6537 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6539 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6540 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6541 instead of a loop around prev_real_insn.
6542 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6546 * config/i386/i386.c (ix86_print_operand): Use conditional
6547 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6550 2018-03-02 Richard Biener <rguenther@suse.de>
6552 PR tree-optimization/84427
6553 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6554 (bitmap_set_subtract_values): Rewrite to handle multiple
6557 (prune_clobbered_mems): Likewise.
6558 (phi_translate): Take edge instead of pred/phiblock.
6559 (phi_translate_1): Likewise.
6560 (phi_translate_set): Likewise. Insert all translated
6561 exprs for a value into the set, keeping possibly multiple
6562 expressions per value.
6563 (compute_antic_aux): Adjust for phi_translate changes.
6564 When intersecting union the expressions and prune those
6565 not in the final value set, keeping possibly multiple
6566 expressions per value. Do not use value-insertion
6567 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6568 all expressions. Add verification that the value-sets
6569 only shrink during iteration.
6570 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6571 (do_pre_regular_insertion): Likewise.
6572 (do_pre_partial_partial_insertion): Likewise.
6574 2018-03-02 Richard Biener <rguenther@suse.de>
6577 * config/darwin.c (saved_debug_info_level): New static global.
6578 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6579 (darwin_asm_lto_end): Restore debug info generation settings.
6581 2018-03-01 Martin Liska <mliska@suse.cz>
6584 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6587 2018-03-01 Richard Biener <rguenther@suse.de>
6590 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6591 type annotation with LTO when debug was disabled at compile-time.
6593 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6595 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6597 (mips_final_postscan_insn): Likewise.
6599 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6601 PR rtl-optimization/84528
6602 * alias.c (init_alias_target): Add commentary.
6603 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6604 a unique base value if the frame pointer is not eliminated
6605 to the stack pointer.
6607 2018-03-01 Tom de Vries <tom@codesourcery.com>
6609 PR rtl-optimization/83327
6610 * lra-int.h (hard_regs_spilled_into): Declare.
6611 * lra.c (hard_regs_spilled_into): Define.
6612 (init_reg_info): Init hard_regs_spilled_into.
6613 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6614 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6615 (process_bb_lives): Handle hard_regs_spilled_into.
6616 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6617 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6619 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6621 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6622 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6623 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6624 * config/rs6000/aix72.h: New file.
6626 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6628 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6629 instead of warning_at with conditional singular and plural messages
6633 * stor-layout.c (update_alignment_for_field): For
6634 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6635 && !DECL_PACKED (field), do the alignment update, just use
6636 only desired_align instead of MAX (type_align, desired_align)
6638 (place_field): Don't do known_align < desired_align handling
6639 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6640 is non-NULL, instead do it after rli->prev_field handling and
6641 only if not within a bitfield word. For DECL_PACKED (field)
6642 use type_align of BITS_PER_UNIT.
6644 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6646 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6647 superfluous parentheses and trailing spaces.
6649 2018-02-28 Richard Biener <rguenther@suse.de>
6651 PR tree-optimization/84584
6652 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6653 SCoPs with fake exit edge.
6655 2018-02-28 Martin Liska <mliska@suse.cz>
6658 * timevar.c (timer::print): Fix format to properly print 100%
6661 2018-02-28 Richard Biener <rguenther@suse.de>
6664 * genmatch.c (capture_info::walk_match): Do not mark
6665 captured expressions without operands as expr_p given
6666 they act more like predicates and should be subject to
6667 "lost tail" side-effect preserving.
6669 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6671 PR rtl-optimization/81611
6672 * auto-inc-dec.c (attempt_change): Move dead note from
6673 mem_insn if it's the next use of regno
6674 (find_address): Take address use of reg holding
6675 non-incremented value. Add parm to limit search to the named
6677 (merge_in_block): Attempt to use a mem insn that is the next
6678 use of the original regno.
6680 2018-02-27 Martin Sebor <msebor@redhat.com>
6683 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6684 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6686 2018-02-27 Martin Sebor <msebor@redhat.com>
6688 PR translation/84207
6689 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6690 n argument to unsigned HOST_WIDE_INT.
6691 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6692 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6693 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6694 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6696 2018-02-27 Richard Biener <rguenther@suse.de>
6698 PR tree-optimization/84512
6699 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6700 Do not use the estimate returned from record_stmt_cost for
6701 the scalar iteration cost but sum properly using add_stmt_cost.
6703 2018-02-27 Richard Biener <rguenther@suse.de>
6705 PR tree-optimization/84466
6706 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6707 Adjust last change to less strictly validate use operands.
6709 2018-02-27 Martin Liska <mliska@suse.cz>
6711 PR gcov-profile/84548
6712 * gcov.c (process_file): Allow partial overlap and consider it
6713 also as group functions.
6714 (output_lines): Properly calculate range of lines for a group.
6716 2018-02-27 Martin Liska <mliska@suse.cz>
6718 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6719 'ggc' suffixes. Change first column width.
6720 (timer::print): Fix formatting of the column.
6722 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6724 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6725 preserve inline entry blocks for the sake of debug inline
6726 entry point markers alone.
6727 (remove_unused_locals): Suggest in comments a better place to
6728 force the preservation of inline entry blocks that are
6729 otherwise unused, but do not preserve them.
6731 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6733 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6735 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6738 * config/i386/constraints.md (Bs): Replace
6739 ix86_indirect_branch_register with
6740 TARGET_INDIRECT_BRANCH_REGISTER.
6742 * config/i386/i386.md (indirect_jump): Likewise.
6743 (tablejump): Likewise.
6744 (*sibcall_memory): Likewise.
6745 (*sibcall_value_memory): Likewise.
6746 Peepholes of indirect call and jump via memory: Likewise.
6747 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6748 (*sibcall_value_GOT_32): Likewise.
6749 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6750 (GOT_memory_operand): Likewise.
6751 (call_insn_operand): Likewise.
6752 (sibcall_insn_operand): Likewise.
6753 (GOT32_symbol_operand): Likewise.
6754 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6756 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6758 PR rtl-optimization/83496
6759 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6760 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6761 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6762 redundant insn, if any.
6763 (relax_delay_slots): Likewise.
6764 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6766 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6768 PR tree-optimization/83965
6769 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6770 that grouped statements are part of a reduction chain. Return
6771 true if the statement is not marked as a reduction itself but
6773 (vect_recog_dot_prod_pattern): Don't check whether the statement
6774 is part of a group here.
6775 (vect_recog_sad_pattern): Likewise.
6776 (vect_recog_widen_sum_pattern): Likewise.
6778 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6781 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6783 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6786 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6788 (ix86_output_indirect_function_return): New prototype.
6789 (ix86_split_simple_return_pop_internal): Likewise.
6790 * config/i386/i386.c (indirect_return_via_cx): New.
6791 (indirect_return_via_cx_bnd): Likewise.
6792 (indirect_thunk_name): Handle return va CX_REG.
6793 (output_indirect_thunk_function): Create alias for
6794 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6795 (ix86_output_indirect_jmp): Remove the bool argument.
6796 (ix86_output_indirect_function_return): New function.
6797 (ix86_split_simple_return_pop_internal): Likewise.
6798 * config/i386/i386.md (*indirect_jump): Don't pass false
6799 to ix86_output_indirect_jmp.
6800 (*tablejump_1): Likewise.
6801 (simple_return_pop_internal): Change it to define_insn_and_split.
6802 Call ix86_split_simple_return_pop_internal to split it for
6804 (simple_return_indirect_internal): Call
6805 ix86_output_indirect_function_return instead of
6806 ix86_output_indirect_jmp.
6808 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6811 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6812 memset and value initialization afterwards.
6814 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6816 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6818 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6821 * common/config/aarch64/aarch64-common.c
6822 (aarch_option_optimization_table[]): Switch
6823 off fomit-frame-pointer
6825 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6826 Chung-Ju Wu <jasonwucj@gmail.com>
6828 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6830 (store_multiple): Ditto.
6832 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6834 * config.gcc: Add --with-cpu support for nds32 target.
6835 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6836 * config/nds32/nds32.opt: Add -mcpu= option.
6838 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6840 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6841 isel=yes): Warn for these deprecated options.
6843 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6845 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6846 ISA_2_5_MASKS_EMBEDDED.
6848 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6850 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6851 p->max as pointers rather than using iterative_hash_expr.
6853 2018-02-23 Carl Love <cel@us.ibm.com>
6855 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6856 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6858 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6859 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6860 P8V_BUILTIN_VEC_VUNSIGNED2.
6862 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6865 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6866 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6868 * lra-constraints.c (curr_insn_transform): Set up
6869 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6871 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6872 * lra-eliminations.c (spill_pseudos): Ditto.
6873 (process_insn_for_elimination): Ditto.
6874 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6875 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6876 LRA_NON_CLOBBERED_ALT.
6878 2018-02-22 Martin Sebor <msebor@redhat.com>
6880 PR tree-optimization/84480
6881 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6882 to maybe_diag_stxncpy_trunc. Call it.
6883 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6884 from gimple_fold_builtin_strcpy. Print inlining stack.
6885 (handle_builtin_stxncpy): Print inlining stack.
6886 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6888 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6891 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6892 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6893 and -fcheck-pointer-bounds are used together.
6894 (indirect_thunk_prefix): New enum.
6895 (indirect_thunk_need_prefix): New function.
6896 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6897 "_nt" instead of "_bnd" for NOTRACK prefix.
6898 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6899 (output_indirect_thunk_function): Likewise.
6901 (ix86_code_end): Update output_indirect_thunk_function calls.
6902 (ix86_output_indirect_branch_via_reg): Replace
6903 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6904 (ix86_output_indirect_branch_via_push): Likewise.
6905 (ix86_output_function_return): Likewise.
6906 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6907 incompatible with -fcf-protection=branch and
6908 -fcheck-pointer-bounds.
6910 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6913 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6914 Change gcc_assert call to output_operand_lossage.
6916 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6918 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6920 2018-02-22 DJ Delorie <dj@redhat.com>
6921 Sebastian Perta <sebastian.perta@renesas.com>
6922 Oleg Endo <olegendo@gcc.gnu.org>
6924 * config/rx/rx.c (rx_rtx_costs): New function.
6925 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6927 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6929 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6931 2018-02-22 Martin Liska <mliska@suse.cz>
6934 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6935 Add "native" as a possible value.
6937 2018-02-22 Martin Liska <mliska@suse.cz>
6940 * config/i386/i386.c (ix86_option_override_internal):
6941 Add "native" as a possible value for -march and -mtune.
6943 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6946 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6947 to all type variants.
6949 PR tree-optimization/84503
6950 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6951 width as info->bitpos + info->bitsize - start.
6952 (merged_store_group::merge_overlapping): Simplify width computation.
6953 (check_no_overlap): New function.
6954 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6955 start + width and last_order of the group, fail if check_no_overlap
6957 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6958 to group if check_no_overlap fails.
6960 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6962 * config/rs6000/altivec.md: Delete contraint arguments to
6963 define_expand, define_split, and define_peephole2, and in
6964 define_insn_and_split if always unused.
6965 * config/rs6000/darwin.md: Ditto.
6966 * config/rs6000/dfp.md: Ditto.
6967 * config/rs6000/rs6000.md: Ditto.
6968 * config/rs6000/sync.md: Ditto.
6969 * config/rs6000/vector.md: Ditto.
6970 * config/rs6000/vsx.md: Ditto.
6972 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6974 * config/rs6000/altivec.md: Write output control strings as braced
6975 blocks instead of double-quoted strings.
6976 * config/rs6000/darwin.md: Ditto.
6977 * config/rs6000/rs6000.md: Ditto.
6978 * config/rs6000/vector.md: Ditto.
6979 * config/rs6000/vsx.md: Ditto.
6981 2018-02-21 Jason Merrill <jason@redhat.com>
6983 PR c++/84314 - ICE with templates and fastcall attribute.
6984 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6986 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6988 * ipa-cp.c (determine_versionability): Fix comment typos.
6990 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6993 * ipa-cp.c (determine_versionability): Do not version functions caling
6996 2018-02-21 Martin Liska <mliska@suse.cz>
6999 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7000 Add "native" as a possible value.
7001 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7002 the macro when native cpu detection is available.
7004 2018-02-21 Martin Liska <mliska@suse.cz>
7007 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7008 Add "native" as a possible value.
7009 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7010 when native cpu detection is available.
7012 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7013 Martin Sebor <msebor@redhat.com>
7015 PR tree-optimization/84478
7016 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7018 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7019 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7020 support which is conservatively correct, for 2 only stay conservative
7021 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7022 argument to the 2 argument get_range_strlen, adjust 6 arg
7023 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7025 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7026 (gimple_fold_builtin_strlen): Pass true as last argument to
7029 2018-02-20 Martin Sebor <msebor@redhat.com>
7032 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7033 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7034 (builtin_memref::builtin_memref): Factor out parts into
7035 set_base_and_offset and call it.
7037 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7040 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7041 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7042 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7043 search at the associated MODE_INT.
7045 2018-02-20 Jeff Law <law@redhat.com>
7048 PR tree-optimization/81592
7050 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7051 for range data rather than using global data.
7052 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7053 range data rather than using global data.
7054 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7055 pass it to children as needed.
7056 (struct directive::fmtresult): Similarly.
7057 (struct directive::set_width): Similarly.
7058 (struct directive::set_precision): Similarly.
7059 (format_integer, format_directive, parse_directive): Similarly.
7060 (format_none): Accept unnamed vr_values parameter.
7061 (format_percent, format_floating, format_character): Similarly.
7062 (format_string, format_plain): Similarly.
7063 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7064 the EVRP range analyzer for range data rather than using global data.
7065 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7066 gimple-ssa-evrp-analyze.h
7067 (class sprintf_dom_walker): Add after_dom_children member function.
7068 Add evrp_range_analyzer member.
7069 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7070 range analyzer as needed.
7071 (sprintf_dom_walker::after_dom_children): New member function.
7072 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7074 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7075 (evrp_range_analyzer::pop_to_marker): Likewise.
7077 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7079 PR tree-optimization/84419
7080 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7081 with the required type if its current type is compatible but
7084 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7087 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7088 after vectorization.
7090 2018-02-20 Martin Liska <mliska@suse.cz>
7093 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7094 possible values if we don't have a hint.
7096 2018-02-20 Martin Liska <mliska@suse.cz>
7100 * final.c (shorten_branches): Build align_tab array with one
7102 * opts.c (finish_options): Add alignment option limit check.
7103 (MAX_CODE_ALIGN): Likewise.
7104 (MAX_CODE_ALIGN_VALUE): Likewise.
7105 * doc/invoke.texi: Document maximum allowed option value for
7106 all -falign-* options.
7108 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7111 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7112 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7113 * var-tracking.c (emit_note_insn_var_location): Remove all references
7114 to NOTE_INSN_CALL_ARG_LOCATION.
7115 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7116 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7117 Use copy_rtx_if_shared.
7118 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7119 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7120 (dwarf2out_var_location): Remove handling of
7121 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7123 * final.c (final_scan_insn): Remove all references to
7124 NOTE_INSN_CALL_ARG_LOCATION.
7125 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7126 before dumping final insns.
7127 * except.c (emit_note_eh_region_end): Remove all references to
7128 NOTE_INSN_CALL_ARG_LOCATION.
7129 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7130 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7131 * config/arc/arc.c (hwloop_optimize): Likewise.
7132 * config/arm/arm.c (create_fix_barrier): Likewise.
7133 * config/s390/s390.c (s390_chunkify_start): Likewise.
7134 * config/sh/sh.c (find_barrier): Likewise.
7135 * config/i386/i386.c (rest_of_insert_endbranch,
7136 ix86_seh_fixup_eh_fallthru): Likewise.
7137 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7138 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7139 * config/frv/frv.c (frv_function_prologue): Likewise.
7140 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7142 (note_outside_basic_block_p): Remove all references to
7143 NOTE_INSN_CALL_ARG_LOCATION.
7144 * gengtype.c (adjust_field_rtx_def): Likewise.
7145 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7147 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7148 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7151 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7154 PR tree-optimization/84452
7155 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7156 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7159 2018-02-19 Martin Liska <mliska@suse.cz>
7162 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7164 2018-02-19 Martin Liska <mliska@suse.cz>
7165 Richard Sandiford <richard.sandiford@linaro.org>
7167 PR tree-optimization/82491
7168 * gimple-fold.c (get_base_constructor): Make earlier bail out
7171 2018-02-19 Carl Love <cel@us.ibm.com>
7173 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7174 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7176 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7177 P8V_BUILTIN_VEC_NEG.
7179 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7181 * config/rl78/rl78.md (movdf): New define expand.
7183 2018-02-19 Martin Liska <mliska@suse.cz>
7186 * doc/invoke.texi: Fix typo.
7187 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7189 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7191 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7192 handle rs6000_single_float and rs6000_double_float specially for
7195 2018-02-16 Jeff Law <law@redhat.com>
7197 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7200 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7203 * ipa-inline.c (inline_small_functions): Fix a typo.
7205 2018-02-16 Nathan Sidwell <nathan@acm.org>
7207 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7209 2018-02-16 Carl Love <cel@us.ibm.com>
7211 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7212 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7213 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7214 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7215 expansion to P8V_BUILTIN_VEC_FLOAT2.
7217 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7219 PR rtl-optimization/70023
7220 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7221 src_regno into account.
7223 2018-02-16 Carl Love <cel@us.ibm.com>
7225 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7226 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7227 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7228 * config/rs6000/rs6000.c: Remove case statements for
7229 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7230 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7231 and P9V_BUILTIN_VEC_VINSERT4B.
7232 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7233 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7234 * config/rs6000/vsx.md:
7235 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7238 2018-02-16 Carl Love <cel@us.ibm.com>
7240 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7242 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7244 * config/rs6000/rs6000-c.c: Add the definitions for
7245 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7246 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7247 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7248 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7249 definition for insert4b and define insn *insert3b_internal.
7250 * doc/extend.texi: Add documentation for vec_extract4b.
7252 2018-02-16 Nathan Sidwell <nathan@acm.org>
7254 * doc/extend.texi (Backwards Compatibility): Mention friend
7255 injection. Note for-scope is deprecated.
7256 * doc/invoke.texi (-ffriend-injection): Deprecate.
7258 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7260 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7261 that moved to I2, also allow destinations that are a paradoxical
7262 subreg (instead of a normal reg).
7264 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7267 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7270 2018-02-16 Richard Biener <rguenther@suse.de>
7272 PR tree-optimization/84037
7273 PR tree-optimization/84016
7275 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7276 Adjust vec_construct for the fact we need additional higher latency
7277 128bit inserts for AVX256 and AVX512 vector builds.
7278 (ix86_add_stmt_cost): Scale vector construction cost for
7281 2018-02-16 Richard Biener <rguenther@suse.de>
7283 PR tree-optimization/84417
7284 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7285 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7286 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7288 2018-02-16 Martin Liska <mliska@suse.cz>
7291 * internal-fn.def (ASAN_CHECK): Set proper flags.
7292 (ASAN_MARK): Likewise.
7294 2018-02-16 Julia Koval <julia.koval@intel.com>
7296 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7297 from PTA_CANNONLAKE.
7299 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7302 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7303 Use ++iter rather than iter++ for std::list iterators.
7304 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7305 defer deleting them until all nodes in the forest are processed. Do
7306 free even leaf nodes. Change to_process into auto_vec.
7309 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7310 * vec.h (vec_default_construct): Use memset instead of placement new
7311 if BROKEN_VALUE_INITIALIZATION is defined.
7312 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7313 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7316 PR rtl-optimization/83723
7317 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7318 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7319 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7321 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7323 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7325 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7327 PR rtl-optimization/81443
7328 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7329 from inner REGs to paradoxical SUBREGs.
7331 2018-02-16 Richard Biener <rguenther@suse.de>
7333 PR tree-optimization/84399
7334 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7335 For operands we can analyze at their definition make sure we can
7336 analyze them at each use as well.
7338 2018-02-16 Richard Biener <rguenther@suse.de>
7340 PR tree-optimization/84190
7341 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7342 volatile accesses if the decl isn't volatile.
7344 2018-02-15 Jason Merrill <jason@redhat.com>
7346 PR c++/84314 - ICE with templates and fastcall attribute.
7347 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7348 TYPE_CANONICAL on an existing type.
7350 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7352 PR tree-optimization/84383
7353 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7354 dstoff nor call operand_equal_p if dstbase is NULL.
7356 PR tree-optimization/84334
7357 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7358 also a CONSTANT_CLASS_P, punt.
7360 2018-02-14 Jim Wilson <jimw@sifive.com>
7362 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7363 first SMALL_OPERAND check. New local min_second_step. Move assert
7364 to where locals are set. Add TARGET_RVC support.
7365 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7367 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7369 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7371 2018-02-14 Martin Sebor <msebor@redhat.com>
7373 PR tree-optimization/83698
7374 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7375 arrays constrain the offset range to their bounds.
7376 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7377 (builtin_access::overlap): Avoid setting the size of overlap if it's
7379 (maybe_diag_overlap): Also consider arrays when deciding what values
7380 of offsets to include in diagnostics.
7382 2018-02-14 Martin Sebor <msebor@redhat.com>
7385 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7386 that correspond to the kind of a declaration.
7388 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7391 * config/pa/pa.md: Load address of PIC label using the linkage table
7392 if the label is nonlocal.
7394 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7396 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7397 warning message if user requests -maltivec=be.
7398 * doc/invoke.texi: Document deprecation of -maltivec=be.
7400 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7403 * config/rs6000/rs6000-c.c: Update definitions for
7404 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7405 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7407 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7410 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7411 add _get_ssp intrinsics. Remove argument from
7412 __builtin_ia32_rdssp[d|q].
7413 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7414 * config/i386/i386-builtin.def: Remove argument from
7415 __builtin_ia32_rdssp[d|q].
7416 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7417 ix86_expand_special_args_builtin for _rdssp[d|q].
7418 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7419 Clear register before usage.
7420 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7421 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7423 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7425 PR tree-optimization/84357
7426 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7427 operand 1 of an ARRAY_REF too.
7429 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7432 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7433 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7435 (set_of_reg): New struct.
7436 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7437 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7438 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7440 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7441 Split into bitclr, bitset, bitinvert patterns if appropriate.
7442 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7443 use rx_fuse_in_memory_bitop.
7444 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7445 to named insn, correct maximum insn length.
7447 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7450 * machmode.def: Define a complex mode for PARTIAL_INT.
7451 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7453 * doc/rtl.texi: Document CSPImode.
7454 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7456 (msp430_hard_regno_nregs_with_padding): Likewise.
7458 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7461 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7463 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7465 PR rtl-optimization/84169
7466 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7467 we generated a parallel as new i3 and we split that to new i2 and i3
7468 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7469 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7470 those to i2, not i1. Partially rewrite this scan code.
7472 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7475 * stor-layout.c (place_field): For variable length fields, adjust
7476 offset_align afterwards not just based on the field's alignment,
7477 but also on the size.
7480 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7481 of exps and logs in the use_exp2 case.
7483 2018-02-13 Jeff Law <law@redhat.com>
7485 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7488 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7491 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7495 * common.opt (gas-loc-support, gas-locview-support): New.
7496 (ginline-points, ginternal-reset-location-views): New.
7497 * doc/invoke.texi: Document them. Use @itemx where intended.
7498 (gvariable-location-views): Adjust.
7499 * target.def (reset_location_view): New.
7500 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7501 (TARGET_RESET_LOCATION_VIEW): New.
7502 * doc/tm.texi: Rebuilt.
7503 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7504 (dwarf2out_default_as_locview_support): New.
7505 (output_asm_line_debug_info): Use option variables.
7506 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7507 (output_loc_list): Likewise.
7508 (add_high_low_attributes): Check option variables.
7509 Don't output entry view attribute in strict mode.
7510 (gen_inlined_subroutine_die): Check option variables.
7511 (dwarf2out_inline_entry): Likewise.
7512 (init_sections_and_labels): Likewise.
7513 (dwarf2out_early_finish): Likewise.
7514 (maybe_reset_location_view): New, from...
7515 (dwarf2out_var_location): ... here. Call it.
7516 * debug.h (dwarf2out_default_as_loc_support): Declare.
7517 (dwarf2out_default_as_locview_support): Declare.
7518 * hooks.c (hook_int_rtx_insn_0): New.
7519 * hooks.h (hook_int_rtx_insn_0): Declare.
7520 * toplev.c (process_options): Take -gas-loc-support and
7521 -gas-locview-support from dwarf2out. Enable
7522 -gvariable-location-views by default only with locview
7523 assembler support. Enable -ginternal-reset-location-views by
7524 default only if the target defines the corresponding hook.
7525 Enable -ginline-points by default if location views are
7526 enabled; force it disabled if statement frontiers are
7528 * tree-inline.c (expand_call_inline): Check option variables.
7529 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7531 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7533 PR tree-optimization/84321
7534 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7535 handling. Also check whether the anti-range contains any values
7536 that satisfy the mask; switch to a VR_RANGE if not.
7538 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7541 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7543 2018-02-13 Martin Jambor <mjambor@suse.cz>
7546 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7547 of call statements, also set location of a load to a temporary.
7549 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7551 * config/rl78/rl78.c (add_vector_labels): New function.
7552 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7553 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7554 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7555 which checks that no arguments are passed.
7556 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7557 * doc/extend.texi: Documentation for the new attribute.
7559 2018-02-13 Andreas Schwab <schwab@suse.de>
7561 * config/riscv/linux.h (CPP_SPEC): Define.
7563 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7566 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7567 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7568 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7569 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7570 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7571 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7572 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7573 temporarily for AES and PCLMUL builtins.
7575 PR tree-optimization/84339
7576 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7577 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7581 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7582 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7583 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7585 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7587 * omp-simd-clone.h: New file.
7588 * omp-simd-clone.c: Include omp-simd-clone.h.
7589 (expand_simd_clones): No longer static.
7590 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7591 cgraph.h and omp-simd-clone.h.
7592 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7593 (vect_recog_widen_shift_pattern): Formatting fix.
7594 (vect_pattern_recog_1): Don't check optab for calls.
7597 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7598 operands[2] into a REG before using gen_lowpart on it.
7600 2018-02-12 Jeff Law <law@redhat.com>
7603 * config/sh/sh.c (find_barrier): Consider a sibling call
7606 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7607 successfully back substituting a reg.
7609 2018-02-12 Richard Biener <rguenther@suse.de>
7611 PR tree-optimization/84037
7612 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7613 parameter, move visited init to caller.
7614 (vect_slp_analyze_operations): Separate cost from validity
7615 check, initialize visited once for all instances.
7616 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7618 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7620 (vect_model_promotion_demotion_cost): Likewise.
7621 (vectorizable_bswap): Guard cost modeling with !slp_node
7622 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7624 (vectorizable_call): Likewise.
7625 (vectorizable_conversion): Likewise.
7626 (vectorizable_assignment): Likewise.
7627 (vectorizable_shift): Likewise.
7628 (vectorizable_operation): Likewise.
7629 (vectorizable_store): Likewise.
7630 (vectorizable_load): Likewise.
7631 (vectorizable_condition): Likewise.
7632 (vectorizable_comparison): Likewise.
7634 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7637 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7638 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7639 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7641 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7644 * params.def (inline-min-speedup): Increase from 8 to 15.
7645 (max-inline-insns-auto): Decrease from 40 to 30.
7646 * ipa-split.c (consider_split): Add some buffer for function to
7647 be considered inlining candidate.
