1 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
7 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
8 (nds32_init_machine_status): Initialize machine->attr_naked_p and
9 machine->attr_no_prologue_p.
10 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
11 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
12 (nds32_expand_epilogue): Consider attr_naked_p.
13 (nds32_expand_epilogue_v3pop): Likewise.
14 (nds32_can_use_return_insn): Likewise.
15 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
16 attr_no_prologue_p fields.
17 * config/nds32/nds32.opt (mret-in-naked-func): New option.
19 2018-05-27 Jakub Jelinek <jakub@redhat.com>
22 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
25 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
27 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
29 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
31 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
33 (*<floatsuffix>floatv2div2sf2): Rename to ...
34 (*float<floatunssuffix>v2div2sf2): ... this.
35 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
36 (float<floatunssuffix>v2div2sf2_mask): ... this.
37 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
38 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
39 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
41 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
43 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
45 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
47 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
49 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
51 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
52 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
53 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
54 gen_ufix_truncv8dfv8si2.
55 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
56 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
57 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
58 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
59 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
60 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
61 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
62 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
64 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
68 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
70 2018-05-25 Jim Wilson <jimw@sifive.com>
72 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
73 * config/riscv/riscv.c (struct machine_function): Add
74 interrupt_handler_p and attribute_checked_p fields.
75 (riscv_attribute_table): Add interrupt.
76 (riscv_interrupt_type_p): New.
77 (riscv_save_reg_p): Save extra regs for interrupt handler.
78 (riscv_use_save_libcall): Return false for interrupt handler.
79 (riscv_first_stack_step): Add forward declaration.
80 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
81 for interrupt handler with large frame. Use it for saved reg list.
82 (riscv_expand_prologue): Move flag_stack_usage_info support to
83 eliminate duplication.
84 (riscv_expand_epilogue): Generate mret for interrupt handler.
85 (riscv_epilogue_uses): New.
86 (riscv_can_use_return_insn): Return false for interrupt handler.
87 (riscv_function_ok_for_sibcall): Likewise.
88 (riscv_set_current_function): Add interrupt handler support.
89 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
90 * config/riscv/riscv.md (UNSPECV_MRET): New.
92 (riscv_frflags, riscv_fsflags): Use tab after opcode.
94 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
96 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
98 PR tree-optimization/85712
99 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
100 this candidate has already been replaced in-situ by a copy.
102 2018-05-25 Jason Merrill <jason@redhat.com>
104 PR c++/80485 - inline function non-zero address.
105 * symtab.c (nonzero_address): Check DECL_COMDAT.
107 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
110 * config/alpha/alpha.md (ashlsi3): New insn pattern.
111 (*ashlsi_se): Rename from *ashldi_se. Define as sign
112 extension of SImode operation. Use const123_operand predicate.
114 (*saddl_se_1): Ditto.
116 (*ssubl_se_1): Ditto.
117 * config/alpha/predicates.md (const123_operand): New predicate.
118 * config/alpha/constraints.md (P): Use IN_RANGE.
120 2018-05-25 Richard Biener <rguenther@suse.de>
122 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
124 (ref_maybe_used_by_stmt_p): Likewise.
125 (stmt_may_clobber_ref_p): Likewise.
126 (stmt_may_clobber_ref_p_1): Likewise.
127 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
129 (ref_maybe_used_by_stmt_p): Likewise.
130 (stmt_may_clobber_ref_p): Likewise.
131 (stmt_may_clobber_ref_p_1): Likewise.
132 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
133 the alias oracle to disambiguate DRs with stmts DR analysis
135 (vect_analyze_data_refs): Do not give up on not analyzable
136 DRs for BB vectorization. Remove code truncating the dataref
139 2018-05-25 Jakub Jelinek <jakub@redhat.com>
142 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
143 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
144 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
146 2018-05-25 Richard Biener <rguenther@suse.de>
148 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
149 function, combining stmt data ref gathering and fatal analysis
151 (vect_analyze_data_refs): Remove now redudnant code and simplify.
152 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
153 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
154 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
155 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
157 2018-05-25 Bin Cheng <bin.cheng@arm.com>
159 PR tree-optimization/85720
160 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
161 SCC if all partitions are builtins.
162 (version_loop_by_alias_check): New parameter. Generate cancelable
163 runtime alias check if all partitions are builtins.
164 (distribute_loop): Update call to above function.
166 2018-05-25 Bin Cheng <bin.cheng@arm.com>
168 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
169 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
170 (parm_default_def_partition_arg): Ditto.
171 (set_parm_default_def_partition): Ditto.
172 (get_parm_default_def_partitions): Ditto and make it static.
173 (get_undefined_value_partitions): Ditto and make it static.
174 (remove_ssa_form): Refactor call to init_var_map here.
175 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
176 computation for loop region.