7648 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7651 2018-02-12 Richard Biener <rguenther@suse.de>
7653 PR tree-optimization/84037
7654 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7655 matched stmts if we cannot swap the non-matched ones.
7657 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7659 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7660 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7661 _mm_maskz_scalef_round_ss): New intrinsics.
7662 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7663 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7664 __builtin_ia32_scalefss_round): Remove.
7665 (__builtin_ia32_scalefsd_mask_round,
7666 __builtin_ia32_scalefss_mask_round): New intrinsics.
7667 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7668 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7669 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7670 "<round_constraint>")): Changed to ...
7671 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7672 "<round_scalar_constraint>")): ... this.
7673 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7674 %0, %1, %2<round_op3>}"): Changed to ...
7675 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7676 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7677 %2<round_scalar_mask_op3>}"): ... this.
7678 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7680 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7682 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7683 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7684 (_mm_maskz_sqrt_round_ss): New intrinsics.
7685 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7686 (__builtin_ia32_sqrtsd_mask_round)
7687 (__builtin_ia32_sqrtss_mask_round): New builtins.
7688 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7689 (__builtin_ia32_sqrtss_round): Remove.
7690 (__builtin_ia32_sqrtsd_mask_round)
7691 (__builtin_ia32_sqrtss_mask_round): New builtins.
7692 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7693 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7694 ((match_operand:VF_128 1 "vector_operand"
7695 "xBm,<round_constraint>")): Changed to ...
7696 ((match_operand:VF_128 1 "vector_operand"
7697 "xBm,<round_scalar_constraint>")): ... this.
7698 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7699 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7700 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7701 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7702 %<iptr>1<round_scalar_mask_op3>}): ... this.
7703 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7704 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7706 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7709 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7710 Cast vec_cmpeq result to correct type.
7711 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7712 Cast vec_cmpgt result to correct type.
7714 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7716 * final.c (final_scan_insn_1): Renamed from...
7717 (final_scan_insn): ... this. New wrapper, to recover
7718 seen from the outermost call in recursive ones.
7719 * config/sparc/sparc.c (output_return): Drop seen from call.
7720 (output_sibcall): Likewise.
7721 * config/visium/visium.c (output_branch): Likewise.
7723 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7725 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7728 2018-02-10 Alan Modra <amodra@gmail.com>
7731 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7732 Specify LR as an input.
7734 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7737 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7738 remove_member_access_dummy_vars): New functions.
7739 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7740 lower_omp_1, execute_lower_omp): Use them.
7742 PR rtl-optimization/84308
7743 * shrink-wrap.c (spread_components): Release todo vector.
7745 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7747 PR rtl-optimization/57193
7748 * ira-color.c (struct allocno_color_data): Add member
7749 conflict_allocno_hard_prefs.
7750 (update_conflict_allocno_hard_prefs): New.
7751 (bucket_allocno_compare_func): Add a preference based on
7752 conflict_allocno_hard_prefs.
7753 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7754 (color_allocnos): Remove a dead code. Initiate
7755 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7757 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7760 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7761 constraint from =wa to wa. Avoid a subreg on the output operand,
7762 instead use a pseudo and subreg it in a move.
7763 (p9_xxbrd_<mode>): Changed to ...
7764 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7765 (p9_xxbrd_v2df): New expander.
7766 (p9_xxbrw_<mode>): Changed to ...
7767 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7768 (p9_xxbrw_v4sf): New expander.
7770 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7772 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7774 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7777 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7778 multiply in 32-bit mode.
7779 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7780 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7783 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7785 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7786 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7787 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7788 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7790 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7793 * dwarf2out.c (is_trivial_indirect_ref): New function.
7794 (dwarf2out_late_global_decl): Do not generate a location
7795 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7796 and that are not defined in the current unit.
7798 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7800 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7801 instead of a libcall for UNORDERED.
7803 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7806 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7807 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7809 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7812 * config/s390/s390.c (s390_set_current_function): Invoke
7813 s390_indirect_branch_settings also if fndecl didn't change.
7815 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7817 * config/rs6000/rs6000.md (blockage): Set length to zero.
7819 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7821 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7823 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7826 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7827 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7831 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7832 PARALLEL incoming that failed vt_get_decl_and_offset check.
7835 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7836 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7838 (get_variable_section): For decls in named .bss* sections pass true as
7839 second argument to bss_initializer_p.
7841 2018-02-09 Marek Polacek <polacek@redhat.com>
7842 Jakub Jelinek <jakub@redhat.com>
7845 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7846 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7847 Sync some changes from cxx_fold_indirect_ref.
7849 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7851 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7853 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7854 (BLOCK_INLINE_ENTRY_LABEL): New.
7855 (dwarf2out_var_location): Disregard inline entry markers.
7856 (inline_entry_data): New struct.
7857 (inline_entry_data_hasher): New hashtable type.
7858 (inline_entry_data_hasher::hash): New.
7859 (inline_entry_data_hasher::equal): New.
7860 (inline_entry_data_table): New variable.
7861 (add_high_low_attributes): Add DW_AT_entry_pc and
7862 DW_AT_GNU_entry_view attributes if a pending entry is found
7863 in inline_entry_data_table. Add old entry_pc attribute only
7864 if debug nonbinding markers are disabled.
7865 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7866 markers are enabled.
7867 (block_within_block_p, dwarf2out_inline_entry): New.
7868 (dwarf2out_finish): Check that no entries remained in
7869 inline_entry_data_table.
7870 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7871 (final_scan_insn, notice_source_line): Likewise.
7872 (rest_of_clean_state): Skip inline entry markers.
7873 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7875 * gimple.c (gimple_build_debug_inline_entry): New.
7876 * gimple.h (enum gimple_debug_subcode): Add
7877 GIMPLE_DEBUG_INLINE_ENTRY.
7878 (gimple_build_debug_inline_entry): Declare.
7879 (gimple_debug_inline_entry_p): New.
7880 (gimple_debug_nonbind_marker_p): Adjust.
7881 * insn-notes.def (INLINE_ENTRY): New.
7882 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7883 inline entry marker notes.
7884 (print_insn): Likewise.
7885 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7886 (INSN_DEBUG_MARKER_KIND): Likewise.
7887 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7888 * tree-inline.c (expand_call_inline): Build and insert
7889 debug_inline_entry stmt.
7890 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7891 inline entry blocks early, if nonbind markers are enabled.
7892 (dump_scope_block): Dump fragment info.
7893 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7894 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7895 (gimple_build_debug_inline_entry): New.
7896 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7897 Enable/disable inline entry points too.
7898 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7899 (DEBUG_INSN): Describe inline entry markers.
7901 * common.opt (gvariable-location-views): New.
7902 (gvariable-location-views=incompat5): New.
7903 * config.in: Rebuilt.
7904 * configure: Rebuilt.
7905 * configure.ac: Test assembler for view support.
7906 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7907 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7908 * dwarf2out.c (var_loc_view): New typedef.
7909 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7910 (dwarf2out_locviews_in_attribute): New.
7911 (dwarf2out_locviews_in_loclist): New.
7912 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7913 (enum dw_line_info_opcode): Add LI_adv_address.
7914 (struct dw_line_info_table): Add view.
7915 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7916 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7917 (zero_view_p): New variable.
7918 (ZERO_VIEW_P): New macro.
7919 (output_asm_line_debug_info): New.
7920 (struct var_loc_node): Add view.
7921 (add_AT_view_list, AT_loc_list): New.
7922 (add_var_loc_to_decl): Add view param. Test it against last.
7923 (new_loc_list): Add view params. Record them.
7924 (AT_loc_list_ptr): Handle loc and view lists.
7925 (view_list_to_loc_list_val_node): New.
7926 (print_dw_val): Handle dw_val_class_view_list.
7927 (size_of_die): Likewise.
7928 (value_format): Likewise.
7929 (loc_list_has_views): New.
7930 (gen_llsym): Set vl_symbol too.
7931 (maybe_gen_llsym, skip_loc_list_entry): New.
7932 (dwarf2out_maybe_output_loclist_view_pair): New.
7933 (output_loc_list): Output view list or entries too.
7934 (output_view_list_offset): New.
7935 (output_die): Handle dw_val_class_view_list.
7936 (output_dwarf_version): New.
7937 (output_compilation_unit_header): Use it.
7938 (output_skeleton_debug_sections): Likewise.
7939 (output_rnglists, output_line_info): Likewise.
7940 (output_pubnames, output_aranges): Update version comments.
7941 (output_one_line_info_table): Output view numbers in asm comments.
7942 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7943 Call maybe_gen_llsym.
7944 (loc_list_from_tree_1): Adjust.
7945 (add_AT_location_description): Create view list attribute if
7946 needed, check it's absent otherwise.
7947 (convert_cfa_to_fb_loc_list): Adjust.
7948 (maybe_emit_file): Call output_asm_line_debug_info for test.
7949 (dwarf2out_var_location): Reset views as needed. Precompute
7950 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7951 attribute. Set view.
7952 (new_line_info_table): Reset next view.
7953 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7954 (dwarf2out_source_line): Likewise. Output view resets and labels to
7955 the assembler, or select appropriate line info opcodes.
7956 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7957 (optimize_string_length): Catch it. Adjust.
7958 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7959 dw_val_class_view_list, and remove it if no longer needed.
7960 (hash_loc_list): Hash view numbers.
7961 (loc_list_hasher::equal): Compare them.
7962 (optimize_location_lists): Check whether a view list symbol is
7963 needed, and whether the locview attribute is present, and
7964 whether they match. Remove the locview attribute if no longer
7966 (index_location_lists): Call skip_loc_list_entry for test.
7967 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7968 Use output_dwarf_version.
7969 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7970 (struct dw_val_node): Add val_view_list.
7971 * final.c (SEEN_NEXT_VIEW): New.
7972 (set_next_view_needed): New.
7973 (clear_next_view_needed): New.
7974 (maybe_output_next_view): New.
7975 (final_start_function): Rename to...
7976 (final_start_function_1): ... this. Take pointer to FIRST,
7977 add SEEN parameter. Emit param bindings in the initial view.
7978 (final_start_function): Reintroduce SEEN-less interface.
7979 (final): Rename to...
7980 (final_1): ... this. Take SEEN parameter. Output final pending
7981 next view at the end.
7982 (final): Reintroduce seen-less interface.
7983 (final_scan_insn): Output pending next view before switching
7984 sections or ending a block. Mark the next view as needed when
7985 outputting variable locations. Notify debug backend of section
7986 changes, and of location view changes.
7987 (rest_of_handle_final): Adjust.
7988 * toplev.c (process_options): Autodetect value for debug variable
7989 location views option. Warn on incompat5 without -gdwarf-5.
7990 * doc/invoke.texi (gvariable-location-views): New.
7991 (gvariable-location-views=incompat5): New.
7992 (gno-variable-location-views): New.
7994 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7996 PR tree-optimization/84136
7997 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7998 that the result of find_edge is non-NULL.
8000 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8003 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8004 storing integer register in SImode. Fix cost of 256 and 512
8005 byte aligned SSE register store.
8007 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8009 * config/i386/i386.c (ix86_multiplication_cost): Fix
8010 multiplication cost for TARGET_AVX512DQ.
8012 2018-02-08 Marek Polacek <polacek@redhat.com>
8014 PR tree-optimization/84238
8015 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8018 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8020 PR tree-optimization/84265
8021 * tree-vect-stmts.c (vectorizable_store): Don't treat
8022 VMAT_CONTIGUOUS accesses as grouped.
8023 (vectorizable_load): Likewise.
8025 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8027 PR tree-optimization/81635
8028 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8029 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8030 (test_round_for_mask): New functions.
8031 (wide_int_cc_tests): Call test_round_for_mask.
8032 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8033 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8034 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8035 range returned by get_range_info.
8037 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8040 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8041 * symtab.c: Include builtins.h
8042 (symtab_node::output_to_lto_symbol_table_p): Move here
8043 from lto-streamer-out.c:output_symbol_p.
8044 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8045 (output_symbol_p): Move all logic to symtab.c
8046 (produce_symtab): Update.
8048 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8050 * config/s390/s390-opts.h (enum indirect_branch): Define.
8051 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8052 (s390_indirect_branch_via_thunk)
8053 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8054 (enum s390_indirect_branch_type): Define.
8055 * config/s390/s390.c (struct s390_frame_layout, struct
8056 machine_function): Remove.
8057 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8058 (indirect_branch_table_label_no, indirect_branch_table_name):
8060 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8061 (enum s390_indirect_branch_option): Define.
8062 (s390_return_addr_from_memory): New function.
8063 (s390_handle_string_attribute): New function.
8064 (s390_attribute_table): Add new attribute handler.
8065 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8066 (s390_indirect_branch_via_thunk): New function.
8067 (s390_indirect_branch_via_inline_thunk): New function.
8068 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8069 sibling call optimization for non z10 compiles.
8070 (s390_emit_call): Force indirect branch target to be a single
8071 register. Add r1 clobber for non-z10 compiles.
8072 (s390_emit_epilogue): Emit return jump via return_use expander.
8073 (s390_reorg): Handle JUMP_INSNs as execute targets.
8074 (s390_option_override_internal): Perform validity checks for the
8075 new command line options.
8076 (s390_indirect_branch_attrvalue): New function.
8077 (s390_indirect_branch_settings): New function.
8078 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8079 (s390_output_indirect_thunk_function): New function.
8080 (s390_code_end): Implement target hook.
8081 (s390_case_values_threshold): Implement target hook.
8082 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8084 * config/s390/s390.h (struct s390_frame_layout)
8085 (struct machine_function): Move here from s390.c.
8086 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8087 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8088 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8089 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8090 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8091 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8092 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8093 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8094 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8095 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8096 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8097 (mnemonic attribute): Add values which aren't recognized
8099 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8100 pattern for branch conversion. Fix mnemonic attribute.
8101 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8102 indirect branch via thunk if requested.
8103 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8104 ("*indirect_jump"): Disable for branch conversion using out of
8106 ("indirect_jump_via_thunk<mode>_z10")
8107 ("indirect_jump_via_thunk<mode>")
8108 ("indirect_jump_via_inlinethunk<mode>_z10")
8109 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8110 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8111 ("casesi_jump_via_inlinethunk<mode>_z10")
8112 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8113 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8114 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8115 ("*indirect2_jump"): Disable for branch conversion.
8116 ("casesi_jump"): Turn into expander and expand patterns for branch
8118 ("return_use"): New expander.
8119 ("*return"): Emit return via thunk and rename it to ...
8120 ("*return<mode>"): ... this one.
8121 * config/s390/s390.opt: Add new options and and enum for the
8124 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8126 * lra-constraints.c (match_reload): Unconditionally use
8127 gen_lowpart_SUBREG, rather than selecting between that
8128 and equivalent gen_rtx_SUBREG code.
8130 2018-02-08 Richard Biener <rguenther@suse.de>
8132 PR tree-optimization/84233
8133 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8134 changed flag instead of boguously re-using phi_inserted.
8136 2018-02-08 Martin Jambor <mjambor@suse.cz>
8138 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8139 static local variables.
8141 2018-02-08 Richard Biener <rguenther@suse.de>
8143 PR tree-optimization/84278
8144 * tree-vect-stmts.c (vectorizable_store): When looking for
8145 smaller vector types to perform grouped strided loads/stores
8146 make sure the mode is supported by the target.
8147 (vectorizable_load): Likewise.
8149 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8151 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8152 Increase LDP/STP opportunities by adding adjacent callee-saves.
8154 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8156 PR rtl-optimization/84068
8157 PR rtl-optimization/83459
8158 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8160 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8162 PR tree-optimization/84224
8163 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8164 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8167 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8170 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8171 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8172 count, remove one USE.
8174 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8176 * doc/install.texi (Configuration): Document the
8177 --with-long-double-format={ibm,ieee} PowerPC configuration
8181 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8182 Convert from define_expand to be define_insn_and_split. Rework
8183 float/double/_Float128 conversions to QI/HI/SImode to work with
8184 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8185 conversions to QI/HImode types did a store and then a load to
8186 truncate the value. For conversions to VSX registers, don't split
8187 the insn, instead emit the code directly. Use the code iterator
8188 any_fix to combine signed and unsigned conversions.
8189 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8190 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8191 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8192 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8193 (fix_<mode>di2_hw): Likewise.
8194 (fixuns_<mode>di2_hw): Likewise.
8195 (fix_<mode>si2_hw): Likewise.
8196 (fixuns_<mode>si2_hw): Likewise.
8197 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8198 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8199 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8200 fix<uns>_trunc<SFDF:mode>si2_p8.
8201 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8203 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8204 (fix<uns>_<mode>_mem): Likewise.
8205 (fctiw<u>z_<mode>_mem): Likewise.
8206 (fix<uns>_<mode>_mem): Likewise.
8207 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8208 the register allocator from doing a direct move to the GPRs to do
8209 a store, and instead use the ISA 3.0 store byte/half-word from
8210 vector register instruction. For IEEE 128-bit floating point,
8211 also optimize stores of 32-bit ints.
8212 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8214 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8216 * genextract.c (push_pathstr_operand): New function to support
8218 (walk_rtx): Call push_pathstr_operand.
8219 (print_path): Support [a-zA-Z].
8221 2018-02-07 Richard Biener <rguenther@suse.de>
8223 PR tree-optimization/84037
8224 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8225 (cse_and_gimplify_to_preheader): Declare.
8226 (vect_get_place_in_interleaving_chain): Likewise.
8227 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8229 (_loop_vec_info::~_loop_vec_info): Delete it.
8230 (cse_and_gimplify_to_preheader): New function.
8231 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8232 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8233 (vectorizable_load): Likewise. For grouped stores always base
8234 the IV on the first element.
8235 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8236 condition before gimplifying.
8238 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8240 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8241 *DIV_EXPR and *MOD_EXPR.
8243 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8246 * config/i386/i386.c (ix86_option_override_internal): Mask out
8247 the CF_SET bit when checking -fcf-protection.
8249 2018-02-07 Tom de Vries <tom@codesourcery.com>
8252 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8255 2018-02-07 Richard Biener <rguenther@suse.de>
8257 PR tree-optimization/84204
8258 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8261 PR tree-optimization/84205
8262 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8263 special-case isl_ast_op_zdiv_r.
8265 PR tree-optimization/84223
8266 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8267 Only add conditions from within the region.
8268 (gather_bbs::after_dom_children): Adjust.
8270 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8273 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8274 * config/avr/avr.md: Only post-reload split REG-REG moves if
8275 either register is GENERAL_REG_P.
8277 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8279 PR tree-optimization/84235
8280 * tree-ssa-scopedtables.c
8281 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8282 if the subtraction is performed in floating point type where NaNs are
8283 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8284 build 1. Formatting fix.
8286 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8289 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8290 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8291 and skip it regardless of bb boundaries. Use CALL_P macro,
8292 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8293 unnecessarily, formatting fix.
8295 2018-02-06 Michael Collison <michael.collison@arm.com>
8297 * config/arm/thumb2.md:
8298 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8299 (*thumb_mov_notscc): Ditto.
8301 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8304 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8305 unsigned_fix, not "s".
8307 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8309 * configure.ac (gcc_fn_eh_frame_ro): New function.
8310 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8311 correct .eh_frame permissions.
8312 * configure: Regenerate.
8314 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8316 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8319 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8321 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8322 Display warning message for -mno-speculate-indirect-jumps.
8324 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8326 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8328 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8330 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8332 PR tree-optimization/84225
8333 * tree-eh.c (find_trapping_overflow): Only call
8334 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8336 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8339 * config/i386/i386.c: Reimplement the check of possible options
8340 -mibt/-mshstk conbination. Change error messages.
8341 * doc/invoke.texi: Fix a typo: remove extra '='.
8343 2018-02-06 Marek Polacek <polacek@redhat.com>
8345 PR tree-optimization/84228
8346 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8348 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8351 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8352 emitted arch directives.
8353 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8354 __ARM_FEATURE_COPROC before changing architectures.
8356 2018-02-06 Richard Biener <rguenther@suse.de>
8358 * config/i386/i386.c (print_reg): Fix typo.
8359 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8361 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8363 * configure: Regenerate.
8365 2018-02-05 Martin Sebor <msebor@redhat.com>
8367 PR tree-optimization/83369
8368 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8371 2018-02-05 Martin Liska <mliska@suse.cz>
8373 * doc/invoke.texi: Cherry-pick upstream r323995.
8375 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8377 * ira.c (ira_init_register_move_cost): Adjust comment.
8379 2018-02-05 Martin Liska <mliska@suse.cz>
8381 PR gcov-profile/84137
8382 * doc/gcov.texi: Fix typo in documentation.
8384 2018-02-05 Martin Liska <mliska@suse.cz>
8386 PR gcov-profile/83879
8387 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8388 using dlopen functionality.
8390 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8392 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8393 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8394 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8395 _mm_maskz_range_round_ss): New intrinsics.
8396 (__builtin_ia32_rangesd128_round)
8397 (__builtin_ia32_rangess128_round): Remove.
8398 (__builtin_ia32_rangesd128_mask_round,
8399 __builtin_ia32_rangess128_mask_round): New builtins.
8400 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8401 __builtin_ia32_rangess128_round): Remove.
8402 (__builtin_ia32_rangesd128_mask_round,
8403 __builtin_ia32_rangess128_mask_round): New builtins.
8404 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8405 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8406 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8407 "<round_saeonly_constraint>")): Changed to ...
8408 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8409 "<round_saeonly_scalar_constraint>")): ... this.
8410 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8411 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8412 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8413 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8414 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8416 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8418 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8420 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8421 Remove all values except native, 8540 and 8548.
8423 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8425 * config/i386/i386.c (ix86_output_function_return): Pass
8426 INVALID_REGNUM, instead of -1, as invalid register number to
8427 indirect_thunk_name and output_indirect_thunk.
8429 2018-02-02 Julia Koval <julia.koval@intel.com>
8431 * config.gcc: Add -march=icelake.
8432 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8433 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8434 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8435 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8436 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8437 (processor_target_table): Add icelake.
8438 (ix86_option_override_internal): Handle new PTAs.
8439 (get_builtin_code_for_version): Handle icelake.
8440 (M_INTEL_COREI7_ICELAKE): New.
8441 (fold_builtin_cpu): Handle icelake.
8442 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8443 * doc/invoke.texi: Add -march=icelake.
8445 2018-02-02 Julia Koval <julia.koval@intel.com>
8447 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8448 to wide_int_bitmask.
8449 * wide-int-bitmask.h: New.
8451 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8454 * config/i386/i386.md: Replace Pmode with word_mode in
8455 builtin_setjmp_setup and builtin_longjmp to support x32.
8457 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8461 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8463 (rs6000_supported_cpu_names): New static variable.
8464 (linux_cpu_translation_table): Likewise.
8465 (elf_platform) <cpu>: Define new static variable and use it.
8466 Translate kernel AT_PLATFORM name to canonical name if needed.
8467 Error if platform name is unknown.
8469 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8472 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8474 2018-02-01 Jeff Law <law@redhat.com>
8477 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8478 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8479 the scratch if RELEASE_VIA_POP is false.
8480 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8481 If we have to save a temporary register, decrement SIZE appropriately.
8482 Pass new arguments to release_scratch_register_on_entry.
8483 (ix86_adjust_stack_and_probe): Likewise.
8484 (ix86_emit_probe_stack_range): Pass new arguments to
8485 release_scratch_register_on_entry.
8487 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8489 PR rtl-optimization/84157
8490 * combine.c (change_zero_ext): Use REG_P predicate in
8491 front of HARD_REGISTER_P predicate.
8493 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8495 * config/avr/avr.c (avr_option_override): Move disabling of
8496 -fdelete-null-pointer-checks to...
8497 * common/config/avr/avr-common.c (avr_option_optimization_table):
8500 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8502 PR tree-optimization/81635
8503 * tree-data-ref.c (split_constant_offset_1): For types that
8504 wrap on overflow, try to use range info to prove that wrapping
8507 2018-02-01 Renlin Li <renlin.li@arm.com>
8510 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8512 (aarch64_register_move_cost): Likewise.
8513 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8515 (REG_CLASS_NAMES): Likewise.
8516 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8517 TAILCALL_ADDR_REGS. Remove IP registers.
8518 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8520 2018-02-01 Richard Biener <rguenther@suse.de>
8522 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8523 for specifying RPO order and allow NULL for that.
8524 * domwalk.c (dom_walker::dom_walker): Likewise.
8525 (dom_walker::walk): Handle NULL RPO order.
8526 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8528 (rewrite_update_dom_walker): Likewise.
8529 (mark_def_dom_walker): Likewise.
8531 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8533 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8534 (aarch64_maybe_expand_sve_subreg_move): Declare.
8535 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8536 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8538 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8539 that are semantically a reverse operation.
8540 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8541 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8542 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8544 (aarch64_can_change_mode_class): For big-endian, forbid changes
8545 between two SVE modes if they have different element sizes.
8547 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8549 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8550 the TImode handling for big-endian targets.
8552 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8554 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8555 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8557 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8558 Remove BSWAP handing for big-endian targets and use the form of
8559 LD1RQ appropariate for the mode.
8561 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8563 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8564 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8567 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8570 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8571 check for operands that need to go through aarch64_sve_reload_be.
8573 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8575 PR tree-optimization/81661
8576 PR tree-optimization/84117
8577 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8578 * tree-eh.c: Include gimplify.h.
8579 (find_trapping_overflow, replace_trapping_overflow,
8580 rewrite_to_non_trapping_overflow): New functions.
8581 * tree-vect-loop.c: Include tree-eh.h.
8582 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8583 * tree-data-ref.c: Include tree-eh.h.
8584 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8586 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8588 PR rtl-optimization/84123
8589 * combine.c (change_zero_ext): Check if hard register satisfies
8590 can_change_dest_mode before calling gen_lowpart_SUBREG.
8592 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8595 * ira.c (ira_init_register_move_cost): Remove assert.
8597 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8599 PR rtl-optimization/84071
8600 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8601 * doc/tm.texi: Regenerate.
8603 2018-01-31 Richard Biener <rguenther@suse.de>
8605 PR tree-optimization/84132
8606 * tree-data-ref.c (analyze_miv_subscript): Properly
8607 check whether evolution_function_is_affine_multivariate_p
8608 before calling gcd_of_steps_may_divide_p.
8610 2018-01-31 Julia Koval <julia.koval@intel.com>
8613 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8614 * config/i386/i386.md (rdpid_rex64) New.
8615 (rdpid): Make 32bit only.
8617 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8620 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8621 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8623 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8626 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8628 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8630 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8632 PR rtl-optimization/84071
8633 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8634 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8636 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8638 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8639 (arc_attribute_table): Add 'aux' attribute.
8640 (arc_in_small_data_p): Consider aux like variables.
8641 (arc_is_aux_reg_p): New function.
8642 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8643 (arc_get_aux_arg): New function.
8644 (prepare_move_operands): Handle aux-register access.
8645 (arc_handle_aux_attribute): New function.
8646 * doc/extend.texi (ARC Variable attributes): Add subsection.
8648 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8650 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8651 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8652 (arc_attribute_table): Add 'uncached' attribute.
8653 (arc_print_operand): Print '.di' flag for uncached memory
8655 (arc_in_small_data_p): Do not consider for small data the uncached
8657 (arc_is_uncached_mem_p): New function.
8658 * config/arc/predicates.md (compact_store_memory_operand): Check
8659 for uncached memory accesses.
8660 (nonvol_nonimm_operand): Likewise.
8661 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8663 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8666 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8667 falign-loops=): Add Optimization flag.