177 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
178 (register_default_def): Delete.
179 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
180 (parm_default_def_partition_arg): Ditto.
181 (set_parm_default_def_partition): Ditto.
182 (get_parm_default_def_partitions): Ditto and make it static.
183 (get_undefined_value_partitions): Ditto and make it static.
184 (coalesce_with_default, coalesce_with_default): Update comment.
185 (create_coalesce_list_for_region): New func factored out from
186 create_outofssa_var_map.
187 (populate_coalesce_list_for_outofssa): New func factored out from
188 create_outofssa_var_map and coalesce_ssa_name.
189 (create_outofssa_var_map): Delete.
190 (coalesce_ssa_name): Refactor to support live range computation.
191 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
192 (get_parm_default_def_partitions): Delete.
193 (get_undefined_value_partitions): Ditto.
194 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
195 computation for loop region.
196 (new_tree_live_info, loe_visit_block): Ditto.
197 (live_worklist, set_var_live_on_entry): Ditto.
198 (calculate_live_on_exit, verify_live_on_entry): Ditto.
199 * tree-ssa-live.h (struct _var_map): New fields.
200 (init_var_map): Change decl.
201 (region_contains_p): New.
203 2018-05-25 Bin Cheng <bin.cheng@arm.com>
205 * tree-ssa-live.h (live_merge_and_clear): Delete.
207 2018-05-25 Richard Biener <rguenther@suse.de>
210 * tree-dump.c (dequeue_and_dump): Remove access to removed
211 operand 2 of a SWITCH_EXPR.
213 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
215 * doc/sourcebuild.texi (vect_double_cond_arith): Include
216 multiplication and division.
217 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
218 (cond_udiv@var{m}, cond_umod@var{m}): Document.
219 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
220 (cond_udiv_optab, cond_umod_optab): New optabs.
221 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
222 (IFN_COND_RDIV): New internal functions.
223 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
224 TRUNC_MOD_EXPR and RDIV_EXPR.
225 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
226 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
228 (SVE_INT_BINARY): Include mult.
229 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
230 (optab, sve_int_op): Handle mult.
231 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
233 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
234 for SVE_INT_BINARY_SD.
236 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
238 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
239 (optab, sve_int_op): Handle div and udiv.
240 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
241 for SVE_INT_BINARY_SD.
242 (*<optab><mode>3): New insn for the same.
244 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
246 * tree-vect-patterns.c: Include predict.h.
247 (vect_recog_divmod_pattern): Restrict check for division support
248 to when optimizing for size.
250 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
252 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
253 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
254 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
255 (gimple_match_op::set_op): Likewise.
256 (gimple_resimplify4): Declare.
257 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
258 (expr::gen_transform): Likewise.
259 (decision_tree::gen): Generate a simplification routine for 4 operands.
260 * gimple-match-head.c (gimple_simplify): Add an overload for
261 4 operands. In the top-level function, handle up to 4 call
262 arguments and call gimple_resimplify4.
263 (gimple_resimplify4): New function.
264 (build_call_internal): Pass a fourth operand.
265 (maybe_push_to_seq): Likewise.
266 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
267 Fold VEC_COND_EXPRs of an operation and a default value into
268 an IFN_COND_* function if possible.
269 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
271 (SVE_COND_FP_BINARY): Include them.
272 (optab, sve_fp_op): Handle them.
273 (SVE_INT_BINARY_REV): New code iterator.
274 (SVE_COND_FP_BINARY_REV): New int iterator.
275 (commutative): New int attribute.
276 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
278 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
280 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
281 (*cond_<optab><mode>): New patterns for reversed operands.
283 2018-05-25 Richard Biener <rguenther@suse.de>
285 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
286 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
287 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
288 (STMT_VINFO_GROUPED_ACCESS): Adjust.
289 * tree-vect-data-refs.c (everywhere): Adjust users.
290 * tree-vect-loop.c (everywhere): Likewise.
291 * tree-vect-slp.c (everywhere): Likewise.
292 * tree-vect-stmts.c (everywhere): Likewise.
293 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
295 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
297 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
299 (gcc_cv_as_section_exclude): ... this.
300 Try Solaris as #exclude syntax.
301 * configure: Regenerate.
302 * config.in: Regenerate.
303 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
305 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
306 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
308 * varasm.c (default_elf_asm_named_section): Don't check if
309 HAVE_GAS_SECTION_EXCLUDE is defined.
311 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
313 * doc/md.texi: Update the documentation of the cond_* optabs
314 to mention the new final operand. Fix GET_MODE_NUNITS call.
315 Describe the scalar case too.
316 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
317 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
319 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
320 (get_conditional_internal_fn): Update comment.
321 * tree-vect-loop.c (vectorizable_reduction): Pass the original
322 accumulator value as a final argument to conditional functions.
323 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
324 a define_expand and add an "else" operand. Assert for now that
325 the else operand is equal to operand 2. Use SVE_INT_BINARY and
326 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
327 (*cond_<optab><mode>): New patterns.
328 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
329 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
330 (UNSPEC_COND_EOR): Delete.
331 (optab): Remove associated mappings.
332 (SVE_INT_BINARY): New code iterator.
333 (sve_int_op): Remove int attribute and add "minus" to the code
335 (SVE_COND_INT_OP): Delete.
336 (SVE_COND_FP_OP): Rename to...
337 (SVE_COND_FP_BINARY): ...this.
339 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
341 * optabs.c (can_reuse_operands_p): New function.
342 (maybe_legitimize_operands): Try to reuse the results for
345 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
347 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
348 Add {q} suffix to insn mnemonic.
350 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
352 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
353 (msp430_warn_func_return): New.
355 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
357 * fold-const.c (tree_nonzero_bits): New function.
358 * fold-const.h (tree_nonzero_bits): Likewise.
359 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
360 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
362 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
366 * varasm.c (assemble_alias): Check ifunc_resolver only on
369 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
372 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
373 when memory input operand is handled.
375 2018-05-24 Luis Machado <luis.machado@linaro.org>
377 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
379 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
381 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
383 * match.pd: Delay FMA folds until after vectorization.
385 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
388 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
389 address check not strict.
391 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
393 * gimple-match.h (gimple_match_op): New class.
394 (mprts_hook): Replace parameters with a gimple_match_op *.
395 (maybe_build_generic_op): Likewise.
396 (gimple_simplified_result_is_gimple_val): Replace parameters with
397 a const gimple_match_op *.
398 (gimple_simplify): Replace code_helper * and tree * parameters with
399 a gimple_match_op * parameter.
400 (gimple_resimplify1): Replace code_helper *, tree and tree *
401 parameters with a gimple_match_op * parameter.
402 (gimple_resimplify2): Likewise.
403 (gimple_resimplify3): Likewise.
404 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
405 parameters with a gimple_match_op * parameter.
406 * gimple-match-head.c (gimple_simplify): Change prototypes of
407 auto-generated functions to take a gimple_match_op * instead of
408 separate code_helper * and tree * parameters. Make the same
409 change in the top-level overload and update calls to the
410 gimple_resimplify routines. Update calls to the auto-generated
411 functions and to maybe_push_res_to_seq in the publicly-facing
412 operation-specific gimple_simplify overloads.
413 (gimple_match_op::MAX_NUM_OPS): Define.
414 (gimple_resimplify1): Replace rcode and ops with a single res_op
415 parameter. Update call to gimple_simplify.
416 (gimple_resimplify2): Likewise.
417 (gimple_resimplify3): Likewise.
418 (mprts_hook): Replace parameters with a gimple_match_op *.
419 (maybe_build_generic_op): Likewise.
420 (build_call_internal): Replace type, nargs and ops with
422 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
423 with a single gimple_match_op *. Update calls to mprts_hook,
424 build_call_internal and gimple_simplified_result_is_gimple_val.
425 Factor out code that is common to the tree_code and combined_fn cases.
426 * genmatch.c (expr::gen_transform): Replace tem_code and
427 tem_ops with a gimple_match_op called tem_op. Update calls
428 to the gimple_resimplify functions and maybe_push_res_to_seq.
429 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
430 res_ops. Update call to the gimple_resimplify functions.
431 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
432 (decision_tree::gen): Make the functions take a gimple_match_op *
433 called res_op instead of separate res_code and res_ops parameters.
434 Update call accordingly.
435 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
436 and ops with a single res_op parameter. Update calls to
437 maybe_build_generic_op and maybe_push_res_to_seq.
438 (fold_stmt_1): Update calls to gimple_simplify and
439 replace_stmt_with_simplification.
440 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
441 and gimple_simplified_result_is_gimple_val.
442 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
444 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
445 with a gimple_match_op *.
446 (vn_nary_build_or_lookup): Likewise. Update call to
447 vn_nary_build_or_lookup_1.
448 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
449 gimple_match_op *. Update calls to the gimple_resimplify routines
450 and to gimple_simplified_result_is_gimple_val.
451 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
452 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
453 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
454 (visit_nary_op): Likewise.
455 (visit_reference_op_load): Likewise.
457 2018-05-23 Luis Machado <luis.machado@linaro.org>
459 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
460 modifier for printing the step amount.
462 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
465 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
468 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
470 * doc/sourcebuild.texi (Endianness): New subsubsection.
472 2018-05-23 Luis Machado <luis.machado@linaro.org>
474 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
475 <prefetch_dynamic_strides>: New const bool field.
476 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
477 prefetch_dynamic_strides.
478 (exynosm1_prefetch_tune): Likewise.
479 (thunderxt88_prefetch_tune): Likewise.