8669 2018-01-30 Jeff Law <law@redhat.com>
8672 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8673 INT_REGISTERS_SAVED. Check it prior to calling
8674 get_scratch_register_on_entry.
8675 (ix86_adjust_stack_and_probe): Similarly.
8676 (ix86_emit_probe_stack_range): Similarly.
8677 (ix86_expand_prologue): Corresponding changes.
8679 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8682 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8683 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8685 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8688 * lra-constraints.c (curr_insn_transform): Process AND in the
8691 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8693 PR rtl-optimization/83986
8694 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8695 dependence against last_pending_memory_flush in addition to
8698 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8700 PR tree-optimization/81611
8701 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8704 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8707 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8710 2018-01-30 Richard Biener <rguenther@suse.de>
8711 Jakub Jelinek <jakub@redhat.com>
8713 PR tree-optimization/84111
8714 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8715 inner loops added during recursion, as they don't have up-to-date
8718 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8721 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8722 (can_inline_edge_by_limits_p): ... here.
8723 (can_early_inline_edge_p, check_callers,
8724 update_caller_keys, update_callee_keys, recursive_inlining,
8725 add_new_edges_to_heap, speculation_useful_p,
8726 inline_small_functions,
8727 inline_small_functions, flatten_function,
8728 inline_to_all_callers_1): Update.
8730 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8732 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8735 2018-01-30 Richard Biener <rguenther@suse.de>
8737 PR tree-optimization/83008
8738 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8739 invariant and constant vector uses in stmts when they need
8742 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8745 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8746 * configure: Regenerate.
8748 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8750 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8752 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8753 Use gen_rtx_REG rather than gen_lowpart.
8755 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8757 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8758 rather than 0 when creating partial subregs.
8760 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8762 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8765 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8768 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8769 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8770 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8771 flags. This restores the settings used before the 2017-07-24.
8772 Turning off pre increment/decrement/modify allows IVOPTS to
8773 optimize DF/SF loops where the index is an int.
8775 2018-01-29 Richard Biener <rguenther@suse.de>
8776 Kelvin Nilsen <kelvin@gcc.gnu.org>
8779 * tree-vect-stmts.c (vectorizable_call): Don't call
8780 targetm.vectorize_builtin_md_vectorized_function if callee is
8783 2018-01-22 Carl Love <cel@us.ibm.com>
8785 * doc/extend.tex: Fix typo in second arg in
8786 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8788 2018-01-29 Richard Biener <rguenther@suse.de>
8790 PR tree-optimization/84086
8791 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8792 (flush_ssaname_freelist): When SSA names were released reset
8793 the SCEV hash table.
8795 2018-01-29 Richard Biener <rguenther@suse.de>
8797 PR tree-optimization/84057
8798 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8799 removed paths when removing edges.
8801 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8803 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8804 -mfunction-return=@var{choice}.
8806 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8809 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8811 (layout::print_source_line): Likewise.
8812 (test_get_line_width_without_trailing_whitespace): Add test cases.
8814 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8817 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8820 2018-01-26 Jim Wilson <jimw@sifive.com>
8822 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8824 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8827 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8829 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8830 and CMP + SUB-immediate -> SUBS.
8832 2018-01-26 Martin Sebor <msebor@redhat.com>
8834 PR tree-optimization/83896
8835 * tree-ssa-strlen.c (get_string_len): Rename...
8836 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8837 Avoid assuming length is constant.
8838 (handle_char_store): Use HOST_WIDE_INT for string length.
8840 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8843 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8844 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8846 2018-01-26 Richard Biener <rguenther@suse.de>
8848 PR rtl-optimization/84003
8849 * dse.c (record_store): Only record redundant stores when
8850 the earlier store aliases at least all accesses the later one does.
8852 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8854 PR rtl-optimization/83985
8855 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8856 REG_CFA_RESTORE insns.
8857 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8858 REG_CFA_RESTORE insns here.
8861 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8862 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8864 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8866 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8867 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8868 (arc_init): Likewise.
8869 (arc_override_options): Likewise.
8870 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8872 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8874 * config/arc/arc.h (TARGET_DBNZ): Define.
8875 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8876 properly set the tune attribute.
8877 (dbnz): Use TARGET_DBNZ guard.
8878 * config/arc/arc.opt (mtune): Add core3 option.
8880 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8882 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8883 recognize new pic like addresses.
8884 (arc_delegitimize_address): Clean up.
8886 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8888 * config/arc/arc-arches.def: Option mrf16 valid for all
8890 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8891 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8892 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8893 * config/arc/arc-tables.opt: Regenerate.
8894 * config/arc/arc.c (arc_conditional_register_usage): Handle
8895 reduced register file case.
8896 (arc_file_start): Set must have build attributes.
8897 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8899 * config/arc/arc.opt (mrf16): Add new option.
8900 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8901 * config/arc/genmultilib.awk: Handle new mrf16 option.
8902 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8903 * config/arc/t-multilib: Regenerate.
8904 * doc/invoke.texi (ARC Options): Document mrf16 option.
8906 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8908 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8909 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8910 (arc_attribute_table): Add 'secure_call' attribute.
8911 (arc_print_operand): Print secure call operand.
8912 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8914 (arc_is_secure_call_p): New function. * config/arc/arc.md
8915 (call_i): Add support for sjli instruction.
8916 (call_value_i): Likewise.
8917 * config/arc/constraints.md (Csc): New constraint.
8919 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8920 John Eric Martin <John.Martin@emmicro-us.com>
8922 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8923 * config/arc/arc.c (_arc_jli_section): New struct.
8924 (arc_jli_section): New type.
8925 (rc_jli_sections): New static variable.
8926 (arc_handle_jli_attribute): New function.
8927 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8928 (arc_file_end): New function.
8929 (TARGET_ASM_FILE_END): Define.
8930 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8931 (arc_add_jli_section): New function.
8932 (jli_call_scan): Likewise.
8933 (arc_reorg): Call jli_call_scan.
8934 (arc_output_addsi): Remove 'S' from printing asm operand.
8935 (arc_is_jli_call_p): New function.
8936 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8938 (movhi_insn): Likewise.
8939 (movsi_insn): Likewise.
8940 (movsi_set_cc_insn): Likewise.
8941 (loadqi_update): Likewise.
8942 (load_zeroextendqisi_update): Likewise.
8943 (load_signextendqisi_update): Likewise.
8944 (loadhi_update): Likewise.
8945 (load_zeroextendhisi_update): Likewise.
8946 (load_signextendhisi_update): Likewise.
8947 (loadsi_update): Likewise.
8948 (loadsf_update): Likewise.
8949 (movsicc_insn): Likewise.
8950 (bset_insn): Likewise.
8951 (bxor_insn): Likewise.
8952 (bclr_insn): Likewise.
8953 (bmsk_insn): Likewise.
8954 (bicsi3_insn): Likewise.
8955 (cmpsi_cc_c_insn): Likewise.
8956 (movsi_ne): Likewise.
8957 (movsi_cond_exec): Likewise.
8958 (clrsbsi2): Likewise.
8969 (call_i): Remove 'S' asm letter, add jli instruction.
8970 (call_value_i): Likewise.
8971 * config/arc/arc.op (mjli-always): New option.
8972 * config/arc/constraints.md (Cji): New constraint.
8973 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8975 (subsf3_fpx): Likewise.
8976 (mulsf3_fpx): Likewise.
8977 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8979 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8981 * doc/invoke.texi (ARC): Document mjli-always option.
8983 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8985 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8986 avoid addition with 0 and use incw and decw where possible.
8988 2018-01-26 Richard Biener <rguenther@suse.de>
8990 PR tree-optimization/81082
8991 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8992 association if it requires casting to unsigned.
8993 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8994 from fold_plusminus_mult_expr to catch important cases late when
8995 range info is available.
8997 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8999 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9000 * configure.ac (hidden_linkonce): New test.
9001 * configure: Regenerate.
9002 * config.in: Regenerate.
9004 2018-01-26 Julia Koval <julia.koval@intel.com>
9006 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9007 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9008 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9009 _mm_mask_bitshuffle_epi64_mask): Fix type.
9010 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9011 USI_FTYPE_V4DI_V4DI_USI): Remove.
9012 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9013 __builtin_ia32_vpshufbitqmb256_mask,
9014 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9015 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9016 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9018 2018-01-26 Alan Modra <amodra@gmail.com>
9021 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9022 UNSPEC_VBPERMQ. Sort other unspecs.
9024 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9026 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9028 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9031 * predict.c (drop_profile): Do not push/pop cfun; update also
9033 (handle_missing_profiles): Fix logic looking for zero profiles.
9035 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9038 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9039 on functions with #pragma omp declare simd or functions with simd
9041 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9042 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9043 Remove trailing \n from warning_at calls.
9045 2018-01-25 Tom de Vries <tom@codesourcery.com>
9048 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9049 for neutered workers.
9051 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9054 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9055 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9057 2018-01-24 Jeff Law <law@redhat.com>
9060 * i386.c (get_probe_interval): Move to earlier point.
9061 (ix86_compute_frame_layout): If -fstack-clash-protection and
9062 the frame is larger than the probe interval, then use pushes
9063 to save registers rather than reg->mem moves.
9064 (ix86_expand_prologue): Remove conditional for int_registers_saved
9067 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9070 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9071 min/max for never referenced object.
9073 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9076 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9078 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9079 attributes from DECL_ATTRIBUTES (decl) without affecting
9080 DECL_ATTRIBUTES (current_function_decl).
9081 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9082 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9084 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9086 PR tree-optimization/83979
9087 * fold-const.c (fold_comparison): Use constant_boolean_node
9088 instead of boolean_{true,false}_node.
9090 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9092 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9095 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9097 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9098 Simplify the clause that sets the length attribute.
9099 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9100 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9101 clause that sets the length attribute.
9102 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9104 2018-01-24 Tom de Vries <tom@codesourcery.com>
9107 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9108 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9109 Add strict parameter.
9110 (prevent_branch_around_nothing): Insert dummy insn between branch to
9111 label and label with no ptx insn inbetween.
9112 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9114 2018-01-24 Tom de Vries <tom@codesourcery.com>
9117 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9118 for neutered threads in warp.
9119 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9121 2018-01-24 Richard Biener <rguenther@suse.de>
9123 PR tree-optimization/83176
9124 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9127 2018-01-24 Richard Biener <rguenther@suse.de>
9129 PR tree-optimization/82819
9130 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9131 code generating pluses that are no-ops in the target precision.
9133 2018-01-24 Richard Biener <rguenther@suse.de>
9136 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9138 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9140 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9141 to merge probabilities.
9142 * predict.c (probably_never_executed): Also mark as cold functions
9143 with global 0 profile and guessed local profile.
9144 * profile-count.c (profile_probability::combine_with_count): New
9146 * profile-count.h (profile_probability::operator*,
9147 profile_probability::operator*=, profile_probability::operator/,
9148 profile_probability::operator/=): Reduce precision to adjusted
9149 and set value to guessed on contradictory divisions.
9150 (profile_probability::combine_with_freq): Remove.
9151 (profile_probability::combine_wiht_count): Declare.
9152 (profile_count::force_nonzero):: Set to adjusted.
9153 (profile_count::probability_in):: Set quality to adjusted.
9154 * tree-ssa-tail-merge.c (replace_block_by): Use
9157 2018-01-23 Andrew Waterman <andrew@sifive.com>
9158 Jim Wilson <jimw@sifive.com>
9160 * config/riscv/riscv.c (riscv_stack_boundary): New.
9161 (riscv_option_override): Set riscv_stack_boundary. Handle
9162 riscv_preferred_stack_boundary_arg.
9163 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9164 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9165 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9166 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9167 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9168 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9170 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9173 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9174 of struct ix86_frame.
9175 (ix86_expand_epilogue): Likewise. Add a local variable for
9176 the reg_save_offset field in struct ix86_frame.
9178 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9180 PR tree-optimization/82604
9181 * tree-loop-distribution.c (enum partition_kind): New enum item
9182 PKIND_PARTIAL_MEMSET.
9183 (partition_builtin_p): Support above new enum item.
9184 (generate_code_for_partition): Ditto.
9185 (compute_access_range): Differentiate cases that equality can be
9186 proven at all loops, the innermost loops or no loops.
9187 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9188 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9189 (finalize_partitions, distribute_loop): Don't fuse partition of
9190 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9191 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9194 2018-01-23 Martin Liska <mliska@suse.cz>
9196 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9197 order to ignore the predictor.
9198 (PRED_POLYMORPHIC_CALL): Likewise.
9199 (PRED_RECURSIVE_CALL): Likewise.
9201 2018-01-23 Martin Liska <mliska@suse.cz>
9203 * tree-profile.c (tree_profiling): Print function header to
9204 aware reader which function we are working on.
9205 * value-prof.c (gimple_find_values_to_profile): Do not print
9206 not interesting value histograms.
9208 2018-01-23 Martin Liska <mliska@suse.cz>
9210 * profile-count.h (enum profile_quality): Add
9211 profile_uninitialized as the first value. Do not number values
9212 as they are zero based.
9213 (profile_count::verify): Update sanity check.
9214 (profile_probability::verify): Likewise.
9216 2018-01-23 Nathan Sidwell <nathan@acm.org>
9218 * doc/invoke.texi (ffor-scope): Deprecate.
9220 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9222 PR tree-optimization/83510
9223 * domwalk.c (set_all_edges_as_executable): New function.
9224 (dom_walker::dom_walker): Convert bool param
9225 "skip_unreachable_blocks" to enum reachability. Move setup of
9226 edge flags to set_all_edges_as_executable and only do it when
9227 reachability is REACHABLE_BLOCKS.
9228 * domwalk.h (enum dom_walker::reachability): New enum.
9229 (dom_walker::dom_walker): Convert bool param
9230 "skip_unreachable_blocks" to enum reachability.
9231 (set_all_edges_as_executable): New decl.
9232 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9233 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9235 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9236 but converting true to REACHABLE_BLOCKS.
9237 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9239 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9240 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9241 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9243 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9244 if check_all_array_refs will be called.
9246 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9248 * tree.c (selftest::test_location_wrappers): Add more test
9251 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9253 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9254 (selftest::test_bit_in_range): Likewise.
9256 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9259 * doc/sourcebuild.texi (vect_float): Say that the selector
9260 only describes the situation when -funsafe-math-optimizations is on.
9261 (vect_float_strict): Document.
9263 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9265 PR tree-optimization/83965
9266 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9267 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9268 instead of checking only for a reduction.
9269 (vect_recog_widen_sum_pattern): Likewise.
9271 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9273 * predict.c (probably_never_executed): Only use precise profile info.
9274 (compute_function_frequency): Skip after inlining hack since we now
9275 have quality checking.
9277 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9279 * profile-count.h (profile_probability::very_unlikely,
9280 profile_probability::unlikely, profile_probability::even): Set
9281 precision to guessed.
9283 2018-01-23 Richard Biener <rguenther@suse.de>
9285 PR tree-optimization/83963
9286 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9287 Properly terminate dominator walk when crossing the exit edge not
9288 when visiting its source block.
9290 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9293 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9294 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9296 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9298 PR tree-optimization/83957
9299 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9300 semicolon after for body surrounded by braces.
9302 PR tree-optimization/83081
9303 * profile-count.h (profile_probability::split): New method.
9304 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9305 Use profile_probability::split.
9306 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9307 when splitting a single conditional jump into 2.
9309 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9311 PR tree-optimization/69452
9312 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9315 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9317 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9318 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9319 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9321 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9323 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9324 * config/rl78/rl78.md (movdi): New define_expand.
9325 * config/rl78/rl78.c (rl78_split_movdi): New function.
9327 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9330 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9332 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9333 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9334 128-bit to produce an UNSPEC move to get the double word with the
9335 signbit and then a shift directly to do signbit.
9336 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9337 implementation with a new version that just does either a direct
9338 move or a regular move. Move memory interface to separate insns.
9339 Move insns so they are next to the expander.
9340 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9341 with signbit move. Split big and little endian case.
9342 (signbit<mode>2_dm_mem_le): Likewise.
9343 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9344 (signbit<mode>2_dm2): Likewise.
9346 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9348 * config/rl78/rl78.md (anddi3): New define_expand.
9350 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9352 * config/rl78/rl78.md (umindi3): New define_expand.
9354 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9356 * config/rl78/rl78.md (smindi3): New define_expand.
9358 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9360 * config/rl78/rl78.md (smaxdi3): New define_expand.
9362 2018-01-22 Carl Love <cel@us.ibm.com>
9364 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9365 LVX_V1TI): Add macro expansion.
9366 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9367 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9368 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9369 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9370 Change check to determine if the instruction is a byte reversing
9371 entry. Fix typo in comment.
9372 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9373 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9374 Add def_builtin calls for new builtins.
9375 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9376 Add define_insn expansion.
9378 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9380 * config/rl78/rl78.md (umaxdi3): New define_expand.
9382 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9384 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9385 for non-QImode registers.
9387 2018-01-22 Richard Biener <rguenther@suse.de>
9389 PR tree-optimization/83963
9390 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9391 including the loop exit block.
9392 (scop_detection::merge_sese): Likewise.
9393 (scop_detection::add_scop): Do it here instead.
9395 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9397 * doc/sourcebuild.texi (arm_softfloat): Document.
9399 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9402 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9403 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9404 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9406 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9407 David Edelsohn <dje.gcc@gmail.com>
9410 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9411 Change "crset eq" to "crset 2".
9412 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9413 (*call_indirect_aix<mode>_nospec): Likewise.
9414 (*call_value_indirect_aix<mode>_nospec): Likewise.
9415 (*call_indirect_elfv2<mode>_nospec): Likewise.
9416 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9417 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9418 change assembly output from . to $.
9419 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9420 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9421 (*tablejump<mode>_internal1_nospec): Likewise.
9423 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9426 * config/sh/sh_optimize_sett_clrt.cc:
9427 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9429 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9431 PR tree-optimization/83940
9432 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9433 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9434 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9435 use it to pass back the definition type.
9436 (vect_check_store_rhs): Likewise rhs_dt_out.
9437 (vect_build_gather_load_calls): Add a mask_dt argument and use
9438 it instead of a call to vect_is_simple_use.
9439 (vectorizable_store): Update calls to vect_check_load_store_mask
9440 and vect_check_store_rhs. Use the dt returned by the latter instead
9441 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9442 instead of calls to vect_is_simple_use. Pass the scalar rather
9443 than the vector operand to vect_is_simple_use when handling
9444 second and subsequent copies of an rhs value.
9445 (vectorizable_load): Update calls to vect_check_load_store_mask
9446 and vect_build_gather_load_calls. Use the cached mask_dt and
9447 gs_info.offset_dt instead of calls to vect_is_simple_use.
9449 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9452 * tree-emutls.c: Include gimplify.h.
9453 (lower_emutls_2): New function.
9454 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9455 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9456 it before further processing.
9459 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9460 UINTVAL (trueop1) instead of INTVAL (op1).
9462 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9466 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9467 INCOMING_FRAME_SP_OFFSET if not defined.
9468 (scan_trace): Add ENTRY argument. If true and
9469 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9470 emit a note to adjust the CFA offset.
9471 (create_cfi_notes): Adjust scan_trace callers.
9472 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9473 INCOMING_FRAME_SP_OFFSET in the CIE.
9474 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9475 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9477 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9478 * doc/tm.texi: Regenerated.
9480 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9482 PR rtl-optimization/83147
9483 * lra-constraints.c (remove_inheritance_pseudos): Use
9484 lra_substitute_pseudo_within_insn.
9486 2018-01-19 Tom de Vries <tom@codesourcery.com>
9487 Cesar Philippidis <cesar@codesourcery.com>
9490 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9492 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9495 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9496 spaces for function labels.
9498 2018-01-19 Martin Liska <mliska@suse.cz>
9500 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9501 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9502 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9503 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9504 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9505 (PRED_CONST_RETURN): Change from 69 to 65.
9506 (PRED_NULL_RETURN): Change from 91 to 71.
9507 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9508 (PRED_LOOP_GUARD): Change from 66 to 73.
9510 2018-01-19 Martin Liska <mliska@suse.cz>
9512 * predict.c (predict_insn_def): Add new assert.
9513 (struct branch_predictor): Change type to signed integer.
9514 (test_prediction_value_range): Amend test to cover
9516 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9517 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9518 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9519 (PRED_LOOP_IV_COMPARE): Likewise.
9520 * predict.h (PROB_UNINITIALIZED): Define new constant.
9522 2018-01-19 Martin Liska <mliska@suse.cz>
9524 * predict.c (dump_prediction): Add new format for
9525 analyze_brprob.py script which is enabled with -details
9527 * profile-count.h (precise_p): New function.
9529 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9531 PR tree-optimization/83922
9532 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9533 there are no statements that need masking.
9534 (vect_active_double_reduction_p): New function.
9535 (vect_analyze_loop_operations): Use it when handling phis that
9536 are not in the loop header.
9538 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9540 PR tree-optimization/83914
9541 * tree-vect-loop.c (vectorizable_induction): Don't convert
9542 init_expr or apply the peeling adjustment for inductions
9543 that are nested within the vectorized loop.
9545 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9547 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9550 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9554 * function.h (gimplify_parameters): Add gimple_seq * argument.
9555 * function.c: Include gimple.h and options.h.
9556 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9557 for the added local temporaries if needed.
9558 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9559 if there are any parameter cleanups, wrap whole body into a
9560 try/finally with the cleanups.
9562 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9565 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9566 Use GET_MODE_CLASS for scalar floating point.
9568 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9572 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9573 Fix call of call_cgraph_insertion_hooks.
9575 2018-01-18 Martin Sebor <msebor@redhat.com>
9577 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9579 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9582 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9585 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9588 * common.opt: (-ffile-prefix-map): New option.
9589 * opts.c (common_handle_option): Defer it.
9590 * opts-global.c (handle_common_deferred_options): Handle it.
9591 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9592 * file-prefix-map.h: New file.
9593 (remap_debug_filename, add_debug_prefix_map): ...here.
9594 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9595 * final.c (debug_prefix_map, add_debug_prefix_map
9596 remap_debug_filename): Move to...
9597 * file-prefix-map.c: New file.
9598 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9599 generalize, get rid of alloca(), use strrchr() instead of strchr().
9600 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9601 Implement in terms of add_prefix_map().
9602 (remap_macro_filename, remap_debug_filename): Implement in term of
9604 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9605 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9606 * dbxout.c: Include file-prefix-map.h.
9607 * varasm.c: Likewise.
9608 * vmsdbgout.c: Likewise.
9609 * xcoffout.c: Likewise.
9610 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9611 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9612 * doc/invoke.texi (-ffile-prefix-map): Document.
9613 (-fdebug-prefix-map): Update description.
9615 2018-01-18 Martin Liska <mliska@suse.cz>
9617 * config/i386/i386.c (indirect_thunk_name): Document that also
9619 (output_indirect_thunk): Document why both instructions
9620 (pause and lfence) are generated.
9622 2018-01-18 Richard Biener <rguenther@suse.de>
9624 PR tree-optimization/83887
9625 * graphite-scop-detection.c
9626 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9627 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9628 (scop_detection::merge_sese): Re-implement with a flood-fill
9629 algorithm that properly finds a SESE region if it exists.
9631 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9634 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9635 pointer_diff optimizations use view_convert instead of convert.
9637 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9639 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9640 Generate different code for -mno-speculate-indirect-jumps.
9641 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9642 (*call_indirect_aix<mode>): Disable for
9643 -mno-speculate-indirect-jumps.
9644 (*call_indirect_aix<mode>_nospec): New define_insn.
9645 (*call_value_indirect_aix<mode>): Disable for
9646 -mno-speculate-indirect-jumps.
9647 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9648 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9649 -mno-speculate-indirect-jumps.
9650 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9652 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9654 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9655 long double type, set the flags for noting the default long double
9656 type, even if we don't pass or return a long double type.
9658 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9661 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9664 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9666 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9667 support for merge[hl].
9668 (fold_mergehl_helper): New helper function.
9669 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9670 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9671 (altivec_vmrglw_direct): Add xxmrglw insn.
9673 2018-01-17 Andrew Waterman <andrew@sifive.com>
9675 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9676 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9678 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9681 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9682 call the lto_location_cache before reading the
9683 DECL_SOURCE_LOCATION of the types.
9685 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9686 Richard Sandiford <richard.sandiford@linaro.org>
9688 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9689 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9690 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9691 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9692 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9694 * config/aarch64/constraints.md (aarch64_movti_operand):
9696 * config/aarch64/predicates.md (Uti): Add new constraint.
9698 2018-01-17 Carl Love <cel@us.ibm.com>
9700 * config/rs6000/vsx.md (define_expand xl_len_r,
9701 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9702 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9704 (define_expand, define_insn): Move the shift left from the
9705 define_insn to the define_expand for lxvl and stxvl instructions.
9706 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9707 and XL_LEN_R definitions to PURE.
9709 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9711 * config/i386/i386.c (indirect_thunk_name): Declare regno
9712 as unsigned int. Compare regno with INVALID_REGNUM.
9713 (output_indirect_thunk): Ditto.
9714 (output_indirect_thunk_function): Ditto.
9715 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9716 in the call to output_indirect_thunk_function.
9718 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9721 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9722 rather than the size of inner_type to determine the stack slot size
9723 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9725 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9728 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9731 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9733 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9734 endian Linux systems to optionally enable multilibs for selecting
9735 the long double type if the user configured an explicit type.
9736 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9737 have no long double multilibs if not defined.
9738 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9739 warn if the user used -mabi={ieee,ibm}longdouble and we built
9740 multilibs for long double.
9741 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9742 appropriate multilib option.
9743 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9745 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9746 for building long double multilibs.
9747 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9749 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9751 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9754 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9756 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9759 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9762 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9765 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9767 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9768 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9770 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9772 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9773 different rtl trees depending on TARGET_64BIT.
9774 (rs6000_gen_lvx): Likewise.
9776 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9778 * config/visium/visium.md (nop): Tweak comment.
9779 (hazard_nop): Likewise.
9781 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9783 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9784 -mspeculate-indirect-jumps.
9785 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9786 for -mno-speculate-indirect-jumps.
9787 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9788 (*call_value_indirect_elfv2<mode>): Disable for
9789 -mno-speculate-indirect-jumps.
9790 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9791 (indirect_jump): Emit different RTL for
9792 -mno-speculate-indirect-jumps.
9793 (*indirect_jump<mode>): Disable for
9794 -mno-speculate-indirect-jumps.
9795 (*indirect_jump<mode>_nospec): New define_insn.
9796 (tablejump): Emit different RTL for
9797 -mno-speculate-indirect-jumps.
9798 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9799 (tablejumpsi_nospec): New define_expand.
9800 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9801 (tablejumpdi_nospec): New define_expand.
9802 (*tablejump<mode>_internal1): Disable for
9803 -mno-speculate-indirect-jumps.
9804 (*tablejump<mode>_internal1_nospec): New define_insn.
9805 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9808 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9810 * caller-save.c (insert_save): Drop unnecessary parameter. All
9813 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9814 Richard Biener <rguenth@suse.de>
9817 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9818 return early, inline manually is_gimple_sizepos. Make sure if we
9819 call gimplify_expr we don't end up with a gimple constant.
9820 * tree.c (variably_modified_type_p): Don't return true for
9821 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9822 * gimplify.h (is_gimple_sizepos): Remove.
9824 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9826 PR tree-optimization/83857
9827 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9828 vectorizable_live_operation for pure SLP statements.
9829 (vectorizable_live_operation): Handle PHIs.
9831 2018-01-16 Richard Biener <rguenther@suse.de>
9833 PR tree-optimization/83867
9834 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9835 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9837 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9840 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9841 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9842 If off is not INTEGER_CST, issue a may not be aligned warning
9843 rather than isn't aligned. Use isn%'t rather than isn't.
9844 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9846 <case MULT_EXPR>: Improve the case when bottom and one of the
9847 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9848 operand, in that case check if the other operand is multiple of
9849 bottom divided by the INTEGER_CST operand.
9851 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9854 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9855 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9856 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9857 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9858 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9859 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9860 (pa_function_arg_size): New function.
9862 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9864 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9865 in a separate statement.
9867 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9869 PR tree-optimization/83847
9870 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9871 group gathers and scatters.
9873 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9875 PR rtl-optimization/86620
9876 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9878 PR rtl-optimization/83213
9879 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9880 to last if both are JUMP_INSNs.
9882 PR tree-optimization/83843
9883 * gimple-ssa-store-merging.c
9884 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9885 store_immediate_info for bswap/nop orig_stores.
9887 2018-01-15 Andrew Waterman <andrew@sifive.com>
9889 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9891 <UDIV>: Increase cost if !TARGET_DIV.
9893 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9895 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9896 (define_attr "cr_logical_3op"): New.
9897 (cceq_ior_compare): Adjust.
9898 (cceq_ior_compare_complement): Adjust.
9899 (*cceq_rev_compare): Adjust.
9900 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9901 (is_cracked_insn): Adjust.
9902 (insn_must_be_first_in_group): Adjust.
9903 * config/rs6000/40x.md: Adjust.
9904 * config/rs6000/440.md: Adjust.
9905 * config/rs6000/476.md: Adjust.
9906 * config/rs6000/601.md: Adjust.
9907 * config/rs6000/603.md: Adjust.
9908 * config/rs6000/6xx.md: Adjust.
9909 * config/rs6000/7450.md: Adjust.
9910 * config/rs6000/7xx.md: Adjust.
9911 * config/rs6000/8540.md: Adjust.
9912 * config/rs6000/cell.md: Adjust.
9913 * config/rs6000/e300c2c3.md: Adjust.
9914 * config/rs6000/e500mc.md: Adjust.
9915 * config/rs6000/e500mc64.md: Adjust.
9916 * config/rs6000/e5500.md: Adjust.
9917 * config/rs6000/e6500.md: Adjust.
9918 * config/rs6000/mpc.md: Adjust.
9919 * config/rs6000/power4.md: Adjust.
9920 * config/rs6000/power5.md: Adjust.
9921 * config/rs6000/power6.md: Adjust.
9922 * config/rs6000/power7.md: Adjust.
9923 * config/rs6000/power8.md: Adjust.
9924 * config/rs6000/power9.md: Adjust.
9925 * config/rs6000/rs64.md: Adjust.
9926 * config/rs6000/titan.md: Adjust.
9928 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9930 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9931 ix86_indirect_branch_register logic.
9933 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9935 * config/i386/constraints.md (Bs): Update
9936 ix86_indirect_branch_register check. Don't check
9937 ix86_indirect_branch_register with GOT_memory_operand.
9939 * config/i386/predicates.md (GOT_memory_operand): Don't check
9940 ix86_indirect_branch_register here.
9941 (GOT32_symbol_operand): Likewise.
9943 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9945 * config/i386/predicates.md (constant_call_address_operand):
9946 Rewrite ix86_indirect_branch_register logic.
9947 (sibcall_insn_operand): Likewise.
9949 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9951 * config/i386/constraints.md (Bs): Replace
9952 ix86_indirect_branch_thunk_register with
9953 ix86_indirect_branch_register.
9955 * config/i386/i386.md (indirect_jump): Likewise.
9956 (tablejump): Likewise.
9957 (*sibcall_memory): Likewise.
9958 (*sibcall_value_memory): Likewise.
9959 Peepholes of indirect call and jump via memory: Likewise.
9960 * config/i386/i386.opt: Likewise.
9961 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9962 (GOT_memory_operand): Likewise.
9963 (call_insn_operand): Likewise.
9964 (sibcall_insn_operand): Likewise.
9965 (GOT32_symbol_operand): Likewise.
9967 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9970 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9971 type rather than type addr's type points to.
9972 (expand_omp_atomic_mutex): Likewise.
9973 (expand_omp_atomic): Likewise.
9975 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9978 * config/i386/i386.c (output_indirect_thunk_function): Use
9979 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9980 for __x86_return_thunk.
9982 2018-01-15 Richard Biener <rguenther@suse.de>
9985 * expmed.c (extract_bit_field_1): Fix typo.
9987 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9990 * config/arm/iterators.md (VF): New mode iterator.
9991 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9992 Remove integer-related logic from pattern.
9993 (neon_vabd<mode>_3): Likewise.
9995 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9998 * common.opt (fstrict-overflow): No longer an alias.
9999 (fwrapv-pointer): New option.
10000 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10001 also for pointer types based on flag_wrapv_pointer.
10002 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10003 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10004 opts->x_flag_wrapv got set.
10005 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10006 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10007 POINTER_TYPE_OVERFLOW_UNDEFINED.
10008 * match.pd: Likewise in address comparison pattern.
10009 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10011 2018-01-15 Richard Biener <rguenther@suse.de>
10014 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10015 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10016 Reset type names to their identifier if their TYPE_DECL doesn't
10017 have linkage (and thus is used for ODR and devirt).
10018 (save_debug_info_for_decl): Remove.
10019 (save_debug_info_for_type): Likewise.
10020 (add_tree_to_fld_list): Adjust.
10021 * tree-pretty-print.c (dump_generic_node): Make dumping of
10022 type names more robust.
10024 2018-01-15 Richard Biener <rguenther@suse.de>
10026 * BASE-VER: Bump to 8.0.1.
10028 2018-01-14 Martin Sebor <msebor@redhat.com>
10031 * builtins.c (check_access): Avoid warning when the no-warning bit
10034 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10036 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10037 * ira-color (allocno_hard_regs_compare): Likewise.
10039 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10042 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10043 Use .pushsection/.popsection.
10045 2018-01-14 Martin Sebor <msebor@redhat.com>
10048 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10050 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10052 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10053 entry from extra_headers.
10054 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10055 extra_headers, make the list bitwise identical to the i?86-*-* one.
10057 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10059 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10060 -mcmodel=large with -mindirect-branch=thunk,
10061 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10062 -mfunction-return=thunk-extern.
10063 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10064 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10065 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10067 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10069 * config/i386/i386.c (print_reg): Print the name of the full
10070 integer register without '%'.
10071 (ix86_print_operand): Handle 'V'.
10072 * doc/extend.texi: Document 'V' modifier.
10074 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10076 * config/i386/constraints.md (Bs): Disallow memory operand for
10077 -mindirect-branch-register.
10079 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10080 (GOT_memory_operand): Likewise.
10081 (call_insn_operand): Likewise.
10082 (sibcall_insn_operand): Likewise.
10083 (GOT32_symbol_operand): Likewise.
10084 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10085 for -mindirect-branch-register.
10086 (tablejump): Likewise.
10087 (*sibcall_memory): Likewise.
10088 (*sibcall_value_memory): Likewise.
10089 Disallow peepholes of indirect call and jump via memory for
10090 -mindirect-branch-register.
10091 (*call_pop): Replace m with Bw.
10092 (*call_value_pop): Likewise.
10093 (*sibcall_pop_memory): Replace m with Bs.
10094 * config/i386/i386.opt (mindirect-branch-register): New option.
10095 * doc/invoke.texi: Document -mindirect-branch-register option.
10097 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10099 * config/i386/i386-protos.h (ix86_output_function_return): New.
10100 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10101 set function_return_type.
10102 (indirect_thunk_name): Add ret_p to indicate thunk for function
10104 (output_indirect_thunk_function): Pass false to
10105 indirect_thunk_name.
10106 (ix86_output_indirect_branch_via_reg): Likewise.
10107 (ix86_output_indirect_branch_via_push): Likewise.
10108 (output_indirect_thunk_function): Create alias for function
10109 return thunk if regno < 0.
10110 (ix86_output_function_return): New function.
10111 (ix86_handle_fndecl_attribute): Handle function_return.
10112 (ix86_attribute_table): Add function_return.
10113 * config/i386/i386.h (machine_function): Add
10114 function_return_type.
10115 * config/i386/i386.md (simple_return_internal): Use
10116 ix86_output_function_return.
10117 (simple_return_internal_long): Likewise.
10118 * config/i386/i386.opt (mfunction-return=): New option.
10119 (indirect_branch): Mention -mfunction-return=.
10120 * doc/extend.texi: Document function_return function attribute.
10121 * doc/invoke.texi: Document -mfunction-return= option.
10123 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10125 * config/i386/i386-opts.h (indirect_branch): New.
10126 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10127 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10128 with local indirect jump when converting indirect call and jump.
10129 (ix86_set_indirect_branch_type): New.
10130 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10131 (indirectlabelno): New.
10132 (indirect_thunk_needed): Likewise.
10133 (indirect_thunk_bnd_needed): Likewise.
10134 (indirect_thunks_used): Likewise.
10135 (indirect_thunks_bnd_used): Likewise.
10136 (INDIRECT_LABEL): Likewise.
10137 (indirect_thunk_name): Likewise.
10138 (output_indirect_thunk): Likewise.
10139 (output_indirect_thunk_function): Likewise.
10140 (ix86_output_indirect_branch_via_reg): Likewise.
10141 (ix86_output_indirect_branch_via_push): Likewise.
10142 (ix86_output_indirect_branch): Likewise.
10143 (ix86_output_indirect_jmp): Likewise.
10144 (ix86_code_end): Call output_indirect_thunk_function if needed.
10145 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10147 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10148 (ix86_attribute_table): Add indirect_branch.
10149 * config/i386/i386.h (machine_function): Add indirect_branch_type
10150 and has_local_indirect_jump.
10151 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10153 (tablejump): Likewise.
10154 (*indirect_jump): Use ix86_output_indirect_jmp.
10155 (*tablejump_1): Likewise.
10156 (simple_return_indirect_internal): Likewise.
10157 * config/i386/i386.opt (mindirect-branch=): New option.
10158 (indirect_branch): New.
10161 (thunk-inline): Likewise.
10162 (thunk-extern): Likewise.
10163 * doc/extend.texi: Document indirect_branch function attribute.
10164 * doc/invoke.texi: Document -mindirect-branch= option.
10166 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10169 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10171 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10173 * ipa-inline.c (want_inline_small_function_p): Return false if
10174 inlining has already failed with CIF_FINAL_ERROR.
10175 (update_caller_keys): Call want_inline_small_function_p before
10177 (update_callee_keys): Likewise.
10179 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10181 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10183 (rs6000_quadword_masked_address_p): Likewise.
10184 (quad_aligned_load_p): Likewise.
10185 (quad_aligned_store_p): Likewise.
10186 (const_load_sequence_p): Add comment to describe the outer-most loop.
10187 (mimic_memory_attributes_and_flags): New function.
10188 (rs6000_gen_stvx): Likewise.
10189 (replace_swapped_aligned_store): Likewise.
10190 (rs6000_gen_lvx): Likewise.
10191 (replace_swapped_aligned_load): Likewise.
10192 (replace_swapped_load_constant): Capitalize argument name in
10193 comment describing this function.
10194 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10195 and stores that access quad-word aligned addresses and replace
10196 with stvx or lvx instructions when appropriate.
10197 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10198 New function prototype.
10199 (rs6000_quadword_masked_address_p): Likewise.
10200 (rs6000_gen_lvx): Likewise.
10201 (rs6000_gen_stvx): Likewise.
10202 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10203 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10204 when memory address is aligned.
10205 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10206 this split to select lvx instruction when memory address is aligned.
10207 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10208 instruction when memory address is aligned.
10209 (*vsx_le_perm_load_v16qi): Likewise.
10210 (four unnamed splitters): Modify to select the stvx instruction
10211 when memory is aligned.
10213 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10215 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10216 which appears in the queue multiple times.
10218 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10219 Alan Hayward <alan.hayward@arm.com>
10220 David Sherwood <david.sherwood@arm.com>
10222 * tree-vectorizer.h (vec_lower_bound): New structure.
10223 (_loop_vec_info): Add check_nonzero and lower_bounds.
10224 (LOOP_VINFO_CHECK_NONZERO): New macro.
10225 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10226 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10227 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10228 fields. Make seg_len the distance travelled, not including the
10230 (dr_direction_indicator): Declare.
10231 (dr_zero_step_indicator): Likewise.
10232 (dr_known_forward_stride_p): Likewise.
10233 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10235 (runtime_alias_check_p): Allow runtime alias checks with
10237 (operator ==): Compare access_size and align.
10238 (prune_runtime_alias_test_list): Rework for new distinction between
10239 the access_size and seg_len.
10240 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10242 (get_segment_min_max): New function.
10243 (create_intersect_range_checks): Use it.
10244 (dr_step_indicator): New function.
10245 (dr_direction_indicator): Likewise.
10246 (dr_zero_step_indicator): Likewise.
10247 (dr_known_forward_stride_p): Likewise.
10248 * tree-loop-distribution.c (data_ref_segment_size): Return
10249 DR_STEP * (niters - 1).
10250 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10252 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10253 (vect_preserves_scalar_order_p): New function, split out from...
10254 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10255 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10256 (vect_vfa_access_size): New function.
10257 (vect_vfa_align): Likewise.
10258 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10259 (dump_lower_bound): New function.
10260 (vect_check_lower_bound): Likewise.
10261 (vect_small_gap_p): Likewise.
10262 (vectorizable_with_step_bound_p): Likewise.
10263 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10264 depencies if the vectorization factor is 1. Convert the checks
10265 for nonzero steps into checks on the bounds of DR_STEP. Try using
10266 a bunds check for variable steps if the minimum required step is
10267 relatively small. Update calls to the dr_with_seg_len
10268 constructor and to vect_compile_time_alias.
10269 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10271 (vect_loop_versioning): Call it.
10272 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10274 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10276 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10277 Alan Hayward <alan.hayward@arm.com>
10278 David Sherwood <david.sherwood@arm.com>
10280 * doc/sourcebuild.texi (vect_scatter_store): Document.
10281 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10283 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10285 * genopinit.c (main): Add supports_vec_scatter_store and
10286 supports_vec_scatter_store_cached to target_optabs.
10287 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10288 IFN_MASK_SCATTER_STORE.
10289 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10291 * internal-fn.h (internal_store_fn_p): Declare.
10292 (internal_fn_stored_value_index): Likewise.
10293 * internal-fn.c (scatter_store_direct): New macro.
10294 (expand_scatter_store_optab_fn): New function.
10295 (direct_scatter_store_optab_supported_p): New macro.
10296 (internal_store_fn_p): New function.
10297 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10298 IFN_MASK_SCATTER_STORE.
10299 (internal_fn_mask_index): Likewise.
10300 (internal_fn_stored_value_index): New function.
10301 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10302 for scatter stores.
10303 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10304 * optabs-query.c (supports_vec_scatter_store_p): New function.
10305 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10306 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10307 true for scatter stores.
10308 (vect_gather_scatter_fn_p): Handle scatter stores too.
10309 (vect_check_gather_scatter): Consider using scatter stores if
10310 supports_vec_scatter_store_p.
10311 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10312 scatter stores too.
10313 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10314 internal_fn_stored_value_index.
10315 (check_load_store_masking): Handle scatter stores too.
10316 (vect_get_store_rhs): Make public.
10317 (vectorizable_call): Use internal_store_fn_p.
10318 (vectorizable_store): Handle scatter store internal functions.
10319 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10320 when deciding whether the end of the group has been reached.
10321 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10322 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10323 (mask_scatter_store<mode>): New insns.
10325 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10326 Alan Hayward <alan.hayward@arm.com>
10327 David Sherwood <david.sherwood@arm.com>
10329 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10330 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10331 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10333 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10334 Use vect_truncate_gather_scatter_offset if we can't treat the
10335 operation as a normal gather load or scatter store.
10336 (get_group_load_store_type): Take the gather_scatter_info
10337 as argument. Try using a gather load or scatter store for
10338 single-element groups.
10339 (get_load_store_type): Update calls to get_group_load_store_type
10340 and vect_use_strided_gather_scatters_p.
10342 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10343 Alan Hayward <alan.hayward@arm.com>
10344 David Sherwood <david.sherwood@arm.com>
10346 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10347 optional tree argument.
10348 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10350 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10351 but continue to use the current value as a fallback.
10352 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10353 to compare the updates.
10354 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10355 (get_load_store_type): Use it when handling a strided access.
10356 (vect_get_strided_load_store_ops): New function.
10357 (vect_get_data_ptr_increment): Likewise.
10358 (vectorizable_load): Handle strided gather loads. Always pass
10359 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10361 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10362 Alan Hayward <alan.hayward@arm.com>
10363 David Sherwood <david.sherwood@arm.com>
10365 * doc/md.texi (gather_load@var{m}): Document.
10366 (mask_gather_load@var{m}): Likewise.
10367 * genopinit.c (main): Add supports_vec_gather_load and
10368 supports_vec_gather_load_cached to target_optabs.
10369 * optabs-tree.c (init_tree_optimization_optabs): Use
10370 ggc_cleared_alloc to allocate target_optabs.
10371 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10372 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10374 * internal-fn.h (internal_load_fn_p): Declare.
10375 (internal_gather_scatter_fn_p): Likewise.
10376 (internal_fn_mask_index): Likewise.
10377 (internal_gather_scatter_fn_supported_p): Likewise.
10378 * internal-fn.c (gather_load_direct): New macro.
10379 (expand_gather_load_optab_fn): New function.
10380 (direct_gather_load_optab_supported_p): New macro.
10381 (direct_internal_fn_optab): New function.
10382 (internal_load_fn_p): Likewise.
10383 (internal_gather_scatter_fn_p): Likewise.
10384 (internal_fn_mask_index): Likewise.
10385 (internal_gather_scatter_fn_supported_p): Likewise.
10386 * optabs-query.c (supports_at_least_one_mode_p): New function.
10387 (supports_vec_gather_load_p): Likewise.
10388 * optabs-query.h (supports_vec_gather_load_p): Declare.
10389 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10390 and memory_type field.
10391 (NUM_PATTERNS): Bump to 15.
10392 * tree-vect-data-refs.c: Include internal-fn.h.
10393 (vect_gather_scatter_fn_p): New function.
10394 (vect_describe_gather_scatter_call): Likewise.
10395 (vect_check_gather_scatter): Try using internal functions for
10396 gather loads. Recognize existing calls to a gather load function.
10397 (vect_analyze_data_refs): Consider using gather loads if
10398 supports_vec_gather_load_p.
10399 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10400 (vect_get_gather_scatter_offset_type): Likewise.
10401 (vect_convert_mask_for_vectype): Likewise.
10402 (vect_add_conversion_to_patterm): Likewise.
10403 (vect_try_gather_scatter_pattern): Likewise.
10404 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10405 (vect_vect_recog_func_ptrs): Add it.
10406 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10407 internal_fn_mask_index and internal_gather_scatter_fn_p.
10408 (check_load_store_masking): Take the gather_scatter_info as an
10409 argument and handle gather loads.
10410 (vect_get_gather_scatter_ops): New function.
10411 (vectorizable_call): Check internal_load_fn_p.
10412 (vectorizable_load): Likewise. Handle gather load internal
10414 (vectorizable_store): Update call to check_load_store_masking.
10415 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10416 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10417 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10418 (aarch64_gather_scale_operand_d): New predicates.
10419 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10420 (mask_gather_load<mode>): New insns.
10422 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10423 Alan Hayward <alan.hayward@arm.com>
10424 David Sherwood <david.sherwood@arm.com>
10426 * optabs.def (fold_left_plus_optab): New optab.
10427 * doc/md.texi (fold_left_plus_@var{m}): Document.
10428 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10429 * internal-fn.c (fold_left_direct): Define.
10430 (expand_fold_left_optab_fn): Likewise.
10431 (direct_fold_left_optab_supported_p): Likewise.
10432 * fold-const-call.c (fold_const_fold_left): New function.
10433 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10434 * tree-parloops.c (valid_reduction_p): New function.
10435 (gather_scalar_reductions): Use it.
10436 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10437 (vect_finish_replace_stmt): Declare.
10438 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10439 (needs_fold_left_reduction_p): New function, split out from...
10440 (vect_is_simple_reduction): ...here. Accept reductions that
10441 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10442 (vect_force_simple_reduction): Also store the reduction type in
10443 the assignment's STMT_VINFO_REDUC_TYPE.
10444 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10445 (merge_with_identity): New function.
10446 (vect_expand_fold_left): Likewise.
10447 (vectorize_fold_left_reduction): Likewise.
10448 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10449 scalar phi in place for it. Check for target support and reject
10450 cases that would reassociate the operation. Defer the transform
10451 phase to vectorize_fold_left_reduction.
10452 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10453 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10454 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10456 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10458 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10459 call to ifc_temp_var.
10461 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10462 Alan Hayward <alan.hayward@arm.com>
10463 David Sherwood <david.sherwood@arm.com>
10465 * target.def (legitimize_address_displacement): Take the original
10466 offset as a poly_int.
10467 * targhooks.h (default_legitimize_address_displacement): Update
10469 * targhooks.c (default_legitimize_address_displacement): Likewise.
10470 * doc/tm.texi: Regenerate.
10471 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10472 as an argument, moving assert of ad->disp == ad->disp_term to...
10473 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10474 Try calling targetm.legitimize_address_displacement before expanding
10475 the address rather than afterwards, and adjust for the new interface.
10476 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10477 Match the new hook interface. Handle SVE addresses.
10478 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10479 new hook interface.
10481 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10483 * Makefile.in (OBJS): Add early-remat.o.
10484 * target.def (select_early_remat_modes): New hook.
10485 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10486 * doc/tm.texi: Regenerate.
10487 * targhooks.h (default_select_early_remat_modes): Declare.
10488 * targhooks.c (default_select_early_remat_modes): New function.
10489 * timevar.def (TV_EARLY_REMAT): New timevar.
10490 * passes.def (pass_early_remat): New pass.
10491 * tree-pass.h (make_pass_early_remat): Declare.
10492 * early-remat.c: New file.
10493 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10495 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10498 Alan Hayward <alan.hayward@arm.com>
10499 David Sherwood <david.sherwood@arm.com>
10501 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10502 vfm1 with a bound_epilog parameter.
10503 (vect_do_peeling): Update calls accordingly, and move the prologue
10504 call earlier in the function. Treat the base bound_epilog as 0 for
10505 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10506 this base when peeling for gaps.
10507 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10508 with fully-masked loops.
10509 (vect_estimate_min_profitable_iters): Handle the single peeled
10510 iteration in that case.
10512 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10513 Alan Hayward <alan.hayward@arm.com>
10514 David Sherwood <david.sherwood@arm.com>
10516 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10517 single-element interleaving even if the size is not a power of 2.
10518 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10519 accesses for single-element interleaving if the group size is
10522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10523 Alan Hayward <alan.hayward@arm.com>
10524 David Sherwood <david.sherwood@arm.com>
10526 * doc/md.texi (fold_extract_last_@var{m}): Document.
10527 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10528 * optabs.def (fold_extract_last_optab): New optab.
10529 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10530 * internal-fn.c (fold_extract_direct): New macro.
10531 (expand_fold_extract_optab_fn): Likewise.
10532 (direct_fold_extract_optab_supported_p): Likewise.
10533 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10534 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10535 EXTRACT_LAST_REDUCTION.
10536 (get_initial_def_for_reduction): Do not create an initial vector
10537 for EXTRACT_LAST_REDUCTION reductions.
10538 (vectorizable_reduction): Leave the scalar phi in place for
10539 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10540 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10541 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10542 transform phase to vectorizable_condition.
10543 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10545 (vect_finish_stmt_generation): ...here.
10546 (vect_finish_replace_stmt): New function.
10547 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10548 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10550 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10552 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10553 Alan Hayward <alan.hayward@arm.com>
10554 David Sherwood <david.sherwood@arm.com>
10556 * doc/md.texi (extract_last_@var{m}): Document.
10557 * optabs.def (extract_last_optab): New optab.
10558 * internal-fn.def (EXTRACT_LAST): New internal function.
10559 * internal-fn.c (cond_unary_direct): New macro.
10560 (expand_cond_unary_optab_fn): Likewise.
10561 (direct_cond_unary_optab_supported_p): Likewise.
10562 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10563 loops using EXTRACT_LAST.
10564 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10565 (extract_last_<mode>): ...this optab.
10566 (vec_extract<mode><Vel>): Update accordingly.
10568 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10569 Alan Hayward <alan.hayward@arm.com>
10570 David Sherwood <david.sherwood@arm.com>
10572 * target.def (empty_mask_is_expensive): New hook.
10573 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10574 * doc/tm.texi: Regenerate.
10575 * targhooks.h (default_empty_mask_is_expensive): Declare.
10576 * targhooks.c (default_empty_mask_is_expensive): New function.
10577 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10578 if the target says that empty masks are expensive.
10579 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10581 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10583 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10584 Alan Hayward <alan.hayward@arm.com>
10585 David Sherwood <david.sherwood@arm.com>
10587 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10588 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10589 (vect_use_loop_mask_for_alignment_p): New function.
10590 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10591 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10592 niters_skip argument. Make sure that the first niters_skip elements
10593 of the first iteration are inactive.
10594 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10595 Update call to vect_set_loop_masks_directly.
10596 (get_misalign_in_elems): New function, split out from...
10597 (vect_gen_prolog_loop_niters): ...here.
10598 (vect_update_init_of_dr): Take a code argument that specifies whether
10599 the adjustment should be added or subtracted.
10600 (vect_update_init_of_drs): Likewise.
10601 (vect_prepare_for_masked_peels): New function.
10602 (vect_do_peeling): Skip prologue peeling if we're using a mask
10603 instead. Update call to vect_update_inits_of_drs.
10604 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10606 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10607 alignment. Do not include the number of peeled iterations in
10608 the minimum threshold in that case.
10609 (vectorizable_induction): Adjust the start value down by
10610 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10611 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10612 Take the number of skipped iterations into account when calculating
10614 * tree-vect-stmts.c (vect_gen_while_not): New function.
10616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10617 Alan Hayward <alan.hayward@arm.com>
10618 David Sherwood <david.sherwood@arm.com>
10620 * doc/sourcebuild.texi (vect_fully_masked): Document.
10621 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10622 default value to 0.
10623 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10625 (vect_analyze_loop_2): ...here. Don't check the vectorization
10626 factor against the number of loop iterations if the loop is
10629 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10630 Alan Hayward <alan.hayward@arm.com>
10631 David Sherwood <david.sherwood@arm.com>
10633 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10634 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10635 (dump_groups): Update accordingly.
10636 (iv_use::mem_type): New member variable.
10637 (address_p): New function.
10638 (record_use): Add a mem_type argument and initialize the new
10640 (record_group_use): Add a mem_type argument. Use address_p.
10641 Remove obsolete null checks of base_object. Update call to record_use.
10642 (find_interesting_uses_op): Update call to record_group_use.
10643 (find_interesting_uses_cond): Likewise.
10644 (find_interesting_uses_address): Likewise.
10645 (get_mem_type_for_internal_fn): New function.
10646 (find_address_like_use): Likewise.
10647 (find_interesting_uses_stmt): Try find_address_like_use before
10648 calling find_interesting_uses_op.