480 (thunderx_prefetch_tune): Likewise.
481 (thunderx2t99_prefetch_tune): Likewise.
482 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
484 (aarch64_override_options_internal): Update to set
485 PARAM_PREFETCH_DYNAMIC_STRIDES.
486 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
487 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
488 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
489 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
490 prefetch-dynamic-strides setting.
492 2018-05-23 Luis Machado <luis.machado@linaro.org>
494 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
495 <minimum_stride>: New const int field.
496 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
497 minimum_stride field defaulting to -1.
498 (exynosm1_prefetch_tune): Likewise.
499 (thunderxt88_prefetch_tune): Likewise.
500 (thunderx_prefetch_tune): Likewise.
501 (thunderx2t99_prefetch_tune): Likewise.
502 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
503 <default_opt_level>: Set to 3.
504 (aarch64_override_options_internal): Update to set
505 PARAM_PREFETCH_MINIMUM_STRIDE.
506 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
507 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
508 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
509 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
510 stride is constant and is below the minimum stride threshold.
512 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
514 * config/arm/arm-cpus.in (mode26): Delete.
515 (armv4): Delete mode26 reference.
516 * config/arm/arm.c (arm_configure_build_target): Delete use of
519 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
521 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
523 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
524 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
526 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
527 (floatunsdidf2): Ditto.
529 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
531 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
532 (fixuns_trunc<mode>si2_avx512f): Ditto.
533 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
534 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
535 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
537 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
539 PR rtl-optimization/79985
540 * df-scan.c (df_insn_refs_collect): Remove special case for
541 global registers and asm statements.
543 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
545 * extend.texi (Global Register Variables): Rewrite the bullet list.
546 Note that the register is available for allocation. Note that access
547 via inline asm must use constraints. Add note about async-signal
548 handlers. Remove paragraph about automagic register selection.
550 2018-05-23 Richard Biener <rguenther@suse.de>
552 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
553 of fixed offset from memset VN.
555 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
557 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
559 (alloc_cand_and_find_basis): Initialize first_interp field.
560 (slsr_process_mul): Modify first_interp field.
561 (slsr_process_add): Likewise.
562 (slsr_process_cast): Modify first_interp field for each new
564 (slsr_process_copy): Likewise.
565 (dump_candidate): Dump first_interp field.
566 (replace_mult_candidate): Process all interpretations, not just
568 (replace_rhs_if_not_dup): Likewise.
569 (replace_one_candidate): Likewise.
571 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
573 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
575 (aarch64_needs_frame_chain): New function.
576 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
578 2018-05-23 Sudakshina Das <sudi.das@arm.com>
581 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
582 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
583 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
584 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
585 as true for strict-align.
586 (aarch64_can_inline_p): Perform checks even when callee has no
587 attributes to check for strict alignment.
588 * doc/extend.texi (AArch64 Function Attributes): Document
590 * doc/invoke.texi: (AArch64 Options): Likewise.
592 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
594 PR tree-optimization/85853
595 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
596 the handling of the root of the node to...
597 (vect_slp_analyze_node_operations_1): ...this new function,
598 and run the whole thing with the child nodes' def types
599 set according to their SLP node's def type.
601 2018-05-23 Richard Biener <rguenther@suse.de>
604 * tree-data-ref.c (create_runtime_alias_checks): Defer
605 and ignore overflow warnings.
607 2018-05-23 Yury Gribov <tetra2005@gmail.com>
609 PR tree-optimization/85822
610 * tree-vrp.c (is_masked_range_test): Fix handling of negative
613 2018-05-23 Richard Biener <rguenther@suse.de>
615 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
616 memset constants via native_interpret_expr.
618 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
621 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
623 (cgraph_node::create_alias): Likewise.
624 (cgraph_node::get_availability): Check ifunc_resolver instead
625 of looking up ifunc attribute.
626 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
627 * varasm.c (do_assemble_alias): Likewise.
628 (assemble_alias): Likewise.
629 (default_binds_local_p_3): Likewise.
630 * cgraph.h (cgraph_node): Add ifunc_resolver.
631 (cgraph_node::only_called_directly_or_aliased_p): Return false
633 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
635 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
636 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
637 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
638 instead of looking up ifunc attribute.
640 2018-05-22 Luis Machado <luis.machado@linaro.org>
642 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
644 2018-05-22 Martin Sebor <msebor@redhat.com>
647 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
648 only when expasion succeeds.
649 (expand_builtin_strcmp): Same.
650 (expand_builtin_strncmp): Same.
652 2018-05-22 Martin Sebor <msebor@redhat.com>
654 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
656 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
657 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
659 * config/aarch64/aarch64-ldpstp.md: Replace uses of
660 aarch64_mem_pair_operand with memory_operand and delete operand swapping
662 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
663 Add check for legitimate_address.
664 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
665 (aarch64_swap_ldrstr_operands): New.