10649 (addr_offset_valid_p): Use the iv mem_type field as the type
10650 of the addressed memory.
10651 (add_autoinc_candidates): Likewise.
10652 (get_address_cost): Likewise.
10653 (split_small_address_groups_p): Use address_p.
10654 (split_address_groups): Likewise.
10655 (add_iv_candidate_for_use): Likewise.
10656 (autoinc_possible_for_pair): Likewise.
10657 (rewrite_groups): Likewise.
10658 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10659 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10660 (get_alias_ptr_type_for_ptr_address): New function.
10661 (rewrite_use_address): Rewrite address uses in calls that were
10662 identified by find_address_like_use.
10664 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10665 Alan Hayward <alan.hayward@arm.com>
10666 David Sherwood <david.sherwood@arm.com>
10668 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10670 * gimple-expr.h (is_gimple_addressable: Likewise.
10671 * gimple-expr.c (is_gimple_address): Likewise.
10672 * internal-fn.c (expand_call_mem_ref): New function.
10673 (expand_mask_load_optab_fn): Use it.
10674 (expand_mask_store_optab_fn): Likewise.
10676 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10677 Alan Hayward <alan.hayward@arm.com>
10678 David Sherwood <david.sherwood@arm.com>
10680 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10681 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10682 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10683 (cond_umax@var{mode}): Document.
10684 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10685 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10686 (cond_umin_optab, cond_umax_optab): New optabs.
10687 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10688 (COND_IOR, COND_XOR): New internal functions.
10689 * internal-fn.h (get_conditional_internal_fn): Declare.
10690 * internal-fn.c (cond_binary_direct): New macro.
10691 (expand_cond_binary_optab_fn): Likewise.
10692 (direct_cond_binary_optab_supported_p): Likewise.
10693 (get_conditional_internal_fn): New function.
10694 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10695 Cope with reduction statements that are vectorized as calls rather
10697 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10698 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10699 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10700 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10701 (UNSPEC_COND_EOR): New unspecs.
10702 (optab): Add mappings for them.
10703 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10704 (sve_int_op, sve_fp_op): New int attributes.
10706 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10707 Alan Hayward <alan.hayward@arm.com>
10708 David Sherwood <david.sherwood@arm.com>
10710 * optabs.def (while_ult_optab): New optab.
10711 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10712 * internal-fn.def (WHILE_ULT): New internal function.
10713 * internal-fn.h (direct_internal_fn_supported_p): New override
10714 that takes two types as argument.
10715 * internal-fn.c (while_direct): New macro.
10716 (expand_while_optab_fn): New function.
10717 (convert_optab_supported_p): Likewise.
10718 (direct_while_optab_supported_p): New macro.
10719 * wide-int.h (wi::udiv_ceil): New function.
10720 * tree-vectorizer.h (rgroup_masks): New structure.
10721 (vec_loop_masks): New typedef.
10722 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10723 and fully_masked_p.
10724 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10725 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10726 (vect_max_vf): New function.
10727 (slpeel_make_loop_iterate_ntimes): Delete.
10728 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10729 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10730 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10731 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10732 internal-fn.h, stor-layout.h and optabs-query.h.
10733 (vect_set_loop_mask): New function.
10734 (add_preheader_seq): Likewise.
10735 (add_header_seq): Likewise.
10736 (interleave_supported_p): Likewise.
10737 (vect_maybe_permute_loop_masks): Likewise.
10738 (vect_set_loop_masks_directly): Likewise.
10739 (vect_set_loop_condition_masked): Likewise.
10740 (vect_set_loop_condition_unmasked): New function, split out from
10741 slpeel_make_loop_iterate_ntimes.
10742 (slpeel_make_loop_iterate_ntimes): Rename to..
10743 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10744 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10745 (vect_do_peeling): Update call accordingly.
10746 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10748 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10749 mask_compare_type, can_fully_mask_p and fully_masked_p.
10750 (release_vec_loop_masks): New function.
10751 (_loop_vec_info): Use it to free the loop masks.
10752 (can_produce_all_loop_masks_p): New function.
10753 (vect_get_max_nscalars_per_iter): Likewise.
10754 (vect_verify_full_masking): Likewise.
10755 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10756 retries, and free the mask rgroups before retrying. Check loop-wide
10757 reasons for disallowing fully-masked loops. Make the final decision
10758 about whether use a fully-masked loop or not.
10759 (vect_estimate_min_profitable_iters): Do not assume that peeling
10760 for the number of iterations will be needed for fully-masked loops.
10761 (vectorizable_reduction): Disable fully-masked loops.
10762 (vectorizable_live_operation): Likewise.
10763 (vect_halve_mask_nunits): New function.
10764 (vect_double_mask_nunits): Likewise.
10765 (vect_record_loop_mask): Likewise.
10766 (vect_get_loop_mask): Likewise.
10767 (vect_transform_loop): Handle the case in which the final loop
10768 iteration might handle a partial vector. Call vect_set_loop_condition
10769 instead of slpeel_make_loop_iterate_ntimes.
10770 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10771 (check_load_store_masking): New function.
10772 (prepare_load_store_mask): Likewise.
10773 (vectorizable_store): Handle fully-masked loops.
10774 (vectorizable_load): Likewise.
10775 (supportable_widening_operation): Use vect_halve_mask_nunits for
10777 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10778 (vect_gen_while): New function.
10779 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10780 (aarch64_uqdec<mode>): New insn.
10782 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10783 Alan Hayward <alan.hayward@arm.com>
10784 David Sherwood <david.sherwood@arm.com>
10786 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10787 (reduc_xor_scal_optab): New optabs.
10788 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10789 (reduc_xor_scal_@var{m}): Document.
10790 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10791 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10792 internal functions.
10793 * fold-const-call.c (fold_const_call): Handle them.
10794 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10795 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10796 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10797 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10798 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10799 (UNSPEC_XORV): New unspecs.
10800 (optab): Add entries for them.
10801 (BITWISEV): New int iterator.
10802 (bit_reduc_op): New int attributes.
10804 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10805 Alan Hayward <alan.hayward@arm.com>
10806 David Sherwood <david.sherwood@arm.com>
10808 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10809 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10810 * optabs.def (vec_shl_insert_optab): New optab.
10811 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10812 (duplicate_and_interleave): Likewise.
10813 * tree-vect-loop.c: Include internal-fn.h.
10814 (neutral_op_for_slp_reduction): New function, split out from
10815 get_initial_defs_for_reduction.
10816 (get_initial_def_for_reduction): Handle option 2 for variable-length
10817 vectors by loading the neutral value into a vector and then shifting
10818 the initial value into element 0.
10819 (get_initial_defs_for_reduction): Replace the code argument with
10820 the neutral value calculated by neutral_op_for_slp_reduction.
10821 Use gimple_build_vector for constant-length vectors.
10822 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10823 but the first group_size elements have a neutral value.
10824 Use duplicate_and_interleave otherwise.
10825 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10826 Update call to get_initial_defs_for_reduction. Handle SLP
10827 reductions for variable-length vectors by creating one vector
10828 result for each scalar result, with the elements associated
10829 with other scalar results stubbed out with the neutral value.
10830 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10831 Require IFN_VEC_SHL_INSERT for double reductions on
10832 variable-length vectors, or SLP reductions that have
10833 a neutral value. Require can_duplicate_and_interleave_p
10834 support for variable-length unchained SLP reductions if there
10835 is no neutral value, such as for MIN/MAX reductions. Also require
10836 the number of vector elements to be a multiple of the number of
10837 SLP statements when doing variable-length unchained SLP reductions.
10838 Update call to vect_create_epilog_for_reduction.
10839 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10840 and remove initial values.
10841 (duplicate_and_interleave): Make public.
10842 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10843 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10845 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10846 Alan Hayward <alan.hayward@arm.com>
10847 David Sherwood <david.sherwood@arm.com>
10849 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10850 (can_duplicate_and_interleave_p): New function.
10851 (vect_get_and_check_slp_defs): Take the vector of statements
10852 rather than just the current one. Remove excess parentheses.
10853 Restriction rejectinon of vect_constant_def and vect_external_def
10854 for variable-length vectors to boolean types, or types for which
10855 can_duplicate_and_interleave_p is false.
10856 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10857 (duplicate_and_interleave): New function.
10858 (vect_get_constant_vectors): Use gimple_build_vector for
10859 constant-length vectors and suitable variable-length constant
10860 vectors. Use duplicate_and_interleave for other variable-length
10861 vectors. Don't defer the update when inserting new statements.
10863 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10864 Alan Hayward <alan.hayward@arm.com>
10865 David Sherwood <david.sherwood@arm.com>
10867 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10868 min_profitable_iters doesn't go negative.
10870 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10871 Alan Hayward <alan.hayward@arm.com>
10872 David Sherwood <david.sherwood@arm.com>
10874 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10875 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10876 * optabs.def (vec_mask_load_lanes_optab): New optab.
10877 (vec_mask_store_lanes_optab): Likewise.
10878 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10879 (MASK_STORE_LANES): Likewise.
10880 * internal-fn.c (mask_load_lanes_direct): New macro.
10881 (mask_store_lanes_direct): Likewise.
10882 (expand_mask_load_optab_fn): Handle masked operations.
10883 (expand_mask_load_lanes_optab_fn): New macro.
10884 (expand_mask_store_optab_fn): Handle masked operations.
10885 (expand_mask_store_lanes_optab_fn): New macro.
10886 (direct_mask_load_lanes_optab_supported_p): Likewise.
10887 (direct_mask_store_lanes_optab_supported_p): Likewise.
10888 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10890 (vect_load_lanes_supported): Likewise.
10891 * tree-vect-data-refs.c (strip_conversion): New function.
10892 (can_group_stmts_p): Likewise.
10893 (vect_analyze_data_ref_accesses): Use it instead of checking
10894 for a pair of assignments.
10895 (vect_store_lanes_supported): Take a masked_p parameter.
10896 (vect_load_lanes_supported): Likewise.
10897 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10898 vect_store_lanes_supported and vect_load_lanes_supported.
10899 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10900 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10901 parameter. Don't allow gaps for masked accesses.
10902 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10903 and vect_load_lanes_supported.
10904 (get_load_store_type): Take a masked_p parameter and update
10905 call to get_group_load_store_type.
10906 (vectorizable_store): Update call to get_load_store_type.
10907 Handle IFN_MASK_STORE_LANES.
10908 (vectorizable_load): Update call to get_load_store_type.
10909 Handle IFN_MASK_LOAD_LANES.
10911 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10912 Alan Hayward <alan.hayward@arm.com>
10913 David Sherwood <david.sherwood@arm.com>
10915 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10917 * config/aarch64/aarch64-protos.h
10918 (aarch64_sve_struct_memory_operand_p): Declare.
10919 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10920 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10921 (VPRED, vpred): Handle SVE structure modes.
10922 * config/aarch64/constraints.md (Utx): New constraint.
10923 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10924 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10925 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10926 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10927 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10928 structure modes. Split into pieces after RA.
10929 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10930 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10932 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10933 SVE structure modes.
10934 (aarch64_classify_address): Likewise.
10935 (sizetochar): Move earlier in file.
10936 (aarch64_print_operand): Handle SVE register lists.
10937 (aarch64_array_mode): New function.
10938 (aarch64_sve_struct_memory_operand_p): Likewise.
10939 (TARGET_ARRAY_MODE): Redefine.
10941 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10942 Alan Hayward <alan.hayward@arm.com>
10943 David Sherwood <david.sherwood@arm.com>
10945 * target.def (array_mode): New target hook.
10946 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10947 * doc/tm.texi: Regenerate.
10948 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10949 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10950 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10951 targetm.array_mode.
10952 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10955 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10956 Alan Hayward <alan.hayward@arm.com>
10957 David Sherwood <david.sherwood@arm.com>
10959 * fold-const.c (fold_binary_loc): Check the argument types
10960 rather than the result type when testing for a vector operation.
10962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10964 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10965 * doc/tm.texi: Regenerate.
10967 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10968 Alan Hayward <alan.hayward@arm.com>
10969 David Sherwood <david.sherwood@arm.com>
10971 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10972 (sve): Document new AArch64 extension.
10973 * doc/md.texi (w): Extend the description of the AArch64
10974 constraint to include SVE vectors.
10975 (Upl, Upa): Document new AArch64 predicate constraints.
10976 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10978 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10979 (msve-vector-bits=): New option.
10980 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10981 SVE when these are disabled.
10982 (sve): New extension.
10983 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10984 modes. Adjust their number of units based on aarch64_sve_vg.
10985 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10986 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10987 aarch64_addr_query_type.
10988 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10989 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10990 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10991 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10992 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10993 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10994 (aarch64_simd_imm_zero_p): Delete.
10995 (aarch64_check_zero_based_sve_index_immediate): Declare.
10996 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10997 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10998 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10999 (aarch64_sve_float_mul_immediate_p): Likewise.
11000 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11001 rather than an rtx.
11002 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11003 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11004 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11005 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11006 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11007 (aarch64_regmode_natural_size): Likewise.
11008 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11009 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11011 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11012 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11013 for VG and the SVE predicate registers.
11014 (V_ALIASES): Add a "z"-prefixed alias.
11015 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11016 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11017 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11018 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11019 (REG_CLASS_NAMES): Add entries for them.
11020 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11021 and the predicate registers.
11022 (aarch64_sve_vg): Declare.
11023 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11024 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11025 (REGMODE_NATURAL_SIZE): Define.
11026 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11028 * config/aarch64/aarch64.c: Include cfgrtl.h.
11029 (simd_immediate_info): Add a constructor for series vectors,
11030 and an associated step field.
11031 (aarch64_sve_vg): New variable.
11032 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11033 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11034 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11035 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11036 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11037 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11038 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11039 (aarch64_get_mask_mode): New functions.
11040 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11041 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11042 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11043 predicate modes and predicate registers. Explicitly restrict
11044 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11045 to store a vector mode if it is recognized by
11046 aarch64_classify_vector_mode.
11047 (aarch64_regmode_natural_size): New function.
11048 (aarch64_hard_regno_caller_save_mode): Return the original mode
11050 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11051 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11052 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11053 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11055 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11056 does not overlap dest if the function is frame-related. Handle
11058 (aarch64_split_add_offset): New function.
11059 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11060 them aarch64_add_offset.
11061 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11062 and update call to aarch64_sub_sp.
11063 (aarch64_add_cfa_expression): New function.
11064 (aarch64_expand_prologue): Pass extra temporary registers to the
11065 functions above. Handle the case in which we need to emit new
11066 DW_CFA_expressions for registers that were originally saved
11067 relative to the stack pointer, but now have to be expressed
11068 relative to the frame pointer.
11069 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11071 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11072 IP0 and IP1 values for SVE frames.
11073 (aarch64_expand_vec_series): New function.
11074 (aarch64_expand_sve_widened_duplicate): Likewise.
11075 (aarch64_expand_sve_const_vector): Likewise.
11076 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11077 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11078 into the register, rather than emitting a SET directly.
11079 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11080 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11081 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11082 (offset_9bit_signed_scaled_p): New functions.
11083 (aarch64_replicate_bitmask_imm): New function.
11084 (aarch64_bitmask_imm): Use it.
11085 (aarch64_cannot_force_const_mem): Reject expressions involving
11086 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11087 (aarch64_classify_index): Handle SVE indices, by requiring
11088 a plain register index with a scale that matches the element size.
11089 (aarch64_classify_address): Handle SVE addresses. Assert that
11090 the mode of the address is VOIDmode or an integer mode.
11091 Update call to aarch64_classify_symbol.
11092 (aarch64_classify_symbolic_expression): Update call to
11093 aarch64_classify_symbol.
11094 (aarch64_const_vec_all_in_range_p): New function.
11095 (aarch64_print_vector_float_operand): Likewise.
11096 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11097 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11098 and the FP immediates 1.0 and 0.5.
11099 (aarch64_print_address_internal): Handle SVE addresses.
11100 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11101 (aarch64_regno_regclass): Handle predicate registers.
11102 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11104 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11105 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11106 (aarch64_convert_sve_vector_bits): New function.
11107 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11108 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11109 rather than an rtx.
11110 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11111 Handle SVE vector and predicate modes. Accept VL-based constants
11112 that need only one temporary register, and VL offsets that require
11113 no temporary registers.
11114 (aarch64_conditional_register_usage): Mark the predicate registers
11115 as fixed if SVE isn't available.
11116 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11117 Return true for SVE vector and predicate modes.
11118 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11119 rather than an unsigned int. Handle SVE modes.
11120 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11122 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11124 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11125 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11126 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11127 (aarch64_sve_float_mul_immediate_p): New functions.
11128 (aarch64_sve_valid_immediate): New function.
11129 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11130 Explicitly reject structure modes. Check for INDEX constants.
11131 Handle PTRUE and PFALSE constants.
11132 (aarch64_check_zero_based_sve_index_immediate): New function.
11133 (aarch64_simd_imm_zero_p): Delete.
11134 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11135 vector modes. Accept constants in the range of CNT[BHWD].
11136 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11137 ask for an Advanced SIMD mode.
11138 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11139 (aarch64_simd_vector_alignment): Handle SVE predicates.
11140 (aarch64_vectorize_preferred_vector_alignment): New function.
11141 (aarch64_simd_vector_alignment_reachable): Use it instead of
11143 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11144 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11146 (MAX_VECT_LEN): Delete.
11147 (expand_vec_perm_d): Add a vec_flags field.
11148 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11149 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11150 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11152 (aarch64_evpc_rev): Rename to...
11153 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11154 (aarch64_evpc_rev_global): New function.
11155 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11156 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11158 (aarch64_evpc_sve_tbl): New function.
11159 (aarch64_expand_vec_perm_const_1): Update after rename of
11160 aarch64_evpc_rev. Handle SVE permutes too, trying
11161 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11162 than aarch64_evpc_tbl.
11163 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11164 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11165 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11166 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11167 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11168 (aarch64_expand_sve_vcond): New functions.
11169 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11170 of aarch64_vector_mode_p.
11171 (aarch64_dwarf_poly_indeterminate_value): New function.
11172 (aarch64_compute_pressure_classes): Likewise.
11173 (aarch64_can_change_mode_class): Likewise.
11174 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11175 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11176 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11177 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11178 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11179 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11180 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11181 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11183 (Dn, Dl, Dr): Accept const as well as const_vector.
11184 (Dz): Likewise. Compare against CONST0_RTX.
11185 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11186 of "vector" where appropriate.
11187 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11188 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11189 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11190 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11191 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11192 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11193 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11194 (v_int_equiv): Extend to SVE modes.
11195 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11197 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11198 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11199 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11200 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11201 (SVE_COND_FP_CMP): New int iterators.
11202 (perm_hilo): Handle the new unpack unspecs.
11203 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11205 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11206 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11207 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11208 (aarch64_equality_operator, aarch64_constant_vector_operand)
11209 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11210 (aarch64_sve_nonimmediate_operand): Likewise.
11211 (aarch64_sve_general_operand): Likewise.
11212 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11213 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11214 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11215 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11216 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11217 (aarch64_sve_float_arith_immediate): Likewise.
11218 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11219 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11220 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11221 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11222 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11223 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11224 (aarch64_sve_float_arith_operand): Likewise.
11225 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11226 (aarch64_sve_float_mul_operand): Likewise.
11227 (aarch64_sve_vec_perm_operand): Likewise.
11228 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11229 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11230 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11231 as well as const_vector.
11232 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11233 in file. Use CONST0_RTX and CONSTM1_RTX.
11234 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11235 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11236 Use aarch64_simd_imm_zero.
11237 * config/aarch64/aarch64-sve.md: New file.
11238 * config/aarch64/aarch64.md: Include it.
11239 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11240 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11241 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11242 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11243 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11244 (sve): New attribute.
11245 (enabled): Disable instructions with the sve attribute unless
11247 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11248 aarch64_expand_mov_immediate.
11249 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11250 CNT[BHSD] immediates.
11251 (movti): Split CONST_POLY_INT moves into two halves.
11252 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11253 Split additions that need a temporary here if the destination
11254 is the stack pointer.
11255 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11256 (*add<mode>3_poly_1): New instruction.
11257 (set_clobber_cc): New expander.
11259 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11261 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11262 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11263 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11264 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11265 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11266 Change innermode from fixed_mode_size to machine_mode.
11267 (simplify_subreg): Update call accordingly. Handle a constant-sized
11268 subreg of a variable-length CONST_VECTOR.
11270 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11271 Alan Hayward <alan.hayward@arm.com>
11272 David Sherwood <david.sherwood@arm.com>
11274 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11275 (add_offset_to_base): New function, split out from...
11276 (create_mem_ref): ...here. When handling a scale other than 1,
11277 check first whether the address is valid without the offset.
11278 Add it into the base if so, leaving the index and scale as-is.
11280 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11283 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11284 fold_for_warn before checking if arg2 is INTEGER_CST.
11286 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11288 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11289 (store_multiple_operation): Delete.
11290 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11291 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11292 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11293 guarded by TARGET_STRING.
11294 (rs6000_output_load_multiple): Delete.
11295 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11296 OPTION_MASK_STRING / TARGET_STRING handling.
11297 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11298 (const rs6000_opt_masks) <"string">: Change mask to 0.
11299 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11300 (MASK_STRING): Delete.
11301 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11303 (load_multiple): Delete.
11310 (store_multiple): Delete.
11317 (movmemsi_8reg): Delete.
11318 (corresponding unnamed define_insn): Delete.
11319 (movmemsi_6reg): Delete.
11320 (corresponding unnamed define_insn): Delete.
11321 (movmemsi_4reg): Delete.
11322 (corresponding unnamed define_insn): Delete.
11323 (movmemsi_2reg): Delete.
11324 (corresponding unnamed define_insn): Delete.
11325 (movmemsi_1reg): Delete.
11326 (corresponding unnamed define_insn): Delete.
11327 * config/rs6000/rs6000.opt (mno-string): New.
11328 (mstring): Replace by deprecation warning stub.
11329 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11331 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11333 * regrename.c (regrename_do_replace): If replacing the same
11334 reg multiple times, try to reuse last created gen_raw_REG.
11337 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11338 main to workaround a bug in GDB.
11340 2018-01-12 Tom de Vries <tom@codesourcery.com>
11343 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11345 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11347 PR rtl-optimization/80481
11348 * ira-color.c (get_cap_member): New function.
11349 (allocnos_conflict_by_live_ranges_p): Use it.
11350 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11351 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11353 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11356 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11357 (*saddl_se_1): Ditto.
11358 (*ssubsi_1): Ditto.
11359 (*ssubl_se_1): Ditto.
11361 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11363 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11364 rather than wi::to_widest for DR_INITs.
11365 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11366 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11367 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11369 (vect_analyze_group_access_1): Note that here.
11371 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11373 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11374 polynomial type sizes.
11376 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11378 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11379 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11380 (gimple_add_tmp_var): Likewise.
11382 2018-01-12 Martin Liska <mliska@suse.cz>
11384 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11385 (gimple_alloc_sizes): Likewise.
11386 (dump_gimple_statistics): Use PRIu64 in printf format.
11387 * gimple.h: Change uint64_t to int.
11389 2018-01-12 Martin Liska <mliska@suse.cz>
11391 * tree-core.h: Use uint64_t instead of int.
11392 * tree.c (tree_node_counts): Likewise.
11393 (tree_node_sizes): Likewise.
11394 (dump_tree_statistics): Use PRIu64 in printf format.
11396 2018-01-12 Martin Liska <mliska@suse.cz>
11398 * Makefile.in: As qsort_chk is implemented in vec.c, add
11399 vec.o to linkage of gencfn-macros.
11400 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11401 passing the info to record_node_allocation_statistics.
11402 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11404 * ggc-common.c (struct ggc_usage): Add operator== and use
11405 it in operator< and compare function.
11406 * mem-stats.h (struct mem_usage): Likewise.
11407 * vec.c (struct vec_usage): Remove operator< and compare
11408 function. Can be simply inherited.
11410 2018-01-12 Martin Jambor <mjambor@suse.cz>
11413 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11414 * tree-ssa-math-opts.c: Include domwalk.h.
11415 (convert_mult_to_fma_1): New function.
11416 (fma_transformation_info): New type.
11417 (fma_deferring_state): Likewise.
11418 (cancel_fma_deferring): New function.
11419 (result_of_phi): Likewise.
11420 (last_fma_candidate_feeds_initial_phi): Likewise.
11421 (convert_mult_to_fma): Added deferring logic, split actual
11422 transformation to convert_mult_to_fma_1.
11423 (math_opts_dom_walker): New type.
11424 (math_opts_dom_walker::after_dom_children): New method, body moved
11425 here from pass_optimize_widening_mul::execute, added deferring logic
11427 (pass_optimize_widening_mul::execute): Moved most of code to
11428 math_opts_dom_walker::after_dom_children.
11429 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11430 * config/i386/i386.c (ix86_option_override_internal): Added
11431 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11433 2018-01-12 Richard Biener <rguenther@suse.de>
11436 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11437 inline instance vars.
11439 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11442 * config/rx/rx.c (rx_is_restricted_memory_address):
11443 Handle SUBREG case.
11445 2018-01-12 Richard Biener <rguenther@suse.de>
11447 PR tree-optimization/80846
11448 * target.def (split_reduction): New target hook.
11449 * targhooks.c (default_split_reduction): New function.
11450 * targhooks.h (default_split_reduction): Declare.
11451 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11452 target requests first reduce vectors by combining low and high
11454 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11455 (get_vectype_for_scalar_type_and_size): Export.
11456 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11457 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11458 * doc/tm.texi: Regenerate.
11459 * config/i386/i386.c (ix86_split_reduction): Implement
11460 TARGET_VECTORIZE_SPLIT_REDUCTION.
11462 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11465 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11466 in PIC mode except for TARGET_VXWORKS_RTP.
11467 * config/sparc/sparc.c: Include cfgrtl.h.
11468 (TARGET_INIT_PIC_REG): Define.
11469 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11470 (sparc_pic_register_p): New predicate.
11471 (sparc_legitimate_address_p): Use it.
11472 (sparc_legitimize_pic_address): Likewise.
11473 (sparc_delegitimize_address): Likewise.
11474 (sparc_mode_dependent_address_p): Likewise.
11475 (gen_load_pcrel_sym): Remove 4th parameter.
11476 (load_got_register): Adjust call to above. Remove obsolete stuff.
11477 (sparc_expand_prologue): Do not call load_got_register here.
11478 (sparc_flat_expand_prologue): Likewise.
11479 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11480 (sparc_use_pseudo_pic_reg): New function.
11481 (sparc_init_pic_reg): Likewise.
11482 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11483 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11485 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11487 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11488 Add item for branch_cost.
11490 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11492 PR rtl-optimization/83565
11493 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11494 not extend the result to a larger mode for rotate operations.
11495 (num_sign_bit_copies1): Likewise.
11497 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11500 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11502 Use values-Xc.o for -pedantic.
11503 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11505 2018-01-12 Martin Liska <mliska@suse.cz>
11508 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11510 (possible_polymorphic_call_targets): Use it.
11511 (ipa_devirt): Likewise.
11513 2018-01-12 Martin Liska <mliska@suse.cz>
11515 * profile-count.h (enum profile_quality): Use 0 as invalid
11516 enum value of profile_quality.
11518 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11520 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11521 -mext-string options.
11523 2018-01-12 Richard Biener <rguenther@suse.de>
11525 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11526 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11527 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11529 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11531 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11533 * configure.ac (--with-long-double-format): Add support for the
11534 configuration option to change the default long double format on
11536 * config.gcc (powerpc*-linux*-*): Likewise.
11537 * configure: Regenerate.
11538 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11539 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11540 used without modification.
11542 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11544 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11545 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11546 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11547 MISC_BUILTIN_SPEC_BARRIER.
11548 (rs6000_init_builtins): Likewise.
11549 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11551 (speculation_barrier): New define_insn.
11552 * doc/extend.texi: Document __builtin_speculation_barrier.
11554 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11557 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11558 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11559 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11561 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11562 integral modes instead of "ss" and "sd".
11563 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11564 vectors with 32-bit and 64-bit elements.
11565 (vecdupssescalarmodesuffix): New mode attribute.
11566 (vec_dup<mode>): Use it.
11568 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11571 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11572 frame if argument is passed on stack.
11574 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11577 * ree.c (combine_reaching_defs): Optimize also
11578 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11579 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11581 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11583 PR middle-end/83189
11584 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11586 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11588 PR middle-end/83718
11589 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11590 after they are computed.
11592 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11594 PR tree-optimization/83695
11595 * gimple-loop-linterchange.cc
11596 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11597 reset cached scev information after interchange.
11598 (pass_linterchange::execute): Remove call to scev_reset_htab.
11600 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11602 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11603 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11604 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11605 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11606 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11607 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11608 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11609 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11610 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11611 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11612 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11613 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11614 (V_lane_reg): Likewise.
11615 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11617 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11618 (vfmal_lane_low<mode>_intrinsic,
11619 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11620 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11621 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11622 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11623 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11624 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11626 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11628 * config/arm/arm-cpus.in (fp16fml): New feature.
11629 (ALL_SIMD): Add fp16fml.
11630 (armv8.2-a): Add fp16fml as an option.
11631 (armv8.3-a): Likewise.
11632 (armv8.4-a): Add fp16fml as part of fp16.
11633 * config/arm/arm.h (TARGET_FP16FML): Define.
11634 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11636 * config/arm/arm-modes.def (V2HF): Define.
11637 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11638 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11639 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11640 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11641 vfmsl_low, vfmsl_high): New set of builtins.
11642 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11643 (vfml_op): New code attribute.
11644 (VFMLHALVES): New int iterator.
11645 (VFML, VFMLSEL): New mode attributes.
11646 (V_reg): Define mapping for V2HF.
11647 (V_hi, V_lo): New mode attributes.
11648 (VF_constraint): Likewise.
11649 (vfml_half, vfml_half_selector): New int attributes.
11650 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11652 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11653 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11655 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11656 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11657 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11658 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11660 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11661 Document new effective target and option set.
11663 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11665 * config/arm/arm-cpus.in (armv8_4): New feature.
11666 (ARMv8_4a): New fgroup.
11667 (armv8.4-a): New arch.
11668 * config/arm/arm-tables.opt: Regenerate.
11669 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11670 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11671 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11672 Add matching rules for -march=armv8.4-a and extensions.
11673 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11675 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11678 * config/rx/rx.md (BW): New mode attribute.
11679 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11681 2018-01-11 Richard Biener <rguenther@suse.de>
11683 PR tree-optimization/83435
11684 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11685 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11686 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11688 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11689 Alan Hayward <alan.hayward@arm.com>
11690 David Sherwood <david.sherwood@arm.com>
11692 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11694 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11695 (aarch64_print_address_internal): Use it to check for a zero offset.
11697 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11698 Alan Hayward <alan.hayward@arm.com>
11699 David Sherwood <david.sherwood@arm.com>
11701 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11702 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11703 Return a poly_int64 rather than a HOST_WIDE_INT.
11704 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11705 rather than a HOST_WIDE_INT.
11706 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11707 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11708 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11709 final_offset from HOST_WIDE_INT to poly_int64.
11710 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11711 to_constant when getting the number of units in an Advanced SIMD
11713 (aarch64_builtin_vectorized_function): Check for a constant number
11715 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11717 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11718 attribute instead of GET_MODE_NUNITS.
11719 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11720 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11721 GET_MODE_SIZE for fixed-size registers.
11722 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11723 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11724 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11725 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11726 (aarch64_print_operand, aarch64_print_address_internal)
11727 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11728 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11729 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11730 Handle polynomial GET_MODE_SIZE.
11731 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11732 wider than SImode without modification.
11733 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11734 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11735 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11736 passing and returning SVE modes.
11737 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11738 rather than GEN_INT.
11739 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11740 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11741 (aarch64_allocate_and_probe_stack_space): Likewise.
11742 (aarch64_layout_frame): Cope with polynomial offsets.
11743 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11744 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11745 polynomial offsets.
11746 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11747 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11748 poly_int64 rather than a HOST_WIDE_INT.
11749 (aarch64_get_separate_components, aarch64_process_components)
11750 (aarch64_expand_prologue, aarch64_expand_epilogue)
11751 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11752 (aarch64_anchor_offset): New function, split out from...
11753 (aarch64_legitimize_address): ...here.
11754 (aarch64_builtin_vectorization_cost): Handle polynomial
11755 TYPE_VECTOR_SUBPARTS.
11756 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11758 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11759 number of elements from the PARALLEL rather than the mode.
11760 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11761 rather than GET_MODE_BITSIZE.
11762 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11763 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11764 (aarch64_expand_vec_perm_const_1): Handle polynomial
11765 d->perm.length () and d->perm elements.
11766 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11767 Apply to_constant to d->perm elements.
11768 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11769 polynomial CONST_VECTOR_NUNITS.
11770 (aarch64_move_pointer): Take amount as a poly_int64 rather
11772 (aarch64_progress_pointer): Avoid temporary variable.
11773 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11774 the mode attribute instead of GET_MODE.
11776 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11777 Alan Hayward <alan.hayward@arm.com>
11778 David Sherwood <david.sherwood@arm.com>
11780 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11781 x exists before using it.
11782 (aarch64_add_constant_internal): Rename to...
11783 (aarch64_add_offset_1): ...this. Replace regnum with separate
11784 src and dest rtxes. Handle the case in which they're different,
11785 including when the offset is zero. Replace scratchreg with an rtx.
11786 Use 2 additions if there is no spare register into which we can
11787 move a 16-bit constant.
11788 (aarch64_add_constant): Delete.
11789 (aarch64_add_offset): Replace reg with separate src and dest
11790 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11791 Use aarch64_add_offset_1.
11792 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11793 an rtx rather than an int. Take the delta as a poly_int64
11794 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11795 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11796 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11797 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11798 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11799 and aarch64_add_sp.
11800 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11801 aarch64_add_constant.
11803 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11805 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11806 Use scalar_float_mode.
11808 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11810 * config/aarch64/aarch64-simd.md
11811 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11812 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11813 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11814 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11815 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11816 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11817 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11818 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11819 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11820 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11822 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11825 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11826 targ_options->x_arm_arch_string is non NULL.
11828 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11830 * config/aarch64/aarch64.h
11831 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11833 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11836 * expmed.c (emit_store_flag_force): Swap if const op0
11837 and change VOIDmode to mode of op0.
11839 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11841 PR rtl-optimization/83761
11842 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11843 than bytes to mode_for_size.
11845 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11847 PR middle-end/83189
11848 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11849 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11852 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11854 PR middle-end/83575
11855 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11856 when in layout mode.
11857 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11858 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11861 2018-01-10 Michael Collison <michael.collison@arm.com>
11863 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11864 * config/aarch64/aarch64-option-extension.def: Add
11865 AARCH64_OPT_EXTENSION of 'fp16fml'.
11866 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11867 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11868 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11869 * config/aarch64/constraints.md (Ui7): New constraint.
11870 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11871 (VFMLA_SEL_W): Ditto.
11874 (VFMLA16_LOW): New int iterator.
11875 (VFMLA16_HIGH): Ditto.
11876 (UNSPEC_FMLAL): New unspec.
11877 (UNSPEC_FMLSL): Ditto.
11878 (UNSPEC_FMLAL2): Ditto.
11879 (UNSPEC_FMLSL2): Ditto.
11880 (f16mac): New code attribute.
11881 * config/aarch64/aarch64-simd-builtins.def
11882 (aarch64_fmlal_lowv2sf): Ditto.
11883 (aarch64_fmlsl_lowv2sf): Ditto.
11884 (aarch64_fmlalq_lowv4sf): Ditto.
11885 (aarch64_fmlslq_lowv4sf): Ditto.
11886 (aarch64_fmlal_highv2sf): Ditto.
11887 (aarch64_fmlsl_highv2sf): Ditto.
11888 (aarch64_fmlalq_highv4sf): Ditto.
11889 (aarch64_fmlslq_highv4sf): Ditto.
11890 (aarch64_fmlal_lane_lowv2sf): Ditto.
11891 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11892 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11893 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11894 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11895 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11896 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11897 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11898 (aarch64_fmlal_lane_highv2sf): Ditto.
11899 (aarch64_fmlsl_lane_highv2sf): Ditto.
11900 (aarch64_fmlal_laneq_highv2sf): Ditto.
11901 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11902 (aarch64_fmlalq_lane_highv4sf): Ditto.
11903 (aarch64_fmlsl_lane_highv4sf): Ditto.
11904 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11905 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11906 * config/aarch64/aarch64-simd.md:
11907 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11908 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11909 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11910 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11911 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11912 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11913 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11914 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11915 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11916 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11917 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11918 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11919 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11920 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11921 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11922 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11923 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11924 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11925 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11926 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11927 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11928 (vfmlsl_low_u32): Ditto.
11929 (vfmlalq_low_u32): Ditto.
11930 (vfmlslq_low_u32): Ditto.
11931 (vfmlal_high_u32): Ditto.
11932 (vfmlsl_high_u32): Ditto.
11933 (vfmlalq_high_u32): Ditto.
11934 (vfmlslq_high_u32): Ditto.
11935 (vfmlal_lane_low_u32): Ditto.
11936 (vfmlsl_lane_low_u32): Ditto.
11937 (vfmlal_laneq_low_u32): Ditto.
11938 (vfmlsl_laneq_low_u32): Ditto.
11939 (vfmlalq_lane_low_u32): Ditto.
11940 (vfmlslq_lane_low_u32): Ditto.
11941 (vfmlalq_laneq_low_u32): Ditto.
11942 (vfmlslq_laneq_low_u32): Ditto.
11943 (vfmlal_lane_high_u32): Ditto.
11944 (vfmlsl_lane_high_u32): Ditto.
11945 (vfmlal_laneq_high_u32): Ditto.
11946 (vfmlsl_laneq_high_u32): Ditto.
11947 (vfmlalq_lane_high_u32): Ditto.
11948 (vfmlslq_lane_high_u32): Ditto.
11949 (vfmlalq_laneq_high_u32): Ditto.
11950 (vfmlslq_laneq_high_u32): Ditto.
11951 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11952 (AARCH64_FL_FOR_ARCH8_4): New.
11953 (AARCH64_ISA_F16FML): New ISA flag.
11954 (TARGET_F16FML): New feature flag for fp16fml.
11955 (doc/invoke.texi): Document new fp16fml option.
11957 2018-01-10 Michael Collison <michael.collison@arm.com>
11959 * config/aarch64/aarch64-builtins.c:
11960 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11961 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11962 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11963 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11964 (AARCH64_ISA_SHA3): New ISA flag.
11965 (TARGET_SHA3): New feature flag for sha3.
11966 * config/aarch64/iterators.md (sha512_op): New int attribute.
11967 (CRYPTO_SHA512): New int iterator.
11968 (UNSPEC_SHA512H): New unspec.
11969 (UNSPEC_SHA512H2): Ditto.
11970 (UNSPEC_SHA512SU0): Ditto.
11971 (UNSPEC_SHA512SU1): Ditto.
11972 * config/aarch64/aarch64-simd-builtins.def
11973 (aarch64_crypto_sha512hqv2di): New builtin.
11974 (aarch64_crypto_sha512h2qv2di): Ditto.
11975 (aarch64_crypto_sha512su0qv2di): Ditto.
11976 (aarch64_crypto_sha512su1qv2di): Ditto.
11977 (aarch64_eor3qv8hi): Ditto.
11978 (aarch64_rax1qv2di): Ditto.
11979 (aarch64_xarqv2di): Ditto.
11980 (aarch64_bcaxqv8hi): Ditto.
11981 * config/aarch64/aarch64-simd.md:
11982 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11983 (aarch64_crypto_sha512su0qv2di): Ditto.
11984 (aarch64_crypto_sha512su1qv2di): Ditto.
11985 (aarch64_eor3qv8hi): Ditto.
11986 (aarch64_rax1qv2di): Ditto.
11987 (aarch64_xarqv2di): Ditto.
11988 (aarch64_bcaxqv8hi): Ditto.
11989 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11990 (vsha512h2q_u64): Ditto.
11991 (vsha512su0q_u64): Ditto.
11992 (vsha512su1q_u64): Ditto.
11993 (veor3q_u16): Ditto.
11994 (vrax1q_u64): Ditto.
11995 (vxarq_u64): Ditto.
11996 (vbcaxq_u16): Ditto.
11997 * config/arm/types.md (crypto_sha512): New type attribute.
11998 (crypto_sha3): Ditto.
11999 (doc/invoke.texi): Document new sha3 option.
12001 2018-01-10 Michael Collison <michael.collison@arm.com>
12003 * config/aarch64/aarch64-builtins.c:
12004 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12005 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12006 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12007 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12008 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12009 (AARCH64_ISA_SM4): New ISA flag.
12010 (TARGET_SM4): New feature flag for sm4.
12011 * config/aarch64/aarch64-simd-builtins.def
12012 (aarch64_sm3ss1qv4si): Ditto.
12013 (aarch64_sm3tt1aq4si): Ditto.
12014 (aarch64_sm3tt1bq4si): Ditto.
12015 (aarch64_sm3tt2aq4si): Ditto.
12016 (aarch64_sm3tt2bq4si): Ditto.
12017 (aarch64_sm3partw1qv4si): Ditto.
12018 (aarch64_sm3partw2qv4si): Ditto.
12019 (aarch64_sm4eqv4si): Ditto.
12020 (aarch64_sm4ekeyqv4si): Ditto.
12021 * config/aarch64/aarch64-simd.md:
12022 (aarch64_sm3ss1qv4si): Ditto.
12023 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12024 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12025 (aarch64_sm4eqv4si): Ditto.
12026 (aarch64_sm4ekeyqv4si): Ditto.
12027 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12028 (sm3part_op): Ditto.
12029 (CRYPTO_SM3TT): Ditto.
12030 (CRYPTO_SM3PART): Ditto.
12031 (UNSPEC_SM3SS1): New unspec.
12032 (UNSPEC_SM3TT1A): Ditto.
12033 (UNSPEC_SM3TT1B): Ditto.
12034 (UNSPEC_SM3TT2A): Ditto.
12035 (UNSPEC_SM3TT2B): Ditto.
12036 (UNSPEC_SM3PARTW1): Ditto.
12037 (UNSPEC_SM3PARTW2): Ditto.
12038 (UNSPEC_SM4E): Ditto.
12039 (UNSPEC_SM4EKEY): Ditto.
12040 * config/aarch64/constraints.md (Ui2): New constraint.
12041 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12042 * config/arm/types.md (crypto_sm3): New type attribute.
12043 (crypto_sm4): Ditto.
12044 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12045 (vsm3tt1aq_u32): Ditto.
12046 (vsm3tt1bq_u32): Ditto.
12047 (vsm3tt2aq_u32): Ditto.
12048 (vsm3tt2bq_u32): Ditto.
12049 (vsm3partw1q_u32): Ditto.
12050 (vsm3partw2q_u32): Ditto.
12051 (vsm4eq_u32): Ditto.
12052 (vsm4ekeyq_u32): Ditto.
12053 (doc/invoke.texi): Document new sm4 option.
12055 2018-01-10 Michael Collison <michael.collison@arm.com>
12057 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12058 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12059 (AARCH64_FL_FOR_ARCH8_4): New.
12060 (AARCH64_FL_V8_4): New flag.
12061 (doc/invoke.texi): Document new armv8.4-a option.
12063 2018-01-10 Michael Collison <michael.collison@arm.com>
12065 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12066 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12067 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12068 * config/aarch64/aarch64-option-extension.def: Add
12069 AARCH64_OPT_EXTENSION of 'sha2'.
12070 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12071 (crypto): Disable sha2 and aes if crypto disabled.
12072 (crypto): Enable aes and sha2 if enabled.
12073 (simd): Disable sha2 and aes if simd disabled.
12074 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12076 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12077 (TARGET_SHA2): New feature flag for sha2.
12078 (TARGET_AES): New feature flag for aes.
12079 * config/aarch64/aarch64-simd.md:
12080 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12081 conditional on TARGET_AES.
12082 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12083 (aarch64_crypto_sha1hsi): Make pattern conditional
12085 (aarch64_crypto_sha1hv4si): Ditto.
12086 (aarch64_be_crypto_sha1hv4si): Ditto.
12087 (aarch64_crypto_sha1su1v4si): Ditto.
12088 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12089 (aarch64_crypto_sha1su0v4si): Ditto.
12090 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12091 (aarch64_crypto_sha256su0v4si): Ditto.
12092 (aarch64_crypto_sha256su1v4si): Ditto.
12093 (doc/invoke.texi): Document new aes and sha2 options.
12095 2018-01-10 Martin Sebor <msebor@redhat.com>
12097 PR tree-optimization/83781
12098 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12101 2018-01-11 Martin Sebor <msebor@gmail.com>
12102 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12104 PR tree-optimization/83501
12105 PR tree-optimization/81703
12107 * tree-ssa-strlen.c (get_string_cst): Rename...
12108 (get_string_len): ...to this. Handle global constants.
12109 (handle_char_store): Adjust.
12111 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12112 Jim Wilson <jimw@sifive.com>
12114 * config/riscv/riscv-protos.h (riscv_output_return): New.
12115 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12116 (riscv_attribute_table, riscv_output_return),
12117 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12118 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12119 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12121 (riscv_expand_prologue): Add early return for naked function.
12122 (riscv_expand_epilogue): Likewise.
12123 (riscv_function_ok_for_sibcall): Return false for naked function.
12124 (riscv_set_current_function): New.
12125 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12126 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12127 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12128 * doc/extend.texi (RISC-V Function Attributes): New.
12130 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12132 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12133 check for 128-bit long double before checking TCmode.
12134 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12135 128-bit long doubles before checking TFmode or TCmode.
12136 (FLOAT128_IBM_P): Likewise.
12138 2018-01-10 Martin Sebor <msebor@redhat.com>
12140 PR tree-optimization/83671
12141 * builtins.c (c_strlen): Unconditionally return zero for the empty
12143 Use -Warray-bounds for warnings.
12144 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12145 for non-constant array indices with COMPONENT_REF, arrays of
12146 arrays, and pointers to arrays.
12147 (gimple_fold_builtin_strlen): Determine and set length range for
12148 non-constant character arrays.
12150 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12152 PR middle-end/81897
12153 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12156 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12158 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12160 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12163 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12164 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12165 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12166 indexed_or_indirect_operand predicate.
12167 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12168 (*vsx_le_perm_load_v8hi): Likewise.
12169 (*vsx_le_perm_load_v16qi): Likewise.
12170 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12171 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12172 (*vsx_le_perm_store_v8hi): Likewise.
12173 (*vsx_le_perm_store_v16qi): Likewise.
12174 (eight unnamed splitters): Likewise.
12176 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12178 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12179 * config/rs6000/emmintrin.h: Likewise.
12180 * config/rs6000/mmintrin.h: Likewise.
12181 * config/rs6000/xmmintrin.h: Likewise.
12183 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12186 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12188 * tree.c (tree_nop_conversion): Return true for location wrapper
12190 (maybe_wrap_with_location): New function.
12191 (selftest::check_strip_nops): New function.
12192 (selftest::test_location_wrappers): New function.
12193 (selftest::tree_c_tests): Call it.
12194 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12195 (maybe_wrap_with_location): New decl.
12196 (EXPR_LOCATION_WRAPPER_P): New macro.
12197 (location_wrapper_p): New inline function.
12198 (tree_strip_any_location_wrapper): New inline function.
12200 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12203 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12204 stack_realign_offset for the largest alignment of stack slot
12206 (ix86_find_max_used_stack_alignment): New function.
12207 (ix86_finalize_stack_frame_flags): Use it. Set
12208 max_used_stack_alignment if we don't realign stack.
12209 * config/i386/i386.h (machine_function): Add
12210 max_used_stack_alignment.
12212 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12214 * config/arm/arm.opt (-mbranch-cost): New option.
12215 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12218 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12221 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12222 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12224 2018-01-10 Richard Biener <rguenther@suse.de>
12227 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12228 early out so it also covers the case where we have a non-NULL
12231 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12233 PR tree-optimization/83753
12234 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12235 for non-strided grouped accesses if the number of elements is 1.
12237 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12240 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12241 * i386.h (TARGET_USE_GATHER): Define.
12242 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12244 2018-01-10 Martin Liska <mliska@suse.cz>
12247 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12248 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12250 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12251 CLEANUP_NO_PARTITIONING is not set.
12253 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12255 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12256 for vectors, as a partial revert of r254296.
12257 * rtl.h (const_vec_p): Delete.
12258 (const_vec_duplicate_p): Don't test for vector CONSTs.
12259 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12260 * expmed.c (make_tree): Likewise.
12263 * common.md (E, F): Use CONSTANT_P instead of checking for
12265 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12266 checking for CONST_VECTOR.
12268 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12270 PR middle-end/83575
12271 * predict.c (force_edge_cold): Handle in more sane way edges
12272 with no prediction.
12274 2018-01-09 Carl Love <cel@us.ibm.com>
12276 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12278 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12279 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12280 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12281 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12282 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12283 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12284 * config/rs6000/rs6000-protos.h: Add extern defition for
12285 rs6000_generate_float2_double_code.
12286 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12288 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12289 (float2_v2df): Add define_expand.
12291 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12294 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12295 op_mode in the force_to_mode call.
12297 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12299 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12300 instead of checking each element individually.
12301 (aarch64_evpc_uzp): Likewise.
12302 (aarch64_evpc_zip): Likewise.
12303 (aarch64_evpc_ext): Likewise.
12304 (aarch64_evpc_rev): Likewise.
12305 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12306 instead of checking each element individually. Return true without
12308 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12309 whether all selected elements come from the same input, instead of
12310 checking each element individually. Remove calls to gen_rtx_REG,
12311 start_sequence and end_sequence and instead assert that no rtl is
12314 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12316 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12317 order of HIGH and CONST checks.
12319 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12321 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12322 if the destination isn't an SSA_NAME.
12324 2018-01-09 Richard Biener <rguenther@suse.de>
12326 PR tree-optimization/83668
12327 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12329 (canonicalize_loop_form): ... here, renamed from ...
12330 (canonicalize_loop_closed_ssa_form): ... this and amended to
12331 swap successor edges for loop exit blocks to make us use
12332 the RPO order we need for initial schedule generation.
12334 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12336 PR tree-optimization/64811
12337 * match.pd: When optimizing comparisons with Inf, avoid
12338 introducing or losing exceptions from comparisons with NaN.
12340 2018-01-09 Martin Liska <mliska@suse.cz>
12343 * asan.c (shadow_mem_size): Add gcc_assert.
12345 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12347 Don't save registers in main().
12350 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12351 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12352 * config/avr/avr.c (avr_set_current_function): Don't error if
12353 naked, OS_task or OS_main are specified at the same time.
12354 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12356 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12358 * common/config/avr/avr-common.c (avr_option_optimization_table):
12359 Switch on -mmain-is-OS_task for optimizing compilations.
12361 2018-01-09 Richard Biener <rguenther@suse.de>
12363 PR tree-optimization/83572
12364 * graphite.c: Include cfganal.h.
12365 (graphite_transform_loops): Connect infinite loops to exit
12366 and remove fake edges at the end.
12368 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12370 * ipa-inline.c (edge_badness): Revert accidental checkin.
12372 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12375 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12376 symbols; not inline clones.
12378 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12381 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12382 hard registers. Formatting fixes.
12384 PR preprocessor/83722
12385 * gcc.c (try_generate_repro): Pass
12386 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12387 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12390 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12391 Kito Cheng <kito.cheng@gmail.com>
12393 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12394 (riscv_leaf_function_p): Delete.
12395 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12397 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12399 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12401 (do_ifelse): New function.
12402 (do_isel): New function.
12403 (do_sub3): New function.
12404 (do_add3): New function.
12405 (do_load_mask_compare): New function.
12406 (do_overlap_load_compare): New function.
12407 (expand_compare_loop): New function.
12408 (expand_block_compare): Call expand_compare_loop() when appropriate.
12409 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12410 option description.
12411 (-mblock-compare-inline-loop-limit): New option.
12413 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12416 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12417 Reverse order of second and third operands in first alternative.
12418 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12419 of first and second elements in UNSPEC_VPERMR vector.
12420 (altivec_expand_vec_perm_le): Likewise.
12422 2018-01-08 Jeff Law <law@redhat.com>
12424 PR rtl-optimizatin/81308
12425 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12426 (process_switch): If group_case_labels makes a change, then set
12428 (pass_convert_switch::execute): If a switch is converted, then
12429 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12431 PR rtl-optimization/81308
12432 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12435 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12437 PR target/83663 - Revert r255946
12438 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12439 generation for cases where splatting a value is not useful.
12440 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12441 across a vec_duplicate and a paradoxical subreg forming a vector
12442 mode to a vec_concat.
12444 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12446 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12447 -march=armv8.3-a variants.
12448 * config/arm/t-multilib: Likewise.
12449 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12451 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12453 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12455 (cceq_ior_compare_complement): Give it a name so I can use it, and
12456 change boolean_or_operator predicate to boolean_operator so it can
12457 be used to generate a crand.
12458 (eqne): New code iterator.
12459 (bd/bd_neg): New code_attrs.
12460 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12461 a single define_insn.
12462 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12463 decrement (bdnzt/bdnzf/bdzt/bdzf).
12464 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12465 with the new names of the branch decrement patterns, and added the
12466 names of the branch decrement conditional patterns.
12468 2018-01-08 Richard Biener <rguenther@suse.de>
12470 PR tree-optimization/83563
12471 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12474 2018-01-08 Richard Biener <rguenther@suse.de>
12476 PR middle-end/83713
12477 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12479 2018-01-08 Richard Biener <rguenther@suse.de>
12481 PR tree-optimization/83685
12482 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12483 references to abnormals.
12485 2018-01-08 Richard Biener <rguenther@suse.de>
12488 * dwarf2out.c (output_indirect_strings): Handle empty
12489 skeleton_debug_str_hash.
12490 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12492 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12494 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12495 (emit_store_direct): Likewise.
12496 (arc_trampoline_adjust_address): Likewise.
12497 (arc_asm_trampoline_template): New function.
12498 (arc_initialize_trampoline): Use asm_trampoline_template.
12499 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12500 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12501 * config/arc/arc.md (flush_icache): Delete pattern.
12503 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12505 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12506 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12509 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12512 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12513 by not USED_FOR_TARGET.
12514 (make_pass_resolve_sw_modes): Likewise.
12516 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12518 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12521 2018-01-08 Richard Biener <rguenther@suse.de>
12523 PR middle-end/83580
12524 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12526 2018-01-08 Richard Biener <rguenther@suse.de>
12528 PR middle-end/83517
12529 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12531 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12533 PR middle-end/81897
12534 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12535 basic blocks with a small number of successors.