666 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
669 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
670 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
672 * config/aarch64/aarch64.md: New patterns to generate stp
674 (store_pair_sw, store_pair_dw): New patterns to generate stp for
675 single words and double words.
676 (load_pair_sw, load_pair_dw): Likewise.
677 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
679 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
681 * config/aarch64/aarch64-ldpstp.md: Modify peephole
682 for different mode ldpstp and add peephole for merged zero stores.
684 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
686 (aarch64_gen_store_pair): Rename calls to match new patterns.
687 (aarch64_gen_load_pair): Rename calls to match new patterns.
688 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
689 (load_pair<DREG:mode><DREG2:mode>): ... This.
690 (store_pair<mode>): Rename to...
691 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
692 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
694 (V_INT_EQUIV): Handle SImode.
695 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
698 2018-05-22 Martin Sebor <msebor@redhat.com>
701 * calls.c (maybe_warn_nonstring_arg): Use string length to set
702 or ajust the presumed bound on an operation to avoid unnecessary
705 2018-05-22 Martin Sebor <msebor@redhat.com>
707 PR tree-optimization/85826
708 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
709 assuming that a DECL necesarily has a constant size.
711 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
714 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
716 2018-05-22 Richard Biener <rguenther@suse.de>
718 PR tree-optimization/85834
719 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
720 non-constant and non-zero memset arguments.
722 2018-05-22 Martin Liska <mliska@suse.cz>
725 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
727 2018-05-22 Richard Biener <rguenther@suse.de>
729 PR tree-optimization/85863
730 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
731 comparisons when vectype is specified.
732 (vectorizable_condition): Do not specify vectype for
733 vect_is_simple_cond when SLP vectorizing.
735 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
738 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
739 define __ibm128 as long double.
740 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
741 as a distinct type when IEEE 128-bit support is enabled.
742 (init_float128_ieee): Fix up conversions between IFmode and IEEE
743 128-bit types to use the correct functions.
744 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
745 convert between 128-bit floating point types that have different
746 modes but the same representation, instead of using gen_lowpart to
748 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
750 (IFKF_reg): New attributes to give the register constraints for
752 (extend<mode>tf2_internal): New insns to mark an explicit
753 conversion between 128-bit floating point types that have a
754 different mode but share the same representation.
756 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
758 PR tree-optimization/85814
759 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
760 a null return from get_strinfo when unsharing the next
761 strinfo in the chain.
763 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
766 * varasm.c (weak_finish): Clean up weak_decls.
768 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
770 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
771 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
772 UNSPEC_UADALP values.
773 * config/aarch64/iterators.md (ABAL): New int iterator.
776 (sur): Add mappings for the above.
777 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
779 (aarch64_<sur>abal<mode>_4): Likewise.
780 (aarch64_<sur>adalp<mode>_3): Likewise.
781 (<sur>sadv16qi): New define_expand.
783 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
785 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
786 (*movdf_internal): Ditto.
787 (*rcpsf2_sse): Ditto.
788 (*rsqrtsf2_sse): Ditto.
789 (*sqrt<mode>2_sse): Ditto.
791 2018-05-21 Tamar Christina <tamar.christina@arm.com>
793 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
795 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
796 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
797 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
798 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
800 * config/aarch64/arm_neon.h: Likewise.
801 * config/aarch64/iterators.md (VQ_I): New.
803 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
805 * config.gcc: Add arc/t-multilib-linux to tmake_file for
807 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
810 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
812 * config/nds32/constraints.md (S): New constraint.
813 * config/nds32/nds32.md (call_internal): Use constraint S.
814 (call_value_internal): Likewise.
815 (sibcall_internal): Likewise.
816 (sibcall_value_internal): Likewise.
818 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
819 Chung-Ju Wu <jasonwucj@gmail.com>
821 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
824 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
825 Chung-Ju Wu <jasonwucj@gmail.com>
827 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
828 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
829 (nds32_rtx_costs_impl): Simplify.
830 (nds32_address_cost_impl): Simplify.
831 (nds32_init_rtx_costs): New function.
832 (nds32_rtx_costs_speed_prefer): Likewise.
833 (nds32_rtx_costs_size_prefer): Likewise.
834 (nds32_address_cost_speed_prefer): Likewise.
835 (nds32_address_cost_speed_fwprop): Likewise.
836 (nds32_address_cost_size_prefer): Likewise.
837 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
838 * config/nds32/nds32.c (nds32_option_override): Use
839 nds32_init_rtx_costs function.
841 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
843 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
844 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
845 (TARGET_PIPELINE_N8): Likewise.
846 (TARGET_PIPELINE_N10): Likewise.
847 (TARGET_PIPELINE_N13): Likewise.