12536 (convert_control_dep_chain_into_preds): Improve handling of
12538 (dump_predicates): Split apart into...
12539 (dump_pred_chain): ...here...
12540 (dump_pred_info): ...and here.
12541 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12542 (can_chain_union_be_invalidated_p): Improve check for invalidation
12544 (uninit_uses_cannot_happen): Avoid unnecessary if
12545 convert_control_dep_chain_into_preds yielded nothing.
12547 2018-01-06 Martin Sebor <msebor@redhat.com>
12549 PR tree-optimization/83640
12550 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12551 subtracting negative offset from size.
12552 (builtin_access::overlap): Adjust offset bounds of the access to fall
12553 within the size of the object if possible.
12555 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12557 PR rtl-optimization/83699
12558 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12559 extract_bit_field_as_subreg to cases in which the extracted
12560 value is also a vector.
12562 * lra-constraints.c (process_alt_operands): Test for the equivalence
12563 substitutions when detecting a possible reload cycle.
12565 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12568 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12569 by default if flag_selective_schedling{,2}. Formatting fixes.
12571 PR rtl-optimization/83682
12572 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12573 if it has non-VECTOR_MODE element mode.
12574 (vec_duplicate_p): Likewise.
12576 PR middle-end/83694
12577 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12578 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12580 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12583 * config/i386/i386-builtin.def
12584 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12585 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12586 Require also OPTION_MASK_ISA_AVX512F in addition to
12587 OPTION_MASK_ISA_GFNI.
12588 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12589 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12590 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12591 to OPTION_MASK_ISA_GFNI.
12592 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12593 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12594 OPTION_MASK_ISA_AVX512BW.
12595 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12596 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12597 addition to OPTION_MASK_ISA_GFNI.
12598 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12599 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12600 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12601 to OPTION_MASK_ISA_GFNI.
12602 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12603 a requirement for all ISAs rather than any of them with a few
12605 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12607 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12608 bitmasks to be enabled with 3 exceptions, instead of requiring any
12609 enabled ISA with lots of exceptions.
12610 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12611 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12612 Change avx512bw in isa attribute to avx512f.
12613 * config/i386/sgxintrin.h: Add license boilerplate.
12614 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12615 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12616 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12617 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12619 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12620 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12621 temporarily sse2 rather than sse if not enabled already.
12624 * config/i386/sse.md (VI248_VLBW): Rename to ...
12625 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12626 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12627 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12628 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12629 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12630 mode iterator instead of VI248_VLBW.
12632 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12634 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12635 (record_modified): Skip clobbers; add debug output.
12636 (param_change_prob): Use sreal frequencies.
12638 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12640 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12641 punt for user-aligned variables.
12643 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12645 * tree-chrec.c (chrec_contains_symbols): Return true for
12648 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12651 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12652 of (x|y) == x for BICS pattern.
12654 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12656 PR tree-optimization/83605
12657 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12658 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12661 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12663 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12664 * config/epiphany/rtems.h: New file.
12666 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12667 Uros Bizjak <ubizjak@gmail.com>
12670 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12671 QIreg_operand instead of register_operand predicate.
12672 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12673 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12674 comments instead of -fmitigate[-_]rop.
12676 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12679 * cgraphunit.c (symbol_table::compile): Switch to text_section
12680 before calling assembly_start debug hook.
12681 * run-rtl-passes.c (run_rtl_passes): Likewise.
12684 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12686 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12687 range_int_cst_p rather than !symbolic_range_p before calling
12688 extract_range_from_multiplicative_op_1.
12690 2018-01-04 Jeff Law <law@redhat.com>
12692 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12693 redundant test in assertion.
12695 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12697 * doc/rtl.texi: Document machine_mode wrapper classes.
12699 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12701 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12702 using tree_to_uhwi.
12704 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12706 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12707 the VEC_PERM_EXPR fold to fail.
12709 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12712 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12713 to switched_sections.
12715 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12718 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12719 test for d.testing.
12721 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12724 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12725 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12727 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12730 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12731 is BLKmode and bitpos not zero or mode change is needed.
12733 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12736 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12739 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12742 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12743 instead of MULT rtx. Update all corresponding splitters.
12744 (*saddl_se): Ditto.
12745 (*ssub<modesuffix>): Ditto.
12746 (*ssubl_se): Ditto.
12747 (*cmp_sadd_di): Update split patterns.
12748 (*cmp_sadd_si): Ditto.
12749 (*cmp_sadd_sidi): Ditto.
12750 (*cmp_ssub_di): Ditto.
12751 (*cmp_ssub_si): Ditto.
12752 (*cmp_ssub_sidi): Ditto.
12753 * config/alpha/predicates.md (const23_operand): New predicate.
12754 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12755 Look for ASHIFT, not MULT inner operand.
12756 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12758 2018-01-04 Martin Liska <mliska@suse.cz>
12760 PR gcov-profile/83669
12761 * gcov.c (output_intermediate_file): Add version to intermediate
12763 * doc/gcov.texi: Document new field 'version' in intermediate
12764 file format. Fix location of '-k' option of gcov command.
12766 2018-01-04 Martin Liska <mliska@suse.cz>
12769 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12771 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12773 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12775 2018-01-03 Martin Sebor <msebor@redhat.com>
12777 PR tree-optimization/83655
12778 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12779 checking calls with invalid arguments.
12781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12783 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12784 (vectorizable_mask_load_store): Delete.
12785 (vectorizable_call): Return false for masked loads and stores.
12786 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12787 instead of gimple_assign_rhs1.
12788 (vectorizable_load): Handle IFN_MASK_LOAD.
12789 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12793 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12795 (vectorizable_mask_load_store): ...here.
12796 (vectorizable_load): ...and here.
12798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12800 * tree-vect-stmts.c (vect_build_all_ones_mask)
12801 (vect_build_zero_merge_argument): New functions, split out from...
12802 (vectorizable_load): ...here.
12804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12806 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12808 (vectorizable_mask_load_store): ...here.
12809 (vectorizable_store): ...and here.
12811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12813 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12815 (vectorizable_mask_load_store): ...here.
12817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12819 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12820 (vect_model_store_cost): Take a vec_load_store_type instead of a
12822 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12823 (vect_model_store_cost): Take a vec_load_store_type instead of a
12825 (vectorizable_mask_load_store): Update accordingly.
12826 (vectorizable_store): Likewise.
12827 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12831 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12832 IFN_MASK_LOAD calls here rather than...
12833 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12836 Alan Hayward <alan.hayward@arm.com>
12837 David Sherwood <david.sherwood@arm.com>
12839 * expmed.c (extract_bit_field_1): For vector extracts,
12840 fall back to extract_bit_field_as_subreg if vec_extract
12843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12844 Alan Hayward <alan.hayward@arm.com>
12845 David Sherwood <david.sherwood@arm.com>
12847 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12848 they are variable or constant sized.
12849 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12850 slots for constant-sized data.
12852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12853 Alan Hayward <alan.hayward@arm.com>
12854 David Sherwood <david.sherwood@arm.com>
12856 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12857 handling COND_EXPRs with boolean comparisons, try to find a better
12858 basis for the mask type than the boolean itself.
12860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12862 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12863 is calculated and how it can be overridden.
12864 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12865 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12867 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
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 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12875 Remove the mode argument.
12876 (aarch64_simd_valid_immediate): Remove the mode and inverse
12878 * config/aarch64/iterators.md (bitsize): New iterator.
12879 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12880 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12881 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12882 aarch64_simd_valid_immediate.
12883 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12884 (aarch64_reg_or_bic_imm): Likewise.
12885 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12886 with an insn_type enum and msl with a modifier_type enum.
12887 Replace element_width with a scalar_mode. Change the shift
12888 to unsigned int. Add constructors for scalar_float_mode and
12889 scalar_int_mode elements.
12890 (aarch64_vect_float_const_representable_p): Delete.
12891 (aarch64_can_const_movi_rtx_p)
12892 (aarch64_simd_scalar_immediate_valid_for_move)
12893 (aarch64_simd_make_constant): Update call to
12894 aarch64_simd_valid_immediate.
12895 (aarch64_advsimd_valid_immediate_hs): New function.
12896 (aarch64_advsimd_valid_immediate): Likewise.
12897 (aarch64_simd_valid_immediate): Remove mode and inverse
12898 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12899 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12900 and aarch64_float_const_representable_p on the result.
12901 (aarch64_output_simd_mov_immediate): Remove mode argument.
12902 Update call to aarch64_simd_valid_immediate and use of
12903 simd_immediate_info.
12904 (aarch64_output_scalar_simd_mov_immediate): Update call
12907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12908 Alan Hayward <alan.hayward@arm.com>
12909 David Sherwood <david.sherwood@arm.com>
12911 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12912 (mode_nunits): Likewise CONST_MODE_NUNITS.
12913 * machmode.def (ADJUST_NUNITS): Document.
12914 * genmodes.c (mode_data::need_nunits_adj): New field.
12915 (blank_mode): Update accordingly.
12916 (adj_nunits): New variable.
12917 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12919 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12920 listed in adj_nunits.
12921 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12922 listed in adj_nunits. Don't emit case statements for such modes.
12923 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12924 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12925 nothing if adj_nunits is nonnull.
12926 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12927 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12928 (emit_mode_fbit): Update use of print_maybe_const_decl.
12929 (emit_move_size): Likewise. Treat the array as non-const
12931 (emit_mode_adjustments): Handle adj_nunits.
12933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12935 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12936 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12937 (VECTOR_MODES): Use it.
12938 (make_vector_modes): Take the prefix as an argument.
12940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12941 Alan Hayward <alan.hayward@arm.com>
12942 David Sherwood <david.sherwood@arm.com>
12944 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12945 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12946 for MODE_VECTOR_BOOL.
12947 * machmode.def (VECTOR_BOOL_MODE): Document.
12948 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12949 (make_vector_bool_mode): New function.
12950 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12952 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12953 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12955 * stor-layout.c (int_mode_for_mode): Likewise.
12956 * tree.c (build_vector_type_for_mode): Likewise.
12957 * varasm.c (output_constant_pool_2): Likewise.
12958 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12959 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12960 for MODE_VECTOR_BOOL.
12961 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12962 of mode class checks.
12963 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12964 instead of a list of mode class checks.
12965 (expand_vector_scalar_condition): Likewise.
12966 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12969 Alan Hayward <alan.hayward@arm.com>
12970 David Sherwood <david.sherwood@arm.com>
12972 * machmode.h (mode_size): Change from unsigned short to
12974 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12975 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12976 or if measurement_type is not polynomial.
12977 (fixed_size_mode::includes_p): Check for constant-sized modes.
12978 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12979 return a poly_uint16 rather than an unsigned short.
12980 (emit_mode_size): Change the type of mode_size from unsigned short
12981 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12982 (emit_mode_adjustments): Cope with polynomial vector sizes.
12983 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12985 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12987 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12988 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12989 * caller-save.c (setup_save_areas): Likewise.
12990 (replace_reg_with_saved_mem): Likewise.
12991 * calls.c (emit_library_call_value_1): Likewise.
12992 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12993 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12994 (gen_lowpart_for_combine): Likewise.
12995 * convert.c (convert_to_integer_1): Likewise.
12996 * cse.c (equiv_constant, cse_insn): Likewise.
12997 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12998 (cselib_subst_to_values): Likewise.
12999 * dce.c (word_dce_process_block): Likewise.
13000 * df-problems.c (df_word_lr_mark_ref): Likewise.
13001 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13002 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13003 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13004 (rtl_for_decl_location): Likewise.
13005 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13006 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13007 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13008 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13009 (expand_expr_real_1): Likewise.
13010 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13011 (pad_below): Likewise.
13012 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13013 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13014 * ira.c (get_subreg_tracking_sizes): Likewise.
13015 * ira-build.c (ira_create_allocno_objects): Likewise.
13016 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13017 (ira_sort_regnos_for_alter_reg): Likewise.
13018 * ira-costs.c (record_operand_costs): Likewise.
13019 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13020 (resolve_simple_move): Likewise.
13021 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13022 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13023 (lra_constraints): Likewise.
13024 (CONST_POOL_OK_P): Reject variable-sized modes.
13025 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13026 (add_pseudo_to_slot, lra_spill): Likewise.
13027 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13028 * optabs-query.c (get_best_extraction_insn): Likewise.
13029 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13030 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13031 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13032 * recog.c (offsettable_address_addr_space_p): Likewise.
13033 * regcprop.c (maybe_mode_change): Likewise.
13034 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13035 * regrename.c (build_def_use): Likewise.
13036 * regstat.c (dump_reg_info): Likewise.
13037 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13038 (find_reloads, find_reloads_subreg_address): Likewise.
13039 * reload1.c (eliminate_regs_1): Likewise.
13040 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13041 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13042 (simplify_binary_operation_1, simplify_subreg): Likewise.
13043 * targhooks.c (default_function_arg_padding): Likewise.
13044 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13045 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13046 (verify_gimple_assign_ternary): Likewise.
13047 * tree-inline.c (estimate_move_cost): Likewise.
13048 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13049 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13050 (get_address_cost_ainc): Likewise.
13051 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13052 (vect_supportable_dr_alignment): Likewise.
13053 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13054 (vectorizable_reduction): Likewise.
13055 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13056 (vectorizable_operation, vectorizable_load): Likewise.
13057 * tree.c (build_same_sized_truth_vector_type): Likewise.
13058 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13059 * var-tracking.c (emit_note_insn_var_location): Likewise.
13060 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13061 (ADDR_VEC_ALIGN): Likewise.
13063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13064 Alan Hayward <alan.hayward@arm.com>
13065 David Sherwood <david.sherwood@arm.com>
13067 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13069 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13070 or if measurement_type is polynomial.
13071 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13072 * combine.c (make_extraction): Likewise.
13073 * dse.c (find_shift_sequence): Likewise.
13074 * dwarf2out.c (mem_loc_descriptor): Likewise.
13075 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13076 (extract_bit_field, extract_low_bits): Likewise.
13077 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13078 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13079 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13080 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13081 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13082 * reload.c (find_reloads): Likewise.
13083 * reload1.c (alter_reg): Likewise.
13084 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13085 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13086 * tree-if-conv.c (predicate_mem_writes): Likewise.
13087 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13088 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13089 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13090 * valtrack.c (dead_debug_insert_temp): Likewise.
13091 * varasm.c (mergeable_constant_section): Likewise.
13092 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13095 Alan Hayward <alan.hayward@arm.com>
13096 David Sherwood <david.sherwood@arm.com>
13098 * expr.c (expand_assignment): Cope with polynomial mode sizes
13099 when assigning to a CONCAT.
13101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13102 Alan Hayward <alan.hayward@arm.com>
13103 David Sherwood <david.sherwood@arm.com>
13105 * machmode.h (mode_precision): Change from unsigned short to
13107 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13109 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13110 or if measurement_type is not polynomial.
13111 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13112 in which the mode is already known to be a scalar_int_mode.
13113 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13114 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13116 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13117 for GET_MODE_PRECISION.
13118 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13119 for GET_MODE_PRECISION.
13120 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13122 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13123 (expand_field_assignment, make_extraction): Likewise.
13124 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13125 (get_last_value): Likewise.
13126 * convert.c (convert_to_integer_1): Likewise.
13127 * cse.c (cse_insn): Likewise.
13128 * expr.c (expand_expr_real_1): Likewise.
13129 * lra-constraints.c (simplify_operand_subreg): Likewise.
13130 * optabs-query.c (can_atomic_load_p): Likewise.
13131 * optabs.c (expand_atomic_load): Likewise.
13132 (expand_atomic_store): Likewise.
13133 * ree.c (combine_reaching_defs): Likewise.
13134 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13135 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13136 * tree.h (type_has_mode_precision_p): Likewise.
13137 * ubsan.c (instrument_si_overflow): Likewise.
13139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13140 Alan Hayward <alan.hayward@arm.com>
13141 David Sherwood <david.sherwood@arm.com>
13143 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13144 polynomial numbers of units.
13145 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13146 (valid_vector_subparts_p): New function.
13147 (build_vector_type): Remove temporary shim and take the number
13148 of units as a poly_uint64 rather than an int.
13149 (build_opaque_vector_type): Take the number of units as a
13150 poly_uint64 rather than an int.
13151 * tree.c (build_vector_from_ctor): Handle polynomial
13152 TYPE_VECTOR_SUBPARTS.
13153 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13154 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13155 (build_vector_from_val): If the number of units is variable,
13156 use build_vec_duplicate_cst for constant operands and
13157 VEC_DUPLICATE_EXPR otherwise.
13158 (make_vector_type): Remove temporary is_constant ().
13159 (build_vector_type, build_opaque_vector_type): Take the number of
13160 units as a poly_uint64 rather than an int.
13161 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13163 * cfgexpand.c (expand_debug_expr): Likewise.
13164 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13165 (store_constructor, expand_expr_real_1): Likewise.
13166 (const_scalar_mask_from_tree): Likewise.
13167 * fold-const-call.c (fold_const_reduction): Likewise.
13168 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13169 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13170 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13171 (fold_relational_const): Likewise.
13172 (native_interpret_vector): Likewise. Change the size from an
13173 int to an unsigned int.
13174 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13175 TYPE_VECTOR_SUBPARTS.
13176 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13177 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13178 duplicating a non-constant operand into a variable-length vector.
13179 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13180 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13181 * ipa-icf.c (sem_variable::equals): Likewise.
13182 * match.pd: Likewise.
13183 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13184 * print-tree.c (print_node): Likewise.
13185 * stor-layout.c (layout_type): Likewise.
13186 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13187 * tree-cfg.c (verify_gimple_comparison): Likewise.
13188 (verify_gimple_assign_binary): Likewise.
13189 (verify_gimple_assign_ternary): Likewise.
13190 (verify_gimple_assign_single): Likewise.
13191 * tree-pretty-print.c (dump_generic_node): Likewise.
13192 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13193 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13194 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13195 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13196 (vect_shift_permute_load_chain): Likewise.
13197 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13198 (expand_vector_condition, optimize_vector_constructor): Likewise.
13199 (lower_vec_perm, get_compute_type): Likewise.
13200 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13201 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13202 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13203 (vect_recog_mask_conversion_pattern): Likewise.
13204 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13205 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13206 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13207 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13208 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13209 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13210 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13211 (supportable_widening_operation): Likewise.
13212 (supportable_narrowing_operation): Likewise.
13213 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13215 * varasm.c (output_constant): Likewise.
13217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13218 Alan Hayward <alan.hayward@arm.com>
13219 David Sherwood <david.sherwood@arm.com>
13221 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13222 so that both the length == 3 and length != 3 cases set up their
13223 own permute vectors. Add comments explaining why we know the
13224 number of elements is constant.
13225 (vect_permute_load_chain): Likewise.
13227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13228 Alan Hayward <alan.hayward@arm.com>
13229 David Sherwood <david.sherwood@arm.com>
13231 * machmode.h (mode_nunits): Change from unsigned char to
13233 (ONLY_FIXED_SIZE_MODES): New macro.
13234 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13235 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13236 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13238 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13239 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13240 or if measurement_type is not polynomial.
13241 * genmodes.c (ZERO_COEFFS): New macro.
13242 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13244 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13245 Use ZERO_COEFFS when emitting initializers.
13246 * data-streamer.h (bp_pack_poly_value): New function.
13247 (bp_unpack_poly_value): Likewise.
13248 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13249 for GET_MODE_NUNITS.
13250 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13251 for GET_MODE_NUNITS.
13252 * tree.c (make_vector_type): Remove temporary shim and make
13253 the real function take the number of units as a poly_uint64
13254 rather than an int.
13255 (build_vector_type_for_mode): Handle polynomial nunits.
13256 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13257 * emit-rtl.c (const_vec_series_p_1): Likewise.
13258 (gen_rtx_CONST_VECTOR): Likewise.
13259 * fold-const.c (test_vec_duplicate_folding): Likewise.
13260 * genrecog.c (validate_pattern): Likewise.
13261 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13262 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13263 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13264 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13265 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13266 * rtlanal.c (subreg_get_info): Likewise.
13267 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13268 (vect_grouped_load_supported): Likewise.
13269 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13270 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13271 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13272 (simplify_const_unary_operation, simplify_binary_operation_1)
13273 (simplify_const_binary_operation, simplify_ternary_operation)
13274 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13275 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13276 instead of CONST_VECTOR_NUNITS.
13277 * varasm.c (output_constant_pool_2): Likewise.
13278 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13279 explicit-encoded elements in the XVEC for variable-length vectors.
13281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13283 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13286 Alan Hayward <alan.hayward@arm.com>
13287 David Sherwood <david.sherwood@arm.com>
13289 * coretypes.h (fixed_size_mode): Declare.
13290 (fixed_size_mode_pod): New typedef.
13291 * builtins.h (target_builtins::x_apply_args_mode)
13292 (target_builtins::x_apply_result_mode): Change type to
13293 fixed_size_mode_pod.
13294 * builtins.c (apply_args_size, apply_result_size, result_vector)
13295 (expand_builtin_apply_args_1, expand_builtin_apply)
13296 (expand_builtin_return): Update accordingly.
13298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13300 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13301 * cselib.c (cselib_hash_rtx): Likewise.
13302 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13303 CONST_VECTOR encoding.
13305 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13306 Jeff Law <law@redhat.com>
13309 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13310 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13311 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13312 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13315 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13316 explicitly probe *sp in a noreturn function if there were any callee
13317 register saves or frame pointer is needed.
13319 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13322 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13323 BLKmode for ternary, binary or unary expressions.
13326 * var-tracking.c (delete_vta_debug_insn): New inline function.
13327 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13328 insns from get_insns () to NULL instead of each bb separately.
13329 Use delete_vta_debug_insn. No longer static.
13330 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13331 delete_vta_debug_insns callers.
13332 * rtl.h (delete_vta_debug_insns): Declare.
13333 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13334 instead of variable_tracking_main.
13336 2018-01-03 Martin Sebor <msebor@redhat.com>
13338 PR tree-optimization/83603
13339 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13340 arguments past the endof the argument list in functions declared
13341 without a prototype.
13342 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13343 Avoid checking when arguments are null.
13345 2018-01-03 Martin Sebor <msebor@redhat.com>
13348 * doc/extend.texi (attribute const): Fix a typo.
13349 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13350 issuing -Wsuggest-attribute for void functions.
13352 2018-01-03 Martin Sebor <msebor@redhat.com>
13354 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13355 offset_int::from instead of wide_int::to_shwi.
13356 (maybe_diag_overlap): Remove assertion.
13357 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13358 * gimple-ssa-sprintf.c (format_directive): Same.
13359 (parse_directive): Same.
13360 (sprintf_dom_walker::compute_format_length): Same.
13361 (try_substitute_return_value): Same.
13363 2018-01-03 Jeff Law <law@redhat.com>
13365 PR middle-end/83654
13366 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13367 non-constant residual for zero at runtime and avoid probing in
13368 that case. Reorganize code for trailing problem to mirror handling
13371 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13373 PR tree-optimization/83501
13374 * tree-ssa-strlen.c (get_string_cst): New.
13375 (handle_char_store): Call get_string_cst.
13377 2018-01-03 Martin Liska <mliska@suse.cz>
13379 PR tree-optimization/83593
13380 * tree-ssa-strlen.c: Include tree-cfg.h.
13381 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13382 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13383 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13385 (strlen_dom_walker::before_dom_children): Call
13386 gimple_purge_dead_eh_edges. Dump tranformation with details
13388 (strlen_dom_walker::before_dom_children): Update call by adding
13389 new argument cleanup_eh.
13390 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13392 2018-01-03 Martin Liska <mliska@suse.cz>
13395 * cif-code.def (VARIADIC_THUNK): New enum value.
13396 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13399 2018-01-03 Jan Beulich <jbeulich@suse.com>
13401 * sse.md (mov<mode>_internal): Tighten condition for when to use
13402 vmovdqu<ssescalarsize> for TI and OI modes.
13404 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13406 Update copyright years.
13408 2018-01-03 Martin Liska <mliska@suse.cz>
13411 * ipa-visibility.c (function_and_variable_visibility): Skip
13412 functions with noipa attribure.
13414 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13416 * gcc.c (process_command): Update copyright notice dates.
13417 * gcov-dump.c (print_version): Ditto.
13418 * gcov.c (print_version): Ditto.
13419 * gcov-tool.c (print_version): Ditto.
13420 * gengtype.c (create_file): Ditto.
13421 * doc/cpp.texi: Bump @copying's copyright year.
13422 * doc/cppinternals.texi: Ditto.
13423 * doc/gcc.texi: Ditto.
13424 * doc/gccint.texi: Ditto.
13425 * doc/gcov.texi: Ditto.
13426 * doc/install.texi: Ditto.
13427 * doc/invoke.texi: Ditto.
13429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13431 * vector-builder.h (vector_builder::m_full_nelts): Change from
13432 unsigned int to poly_uint64.
13433 (vector_builder::full_nelts): Update prototype accordingly.
13434 (vector_builder::new_vector): Likewise.
13435 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13436 (vector_builder::operator ==): Likewise.
13437 (vector_builder::finalize): Likewise.
13438 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13439 Take the number of elements as a poly_uint64 rather than an
13441 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13442 from unsigned int to poly_uint64.
13443 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13444 (vec_perm_indices::new_vector): Likewise.
13445 (vec_perm_indices::length): Likewise.
13446 (vec_perm_indices::nelts_per_input): Likewise.
13447 (vec_perm_indices::input_nelts): Likewise.
13448 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13449 number of elements per input as a poly_uint64 rather than an
13450 unsigned int. Use the original encoding for variable-length
13451 vectors, rather than clamping each individual element.
13452 For the second and subsequent elements in each pattern,
13453 clamp the step and base before clamping their sum.
13454 (vec_perm_indices::series_p): Handle polynomial element counts.
13455 (vec_perm_indices::all_in_range_p): Likewise.
13456 (vec_perm_indices_to_tree): Likewise.
13457 (vec_perm_indices_to_rtx): Likewise.
13458 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13459 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13460 (tree_vector_builder::new_binary_operation): Handle polynomial
13461 element counts. Return false if we need to know the number
13462 of elements at compile time.
13463 * fold-const.c (fold_vec_perm): Punt if the number of elements
13464 isn't known at compile time.
13466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13468 * vec-perm-indices.h (vec_perm_builder): Change element type
13469 from HOST_WIDE_INT to poly_int64.
13470 (vec_perm_indices::element_type): Update accordingly.
13471 (vec_perm_indices::clamp): Handle polynomial element_types.
13472 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13473 (vec_perm_indices::all_in_range_p): Likewise.
13474 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13476 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13477 polynomial vec_perm_indices element types.
13478 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13479 * fold-const.c (fold_vec_perm): Likewise.
13480 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13481 * tree-vect-generic.c (lower_vec_perm): Likewise.
13482 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13483 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13484 element type to HOST_WIDE_INT.
13486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13487 Alan Hayward <alan.hayward@arm.com>
13488 David Sherwood <david.sherwood@arm.com>
13490 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13491 rather than an int. Use plus_constant.
13492 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13493 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13496 Alan Hayward <alan.hayward@arm.com>
13497 David Sherwood <david.sherwood@arm.com>
13499 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13500 a HOST_WIDE_INT to a poly_int64.
13502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13503 Alan Hayward <alan.hayward@arm.com>
13504 David Sherwood <david.sherwood@arm.com>
13506 * calls.c (load_register_parameters): Cope with polynomial
13507 mode sizes. Require a constant size for BLKmode parameters
13508 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13509 forces a parameter to be padded at the lsb end in order to
13510 fill a complete number of words, require the parameter size
13511 to be ordered wrt UNITS_PER_WORD.