848 (TARGET_PIPELINE_GRAYWOLF): Likewise.
850 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
852 * config/nds32/nds32-fpu.md: Update copyright year.
854 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
856 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
858 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
860 * config/nds32/nds32.c
861 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
862 * config/nds32/nds32.opt (minline-asm-r15): New option.
864 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
866 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
868 * config/nds32/nds32.md (abssi2): New pattern.
870 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
872 * config/i386/i386.md (rex64namesuffix): New mode attribute.
873 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
874 Merge insn pattern from sse_cvtsi2ss<round_name> and
875 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
876 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
877 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
878 using SWI48 mode iterator.
879 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
880 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
881 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
882 pattern from sse_cvttss2si<round_saeonly_name>
883 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
884 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
885 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
886 using SWI48 mode iterator.
887 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
888 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
889 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
890 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
891 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
892 using SWI48 mode iterator.
893 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
894 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
895 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
896 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
897 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
899 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
900 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
901 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
902 pattern from sse_cvttsd2si<round_saeonly_name>
903 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
905 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
907 * config/nds32/nds32-md-auxiliary.c
908 (nds32_valid_smw_lwm_base_p): Refine.
909 (nds32_output_smw_single_word): Refine.
910 (nds32_output_smw_double_word): New.
911 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
913 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
915 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
916 (nds32_output_stack_pop): Refine.
917 (nds32_expand_unaligned_load): Refine.
918 (nds32_expand_unaligned_store): Refine.
920 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
921 Chung-Ju Wu <jasonwucj@gmail.com>
923 * config/nds32/constants.md: Add TP_REGNUM constant.
924 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
925 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
927 * config/nds32/nds32-doubleword.md: Consider flag_pic.
928 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
929 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
930 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
931 and PIC code generation.
932 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
934 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
936 * config/nds32/nds32.md: Support TLS and PIC.
937 * config/nds32/nds32.c: Support TLS and PIC.
938 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
939 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
942 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
944 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
947 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
948 Chung-Ju Wu <jasonwucj@gmail.com>
950 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
951 * config/nds32/nds32-md-auxiliary.c
952 (symbolic_reference_mentioned_p): New.
953 (nds32_legitimize_ict_address): New.
954 (nds32_expand_ict_move): New.
955 (nds32_indirect_call_referenced_p): New.
956 (nds32_symbol_binds_local_p): Delete.
957 (nds32_long_call_p): Modify.
958 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
959 * config/nds32/nds32-protos.h
960 (symbolic_reference_mentioned_p): Declare.
961 (nds32_legitimize_ict_address): Declare.
962 (nds32_expand_ict_move): Declare.
963 (nds32_indirect_call_referenced_p): Declare.
964 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
965 (nds32_relax_group): Use nds32_ict_const_p as condition.
966 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
967 (nds32_asm_file_start): Output ict_model directive in asm code.
968 (nds32_legitimate_address_p): Consider indirect call.
969 (nds32_print_operand): Consider indirect call.
970 (nds32_print_operand_address): Consider indirect call.
971 (nds32_insert_attributes): Handle "indirect_call" attribute.
972 (TARGET_LEGITIMATE_ADDRESS_P): Define.
973 (TARGET_LEGITIMATE_CONSTANT_P): Define.
974 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
975 (TARGET_DELEGITIMIZE_ADDRESS): Define.
976 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
977 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
978 (TARGET_ICT_MODEL_SMALL): Define.
979 (TARGET_ICT_MODEL_LARGE): Define.
980 * config/nds32/nds32.md (movsi): Consider ict model.
981 (call, call_value): Consider ict model.
982 (sibcall, sibcall_value): Consider ict model.
983 * config/nds32/nds32.opt (mict-model): New option.
984 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
987 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
988 Monk Chiang <sh.chiang04@gmail.com>
989 Jim Wilson <jimw@sifive.com>
991 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
992 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
993 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
994 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
995 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
996 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
997 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
998 compute save_libcall_adjustment properly.
999 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1000 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1001 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1002 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1003 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1004 (ABI_SPEC): Handle mabi=ilp32e.
1005 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1006 (RVE): Add RVE mask.
1007 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1008 <-march>: Add rv32e as an example.
1010 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1013 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1014 (intra_create_variable_infos): Handle C++ constructors.
1016 2018-05-18 Martin Liska <mliska@suse.cz>
1018 * passes.def: Remove a redundant pass.
1020 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1023 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1025 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1027 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1029 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1031 (armv2, armv2a, armv3, armv3m): Delete architectures.
1032 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1033 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1034 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1036 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1037 (*mulsidi3adddi): Likewise.
1038 (mulsidi3): Likewise.
1039 (*mulsidi3_nov6): Likewise.
1040 (umulsidi3): Likewise.
1041 (umulsidi3_nov6): Likewise.
1042 (umaddsidi4): Likewise.
1043 (*umulsidi3adddi): Likewise.
1044 (smulsi3_highpart): Likewise.
1045 (*smulsi3_highpart_nov6): Likewise.
1046 (umulsi3_highpart): Likewise.
1047 (*umulsi3_highpart_nov6): Likewise.
1048 * config/arm/arm.h (arm_arch3m): Delete.
1049 * config/arm/arm.c (arm_arch3m): Delete.
1050 (arm_option_override_internal): Update armv3-related comment.
1051 (arm_configure_build_target): Delete use of isa_bit_mode32.
1052 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1053 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1054 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1056 (mulusa3): Likewise.
1057 * config/arm/arm-protos.h (arm_arch3m): Delete.
1058 * config/arm/arm-tables.opt: Regenerate.
1059 * config/arm/arm-tune.md: Likewise.
1060 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1061 deleted architectures.
1063 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1065 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1066 (armv5t, armv5te): New features.
1067 (ARMv5, ARMv5e): Delete fgroups.
1068 (ARMv5t, ARMv5te): Adjust for above changes.
1070 (armv5, armv5e): Delete arches.
1071 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1073 (*call_reg_arm): Likewise.
1074 (*call_value_reg_armv5): Likewise.
1075 (*call_value_reg_arm): Likewise.
1076 (*call_symbol): Likewise.
1077 (*call_value_symbol): Likewise.
1078 (*sibcall_insn): Likewise.
1079 (*sibcall_value_insn): Likewise.
1081 (prefetch): Likewise.
1082 (define_split and define_peephole2 dependent on arm_arch5):
1084 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1086 (TARGET_ARM_QBIT): Likewise.
1087 (TARGET_DSP_MULTIPLY): Likewise.
1088 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1089 (arm_arch5, arm_arch5e): Delete.
1090 (arm_arch5t, arm_arch5te): Declare.
1091 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1092 (arm_arch5t): Declare.
1093 (arm_option_reconfigure_globals): Update for the above.
1094 (arm_options_perform_arch_sanity_checks): Update comment, replace
1095 use of arm_arch5 with arm_arch5t.
1096 (use_return_insn): Likewise.
1097 (arm_emit_call_insn): Likewise.
1098 (output_return_instruction): Likewise.
1099 (arm_final_prescan_insn): Likewise.
1100 (arm_coproc_builtin_available): Likewise.
1101 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1102 arm_arch5e with arm_arch5t and arm_arch5te.
1103 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1104 (arm_arch5t, arm_arch5te): Declare.
1105 * config/arm/arm-tables.opt: Regenerate.
1106 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1107 * config/arm/t-multilib: Likewise.
1108 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1109 instead of arm_arch5.
1110 (*call_reg_thumb1): Likewise.
1111 (*call_value_reg_thumb1_v5): Likewise.
1112 (*call_value_reg_thumb1): Likewise.
1113 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1115 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1117 2018-05-18 Martin Liska <mliska@suse.cz>
1119 PR gcov-profile/84846
1120 * doc/gcov.texi: Document -t option of gcov tool.
1122 2018-05-18 Martin Liska <mliska@suse.cz>
1124 PR gcov-profile/84846
1125 * gcov.c (print_usage): Add new -t option.
1126 (process_args): Handle the option.
1127 (generate_results): Use stdout as output when requested by
1130 2018-05-18 Martin Liska <mliska@suse.cz>
1132 PR gcov-profile/84846
1133 * coverage.c (coverage_init): Write PWD to .gcno file.
1134 * doc/gcov.texi: Document how working directory is printed.
1135 * gcov-dump.c (dump_gcov_file): Print PWD.
1136 * gcov.c (output_intermediate_file): Likewise.
1137 (read_graph_file): Read PWD string.
1138 (output_lines): Print PWD.
1140 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1143 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1144 for retval and return false if all args to phi are zero.
1146 2018-05-18 Richard Biener <rguenther@suse.de>
1148 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1150 (evrp_dom_walker::before_dom_children): Call it.
1152 2018-05-18 Richard Biener <rguenther@suse.de>
1154 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1155 results when processing array refs with variable index.
1157 2018-05-18 Toon Moene <toon@moene.org>
1159 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1160 directly after that of -floop-interchange. Indicate that both
1161 options are enabled by default when specifying -O3.
1163 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1165 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1166 iterator. Delete separate integer-mode vec_set<mode> expander.
1167 (aarch64_simd_vec_setv2di): Delete.
1168 (vec_setv2di): Delete.
1169 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1170 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1171 the "w, r" alternative.
1173 2018-05-18 Martin Liska <mliska@suse.cz>
1175 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1176 * tree-pass.h (make_pass_lower_switch_O0): New function.
1177 * tree-switch-conversion.c (node_has_low_bound): Remove.
1178 (node_has_high_bound): Likewise.