13513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13514 Alan Hayward <alan.hayward@arm.com>
13515 David Sherwood <david.sherwood@arm.com>
13517 * reload1.c (spill_stack_slot_width): Change element type
13518 from unsigned int to poly_uint64_pod.
13519 (alter_reg): Treat mode sizes as polynomial.
13521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13522 Alan Hayward <alan.hayward@arm.com>
13523 David Sherwood <david.sherwood@arm.com>
13525 * reload.c (complex_word_subreg_p): New function.
13526 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
13533 operands whose sizes aren't ordered.
13534 (match_reload): Refer to this check here.
13536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13537 Alan Hayward <alan.hayward@arm.com>
13538 David Sherwood <david.sherwood@arm.com>
13540 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13541 that the mode size is in the set {1, 2, 4, 8, 16}.
13543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13544 Alan Hayward <alan.hayward@arm.com>
13545 David Sherwood <david.sherwood@arm.com>
13547 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13548 Use plus_constant instead of gen_rtx_PLUS.
13550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13551 Alan Hayward <alan.hayward@arm.com>
13552 David Sherwood <david.sherwood@arm.com>
13554 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13555 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13556 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13557 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13558 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13559 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13560 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13561 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13562 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13563 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13565 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13566 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13567 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13568 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13569 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13570 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13571 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13572 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13573 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13574 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13576 * expr.c (emit_move_resolve_push): Treat the input and result
13577 of PUSH_ROUNDING as a poly_int64.
13578 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13579 (emit_push_insn): Likewise.
13580 * lra-eliminations.c (mark_not_eliminable): Likewise.
13581 * recog.c (push_operand): Likewise.
13582 * reload1.c (elimination_effects): Likewise.
13583 * rtlanal.c (nonzero_bits1): Likewise.
13584 * calls.c (store_one_arg): Likewise. Require the padding to be
13585 known at compile time.
13587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13588 Alan Hayward <alan.hayward@arm.com>
13589 David Sherwood <david.sherwood@arm.com>
13591 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13592 Use plus_constant instead of gen_rtx_PLUS.
13594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13595 Alan Hayward <alan.hayward@arm.com>
13596 David Sherwood <david.sherwood@arm.com>
13598 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13599 rather than an int.
13601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13602 Alan Hayward <alan.hayward@arm.com>
13603 David Sherwood <david.sherwood@arm.com>
13605 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13606 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13607 via stack temporaries. Treat the mode size as polynomial too.
13609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13610 Alan Hayward <alan.hayward@arm.com>
13611 David Sherwood <david.sherwood@arm.com>
13613 * expr.c (expand_expr_real_2): When handling conversions involving
13614 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13615 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13616 as a poly_uint64 too.
13618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13619 Alan Hayward <alan.hayward@arm.com>
13620 David Sherwood <david.sherwood@arm.com>
13622 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13625 Alan Hayward <alan.hayward@arm.com>
13626 David Sherwood <david.sherwood@arm.com>
13628 * combine.c (can_change_dest_mode): Handle polynomial
13629 REGMODE_NATURAL_SIZE.
13630 * expmed.c (store_bit_field_1): Likewise.
13631 * expr.c (store_constructor): Likewise.
13632 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13633 and polynomial REGMODE_NATURAL_SIZE.
13634 (gen_lowpart_common): Likewise.
13635 * reginfo.c (record_subregs_of_mode): Likewise.
13636 * rtlanal.c (read_modify_subreg_p): Likewise.
13638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13639 Alan Hayward <alan.hayward@arm.com>
13640 David Sherwood <david.sherwood@arm.com>
13642 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13643 numbers of elements.
13645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13646 Alan Hayward <alan.hayward@arm.com>
13647 David Sherwood <david.sherwood@arm.com>
13649 * match.pd: Cope with polynomial numbers of vector elements.
13651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13652 Alan Hayward <alan.hayward@arm.com>
13653 David Sherwood <david.sherwood@arm.com>
13655 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13656 in a POINTER_PLUS_EXPR.
13658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13659 Alan Hayward <alan.hayward@arm.com>
13660 David Sherwood <david.sherwood@arm.com>
13662 * omp-simd-clone.c (simd_clone_subparts): New function.
13663 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13664 (ipa_simd_modify_function_body): Likewise.
13666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13667 Alan Hayward <alan.hayward@arm.com>
13668 David Sherwood <david.sherwood@arm.com>
13670 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13671 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13672 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13673 (expand_vector_condition, vector_element): Likewise.
13674 (subparts_gt): New function.
13675 (get_compute_type): Use subparts_gt.
13676 (count_type_subparts): Delete.
13677 (expand_vector_operations_1): Use subparts_gt instead of
13678 count_type_subparts.
13680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13681 Alan Hayward <alan.hayward@arm.com>
13682 David Sherwood <david.sherwood@arm.com>
13684 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13685 (vect_compile_time_alias): ...this new function. Do the calculation
13686 on poly_ints rather than trees.
13687 (vect_prune_runtime_alias_test_list): Update call accordingly.
13689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13690 Alan Hayward <alan.hayward@arm.com>
13691 David Sherwood <david.sherwood@arm.com>
13693 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13695 (vect_schedule_slp_instance): Likewise.
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 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13702 constant and extern definitions for variable-length vectors.
13703 (vect_get_constant_vectors): Note that the number of units
13704 is known to be constant.
13706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13707 Alan Hayward <alan.hayward@arm.com>
13708 David Sherwood <david.sherwood@arm.com>
13710 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13711 of units as polynomial. Choose between WIDE and NARROW based
13714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13715 Alan Hayward <alan.hayward@arm.com>
13716 David Sherwood <david.sherwood@arm.com>
13718 * tree-vect-stmts.c (simd_clone_subparts): New function.
13719 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13722 Alan Hayward <alan.hayward@arm.com>
13723 David Sherwood <david.sherwood@arm.com>
13725 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13726 vectors as polynomial. Use build_index_vector for
13727 IFN_GOMP_SIMD_LANE.
13729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13730 Alan Hayward <alan.hayward@arm.com>
13731 David Sherwood <david.sherwood@arm.com>
13733 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13734 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13735 for variable-length vectors.
13736 (vectorizable_mask_load_store): Treat the number of units as
13737 polynomial, asserting that it is constant if the condition has
13738 already been enforced.
13739 (vectorizable_store, vectorizable_load): Likewise.
13741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13742 Alan Hayward <alan.hayward@arm.com>
13743 David Sherwood <david.sherwood@arm.com>
13745 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13746 of units as polynomial. Punt if we can't tell at compile time
13747 which vector contains the final result.
13749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13750 Alan Hayward <alan.hayward@arm.com>
13751 David Sherwood <david.sherwood@arm.com>
13753 * tree-vect-loop.c (vectorizable_induction): Treat the number
13754 of units as polynomial. Punt on SLP inductions. Use an integer
13755 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13756 cast of such a series for variable-length floating-point
13759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13760 Alan Hayward <alan.hayward@arm.com>
13761 David Sherwood <david.sherwood@arm.com>
13763 * tree.h (build_index_vector): Declare.
13764 * tree.c (build_index_vector): New function.
13765 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13766 of units as polynomial, forcibly converting it to a constant if
13767 vectorizable_reduction has already enforced the condition.
13768 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13769 to create a {1,2,3,...} vector.
13770 (vectorizable_reduction): Treat the number of units as polynomial.
13771 Choose vectype_in based on the largest scalar element size rather
13772 than the smallest number of units. Enforce the restrictions
13775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13776 Alan Hayward <alan.hayward@arm.com>
13777 David Sherwood <david.sherwood@arm.com>
13779 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13780 number of units as polynomial.
13782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13783 Alan Hayward <alan.hayward@arm.com>
13784 David Sherwood <david.sherwood@arm.com>
13786 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13787 * target.def (autovectorize_vector_sizes): Return the vector sizes
13788 by pointer, using vector_sizes rather than a bitmask.
13789 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13790 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13791 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13793 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13794 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13795 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13796 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13797 * omp-general.c (omp_max_vf): Likewise.
13798 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13799 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13800 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13801 * tree-vect-slp.c (vect_slp_bb): Likewise.
13802 * doc/tm.texi: Regenerate.
13803 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13805 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13806 the vector size as a poly_uint64 rather than an unsigned int.
13807 (current_vector_size): Change from an unsigned int to a poly_uint64.
13808 (get_vectype_for_scalar_type): Update accordingly.
13809 * tree.h (build_truth_vector_type): Take the size and number of
13810 units as a poly_uint64 rather than an unsigned int.
13811 (build_vector_type): Add a temporary overload that takes
13812 the number of units as a poly_uint64 rather than an unsigned int.
13813 * tree.c (make_vector_type): Likewise.
13814 (build_truth_vector_type): Take the number of units as a poly_uint64
13815 rather than an unsigned int.
13817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13818 Alan Hayward <alan.hayward@arm.com>
13819 David Sherwood <david.sherwood@arm.com>
13821 * target.def (get_mask_mode): Take the number of units and length
13822 as poly_uint64s rather than unsigned ints.
13823 * targhooks.h (default_get_mask_mode): Update accordingly.
13824 * targhooks.c (default_get_mask_mode): Likewise.
13825 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13826 * doc/tm.texi: Regenerate.
13828 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13829 Alan Hayward <alan.hayward@arm.com>
13830 David Sherwood <david.sherwood@arm.com>
13832 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13833 * omp-general.c (omp_max_vf): Likewise.
13834 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13835 (expand_omp_simd): Handle polynomial safelen.
13836 * omp-low.c (omplow_simd_context): Add a default constructor.
13837 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13838 (lower_rec_simd_input_clauses): Update accordingly.
13839 (lower_rec_input_clauses): Likewise.
13841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13842 Alan Hayward <alan.hayward@arm.com>
13843 David Sherwood <david.sherwood@arm.com>
13845 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13846 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13847 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13848 (vect_analyze_slp_cost): Likewise.
13849 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13850 (vect_model_load_cost): Likewise.
13852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13853 Alan Hayward <alan.hayward@arm.com>
13854 David Sherwood <david.sherwood@arm.com>
13856 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13857 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13858 from an unsigned int * to a poly_uint64_pod *.
13859 (calculate_unrolling_factor): New function.
13860 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13863 Alan Hayward <alan.hayward@arm.com>
13864 David Sherwood <david.sherwood@arm.com>
13866 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13867 from an unsigned int to a poly_uint64.
13868 (_loop_vec_info::slp_unrolling_factor): Likewise.
13869 (_loop_vec_info::vectorization_factor): Change from an int
13871 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13872 (vect_get_num_vectors): New function.
13873 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13874 (vect_get_num_copies): Use vect_get_num_vectors.
13875 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13876 to an unsigned int *.
13877 (vect_analyze_data_refs): Change min_vf from an int * to a
13879 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13880 than an unsigned HOST_WIDE_INT.
13881 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13882 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13883 to an unsigned int *.
13884 (vect_analyze_data_ref_dependences): Likewise.
13885 (vect_compute_data_ref_alignment): Handle polynomial vf.
13886 (vect_enhance_data_refs_alignment): Likewise.
13887 (vect_prune_runtime_alias_test_list): Likewise.
13888 (vect_shift_permute_load_chain): Likewise.
13889 (vect_supportable_dr_alignment): Likewise.
13890 (dependence_distance_ge_vf): Take the vectorization factor as a
13891 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13892 (vect_analyze_data_refs): Change min_vf from an int * to a
13894 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13895 vfm1 as a poly_uint64 rather than an int. Make the same change
13896 for the returned bound_scalar.
13897 (vect_gen_vector_loop_niters): Handle polynomial vf.
13898 (vect_do_peeling): Likewise. Update call to
13899 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13900 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13902 * tree-vect-loop.c (vect_determine_vectorization_factor)
13903 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13904 (vect_get_known_peeling_cost): Likewise.
13905 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13906 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13907 (vect_transform_loop): Likewise. Use the lowest possible VF when
13908 updating the upper bounds of the loop.
13909 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13910 rather than an int.
13911 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13912 polynomial unroll factors.
13913 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13914 (vect_make_slp_decision): Likewise.
13915 (vect_supported_load_permutation_p): Likewise, and polynomial
13917 (vect_analyze_slp_cost): Handle polynomial vf.
13918 (vect_slp_analyze_node_operations): Likewise.
13919 (vect_slp_analyze_bb_1): Likewise.
13920 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13921 than an unsigned HOST_WIDE_INT.
13922 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13923 (vectorizable_load): Handle polynomial vf.
13924 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13926 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13929 Alan Hayward <alan.hayward@arm.com>
13930 David Sherwood <david.sherwood@arm.com>
13932 * match.pd: Handle bit operations involving three constants
13933 and try to fold one pair.
13935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13937 * tree-vect-loop-manip.c: Include gimple-fold.h.
13938 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13939 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13940 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13941 Add a path that uses a step of VF instead of 1, but disable it
13943 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13944 and niters_no_overflow parameters. Update calls to
13945 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13946 Create a new SSA name if the latter choses to use a ste other
13947 than zero, and return it via niters_vector_mult_vf_var.
13948 * tree-vect-loop.c (vect_transform_loop): Update calls to
13949 vect_do_peeling, vect_gen_vector_loop_niters and
13950 slpeel_make_loop_iterate_ntimes.
13951 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13952 (vect_gen_vector_loop_niters): Update declarations after above changes.
13954 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13956 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13957 128-bit round to integer instructions.
13958 (ceil<mode>2): Likewise.
13959 (btrunc<mode>2): Likewise.
13960 (round<mode>2): Likewise.
13962 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13964 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13965 unaligned VSX load/store on P8/P9.
13966 (expand_block_clear): Allow the use of unaligned VSX
13967 load/store on P8/P9.
13969 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13971 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13973 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13974 swap associated with both a load and a store.
13976 2018-01-02 Andrew Waterman <andrew@sifive.com>
13978 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13979 * config/riscv/riscv.md (clear_cache): Use it.
13981 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13983 * web.c: Remove out-of-date comment.
13985 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13987 * expr.c (fixup_args_size_notes): Check that any existing
13988 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13989 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13990 (emit_single_push_insn): ...here.
13992 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13994 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13995 (const_vector_encoded_nelts): New function.
13996 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13997 (const_vector_int_elt, const_vector_elt): Declare.
13998 * emit-rtl.c (const_vector_int_elt_1): New function.
13999 (const_vector_elt): Likewise.
14000 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14001 of CONST_VECTOR_ELT.
14003 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14005 * expr.c: Include rtx-vector-builder.h.
14006 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14007 directly on the tree encoding.
14008 (const_vector_from_tree): Likewise.
14009 * optabs.c: Include rtx-vector-builder.h.
14010 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14011 sequence of "u" values.
14012 * vec-perm-indices.c: Include rtx-vector-builder.h.
14013 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14014 directly on the vec_perm_indices encoding.
14016 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14018 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14019 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14020 * rtx-vector-builder.h: New file.
14021 * rtx-vector-builder.c: Likewise.
14022 * rtl.h (rtx_def::u2): Add a const_vector field.
14023 (CONST_VECTOR_NPATTERNS): New macro.
14024 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14025 (CONST_VECTOR_DUPLICATE_P): Likewise.
14026 (CONST_VECTOR_STEPPED_P): Likewise.
14027 (CONST_VECTOR_ENCODED_ELT): Likewise.
14028 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14029 (unwrap_const_vec_duplicate): Likewise.
14030 (const_vec_series_p): Check for a non-duplicated vector encoding.
14031 Say that the function only returns true for integer vectors.
14032 * emit-rtl.c: Include rtx-vector-builder.h.
14033 (gen_const_vec_duplicate_1): Delete.
14034 (gen_const_vector): Call gen_const_vec_duplicate instead of
14035 gen_const_vec_duplicate_1.
14036 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14037 (gen_const_vec_duplicate): Use rtx_vector_builder.
14038 (gen_const_vec_series): Likewise.
14039 (gen_rtx_CONST_VECTOR): Likewise.
14040 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14041 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14042 Build a new vector rather than modifying a CONST_VECTOR in-place.
14043 (handle_special_swappables): Update call accordingly.
14044 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14045 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14046 Build a new vector rather than modifying a CONST_VECTOR in-place.
14047 (handle_special_swappables): Update call accordingly.
14049 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14051 * simplify-rtx.c (simplify_const_binary_operation): Use
14052 CONST_VECTOR_ELT instead of XVECEXP.
14054 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14056 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14057 the selector elements to be different from the data elements
14058 if the selector is a VECTOR_CST.
14059 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14060 ssizetype for the selector.
14062 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14064 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14065 before testing each element individually.
14066 * tree-vect-generic.c (lower_vec_perm): Likewise.
14068 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14070 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14071 * selftest-run-tests.c (selftest::run_tests): Call it.
14072 * vector-builder.h (vector_builder::operator ==): New function.
14073 (vector_builder::operator !=): Likewise.
14074 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14075 (vec_perm_indices::all_from_input_p): New function.
14076 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14077 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14078 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14079 instead of reading the VECTOR_CST directly. Detect whether both
14080 vector inputs are the same before constructing the vec_perm_indices,
14081 and update the number of inputs argument accordingly. Use the
14082 utility functions added above. Only construct sel2 if we need to.
14084 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14086 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14087 the broadcast of the low byte.
14088 (expand_mult_highpart): Use an explicit encoding for the permutes.
14089 * optabs-query.c (can_mult_highpart_p): Likewise.
14090 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14091 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14092 (vectorizable_bswap): Likewise.
14093 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14094 explicit encoding for the power-of-2 permutes.
14095 (vect_permute_store_chain): Likewise.
14096 (vect_grouped_load_supported): Likewise.
14097 (vect_permute_load_chain): Likewise.
14099 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14101 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14102 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14103 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14104 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14105 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14106 (vect_gen_perm_mask_any): Likewise.
14108 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14110 * int-vector-builder.h: New file.
14111 * vec-perm-indices.h: Include int-vector-builder.h.
14112 (vec_perm_indices): Redefine as an int_vector_builder.
14113 (auto_vec_perm_indices): Delete.
14114 (vec_perm_builder): Redefine as a stand-alone class.
14115 (vec_perm_indices::vec_perm_indices): New function.
14116 (vec_perm_indices::clamp): Likewise.
14117 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14118 (vec_perm_indices::new_vector): New function.
14119 (vec_perm_indices::new_expanded_vector): Update for new
14120 vec_perm_indices class.
14121 (vec_perm_indices::rotate_inputs): New function.
14122 (vec_perm_indices::all_in_range_p): Operate directly on the
14123 encoded form, without computing elided elements.
14124 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14125 encoding. Update for new vec_perm_indices class.
14126 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14127 the given vec_perm_builder.
14128 (expand_vec_perm_var): Update vec_perm_builder constructor.
14129 (expand_mult_highpart): Use vec_perm_builder instead of
14130 auto_vec_perm_indices.
14131 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14132 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14133 or double series encoding as appropriate.
14134 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14135 vec_perm_indices instead of auto_vec_perm_indices.
14136 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14137 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14138 (vect_permute_store_chain): Likewise.
14139 (vect_grouped_load_supported): Likewise.
14140 (vect_permute_load_chain): Likewise.
14141 (vect_shift_permute_load_chain): Likewise.
14142 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14143 (vect_transform_slp_perm_load): Likewise.
14144 (vect_schedule_slp_instance): Likewise.
14145 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14146 (vectorizable_mask_load_store): Likewise.
14147 (vectorizable_bswap): Likewise.
14148 (vectorizable_store): Likewise.
14149 (vectorizable_load): Likewise.
14150 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14151 vec_perm_indices instead of auto_vec_perm_indices. Use
14152 tree_to_vec_perm_builder to read the vector from a tree.
14153 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14154 vec_perm_builder instead of a vec_perm_indices.
14155 (have_whole_vector_shift): Use vec_perm_builder and
14156 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14157 truncation to calc_vec_perm_mask_for_shift.
14158 (vect_create_epilog_for_reduction): Likewise.
14159 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14160 from auto_vec_perm_indices to vec_perm_indices.
14161 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14162 instead of changing individual elements.
14163 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14164 the vector in d.perm.
14165 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14166 from auto_vec_perm_indices to vec_perm_indices.
14167 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14168 instead of changing individual elements.
14169 (arm_vectorize_vec_perm_const): Use new_vector to install
14170 the vector in d.perm.
14171 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14172 Update vec_perm_builder constructor.
14173 (rs6000_expand_interleave): Likewise.
14174 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14175 (rs6000_expand_interleave): Likewise.
14177 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14179 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14180 to qimode could truncate the indices.
14181 * optabs.c (expand_vec_perm_var): Likewise.
14183 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14185 * Makefile.in (OBJS): Add vec-perm-indices.o.
14186 * vec-perm-indices.h: New file.
14187 * vec-perm-indices.c: Likewise.
14188 * target.h (vec_perm_indices): Replace with a forward class
14190 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14191 * optabs.h: Include vec-perm-indices.h.
14192 (expand_vec_perm): Delete.
14193 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14194 (expand_vec_perm_const): Declare.
14195 * target.def (vec_perm_const_ok): Replace with...
14196 (vec_perm_const): ...this new hook.
14197 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14198 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14199 * doc/tm.texi: Regenerate.
14200 * optabs.def (vec_perm_const): Delete.
14201 * doc/md.texi (vec_perm_const): Likewise.
14202 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14203 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14204 expand_vec_perm for constant permutation vectors. Assert that
14205 the mode of variable permutation vectors is the integer equivalent
14206 of the mode that is being permuted.
14207 * optabs-query.h (selector_fits_mode_p): Declare.
14208 * optabs-query.c: Include vec-perm-indices.h.
14209 (selector_fits_mode_p): New function.
14210 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14211 is defined, instead of checking whether the vec_perm_const_optab
14212 exists. Use targetm.vectorize.vec_perm_const instead of
14213 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14214 fit in the vector mode before using a variable permute.
14215 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14216 vec_perm_indices instead of an rtx.
14217 (expand_vec_perm): Replace with...
14218 (expand_vec_perm_const): ...this new function. Take the selector
14219 as a vec_perm_indices rather than an rtx. Also take the mode of
14220 the selector. Update call to shift_amt_for_vec_perm_mask.
14221 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14222 Use vec_perm_indices::new_expanded_vector to expand the original
14223 selector into bytes. Check whether the indices fit in the vector
14224 mode before using a variable permute.
14225 (expand_vec_perm_var): Make global.
14226 (expand_mult_highpart): Use expand_vec_perm_const.
14227 * fold-const.c: Includes vec-perm-indices.h.
14228 * tree-ssa-forwprop.c: Likewise.
14229 * tree-vect-data-refs.c: Likewise.
14230 * tree-vect-generic.c: Likewise.
14231 * tree-vect-loop.c: Likewise.
14232 * tree-vect-slp.c: Likewise.
14233 * tree-vect-stmts.c: Likewise.
14234 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14236 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14237 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14238 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14239 (aarch64_vectorize_vec_perm_const): ...this new function.
14240 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14241 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14242 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14243 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14244 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14245 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14246 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14248 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14249 check for NEON modes.
14250 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14251 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14252 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14253 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14255 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14256 the old VEC_PERM_CONST conditions.
14257 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14258 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14259 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14260 (ia64_vectorize_vec_perm_const_ok): Merge into...
14261 (ia64_vectorize_vec_perm_const): ...this new function.
14262 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14263 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14264 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14265 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14266 * config/mips/mips.c (mips_expand_vec_perm_const)
14267 (mips_vectorize_vec_perm_const_ok): Merge into...
14268 (mips_vectorize_vec_perm_const): ...this new function.
14269 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14270 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14271 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14272 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14273 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14274 (rs6000_expand_vec_perm_const): Delete.
14275 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14277 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14278 (altivec_expand_vec_perm_const_le): Take each operand individually.
14279 Operate on constant selectors rather than rtxes.
14280 (altivec_expand_vec_perm_const): Likewise. Update call to
14281 altivec_expand_vec_perm_const_le.
14282 (rs6000_expand_vec_perm_const): Delete.
14283 (rs6000_vectorize_vec_perm_const_ok): Delete.
14284 (rs6000_vectorize_vec_perm_const): New function.
14285 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14286 an element count and rtx array.
14287 (rs6000_expand_extract_even): Update call accordingly.
14288 (rs6000_expand_interleave): Likewise.
14289 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14290 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14291 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14292 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14293 (rs6000_expand_vec_perm_const): Delete.
14294 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14295 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14296 (altivec_expand_vec_perm_const_le): Take each operand individually.
14297 Operate on constant selectors rather than rtxes.
14298 (altivec_expand_vec_perm_const): Likewise. Update call to
14299 altivec_expand_vec_perm_const_le.
14300 (rs6000_expand_vec_perm_const): Delete.
14301 (rs6000_vectorize_vec_perm_const_ok): Delete.
14302 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14303 reference to the SPE evmerge intructions.
14304 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14305 an element count and rtx array.
14306 (rs6000_expand_extract_even): Update call accordingly.
14307 (rs6000_expand_interleave): Likewise.
14308 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14309 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14311 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14313 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14315 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14316 vector mode and that that mode matches the mode of the data
14318 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14319 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14320 directly using expand_vec_perm_1 when forcing selectors into
14322 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14324 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14326 * optabs-query.h (can_vec_perm_p): Delete.
14327 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14328 * optabs-query.c (can_vec_perm_p): Split into...
14329 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14330 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14331 particular selector is valid.
14332 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14333 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14334 (vect_grouped_load_supported): Likewise.
14335 (vect_shift_permute_load_chain): Likewise.
14336 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14337 (vect_transform_slp_perm_load): Likewise.
14338 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14339 (vectorizable_bswap): Likewise.
14340 (vect_gen_perm_mask_checked): Likewise.
14341 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14342 implementations of variable permutation vectors into account
14343 when deciding which selector to use.
14344 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14345 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14346 with a false third argument.
14347 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14348 to test whether the constant selector is valid and can_vec_perm_var_p
14349 to test whether a variable selector is valid.
14351 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14353 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14354 * optabs-query.c (can_vec_perm_p): Likewise.
14355 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14356 instead of vec_perm_indices.
14357 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14358 (vect_gen_perm_mask_checked): Likewise,
14359 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14360 (vect_gen_perm_mask_checked): Likewise,
14362 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14364 * optabs-query.h (qimode_for_vec_perm): Declare.
14365 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14366 (qimode_for_vec_perm): ...this new function.
14367 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14369 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14371 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14372 does not have a conditional at the top.
14374 2018-01-02 Richard Biener <rguenther@suse.de>
14376 * ipa-inline.c (big_speedup_p): Fix expression.
14378 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14381 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14384 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14388 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14389 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14390 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14391 cond_taken_branch_cost 3->4.
14393 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14395 PR tree-optimization/83581
14396 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14397 TODO_cleanup_cfg if any changes have been made.
14399 PR middle-end/83608
14400 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14401 convert_modes if target mode has the right side, but different mode
14404 PR middle-end/83609
14405 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14406 last argument when extracting from CONCAT. If either from_real or
14407 from_imag is NULL, use expansion through memory. If result is not
14408 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14409 the parts directly to inner mode, if even that fails, use expansion
14412 PR middle-end/83623
14413 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14414 check for bswap in mode rather than HImode and use that in expand_unop
14417 Copyright (C) 2018 Free Software Foundation, Inc.
14419 Copying and distribution of this file, with or without modification,
14420 are permitted in any medium without royalty provided the copyright
14421 notice and this notice are preserved.