1179 (node_is_bounded): Likewise.
1180 (class pass_lower_switch): Make it a template type and create
1182 (pass_lower_switch::execute): Add template argument.
1183 (make_pass_lower_switch): New function.
1184 (make_pass_lower_switch_O0): New function.
1185 (do_jump_if_equal): Remove.
1186 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1187 all the hard work to tree optimization passes.
1189 2018-05-18 Martin Liska <mliska@suse.cz>
1191 * dbgcnt.c (limit_low): Renamed from limit.
1192 (limit_high): New variable.
1193 (dbg_cnt_is_enabled): Check for upper limit.
1194 (dbg_cnt): Adjust dumping.
1195 (dbg_cnt_set_limit_by_index): Add new argument for high
1197 (dbg_cnt_set_limit_by_name): Likewise.
1198 (dbg_cnt_process_single_pair): Parse new format.
1199 (dbg_cnt_process_opt): Use strtok.
1200 (dbg_cnt_list_all_counters): Remove 'value' and add
1202 * doc/invoke.texi: Document changes.
1204 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1206 * doc/sourcebuild.texi (scalar_all_fma): Document.
1207 * tree.def (FMA_EXPR): Delete.
1208 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1209 * internal-fn.c (ternary_direct): New macro.
1210 (expand_ternary_optab_fn): Likewise.
1211 (direct_ternary_optab_supported_p): Likewise.
1212 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1213 * builtins.c (fold_builtin_fma): Delete.
1214 (fold_builtin_3): Don't call it.
1215 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1216 * expr.c (expand_expr_real_2): Likewise.
1217 * fold-const.c (operand_equal_p): Likewise.
1218 (fold_ternary_loc): Likewise.
1219 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1220 * gimple.c (DEFTREECODE): Likewise.
1221 * gimplify.c (gimplify_expr): Likewise.
1222 * optabs-tree.c (optab_for_tree_code): Likewise.
1223 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1224 * tree-eh.c (operation_could_trap_p): Likewise.
1225 (stmt_could_throw_1_p): Likewise.
1226 * tree-inline.c (estimate_operator_cost): Likewise.
1227 * tree-pretty-print.c (dump_generic_node): Likewise.
1228 (op_code_prio): Likewise.
1229 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1230 * tree-ssa-operands.c (get_expr_operands): Likewise.
1231 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1232 * fold-const-call.h (fold_fma): Delete.
1233 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1234 CFN_FNMA and CFN_FNMS.
1236 * genmatch.c (combined_fn): New enum.
1237 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1238 (commutative_op): New function.
1239 (commutate): Use it. Handle more than 2 operands.
1240 (dt_operand::gen_gimple_expr): Use commutative_op.
1241 (parser::parse_expr): Allow :c to be used with non-binary
1242 operators if the commutative operand is known.
1243 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1244 CFN_FMS, CFN_FNMA and CFN_FNMS.
1245 (backprop::process_assign_use): Remove FMA_EXPR handling.
1246 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1247 (gen_hsa_fma): New function.
1248 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1249 IFN_FNMA and IFN_FNMS.
1250 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1251 * gimple-fold.h (follow_all_ssa_edges): Declare.
1252 * gimple-fold.c (follow_all_ssa_edges): New function.
1253 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1254 gimple_build interface and use follow_all_ssa_edges to fold the result.
1255 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1256 instead of checking for optabs directly.
1257 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1258 rather than FMA_EXPRs.
1259 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1260 call to IFN_FMA instead of an FMA_EXPR.
1262 2018-05-17 Jim Wilson <jimw@sifive.com>
1264 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1265 sign extended promoted subreg, and the range does not have the sign bit
1266 set, then do a sign extend.
1268 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1269 test, check for sign extended subreg and/or constant operands, and
1270 do a sign extend in that case.
1272 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1274 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1275 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1277 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1278 Change logics_shift_reg to logics_shift_imm.
1279 (thunderx2t99_fp_loadpair_basic): Delete.
1280 (thunderx2t99_fp_storepair_basic): Delete.
1281 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1282 (thunderx2t99_asimd_polynomial): Delete.
1283 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1284 and neon_fp_mul_d_scalar_q.
1285 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1286 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1287 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1288 (thunderx2t99_asimd_lut): Add missing tbl types.
1289 (thunderx2t99_asimd_ext): Delete.
1290 (thunderx2t99_asimd_load1_1_mult): Delete.
1291 (thunderx2t99_asimd_load1_2_mult): Delete.
1292 (thunderx2t99_asimd_load1_ldp): New.
1293 (thunderx2t99_asimd_load1): New.
1294 (thunderx2t99_asimd_load2): Add missing *load2* types.
1295 (thunderx2t99_asimd_load3): New.
1296 (thunderx2t99_asimd_load4): New.
1297 (thunderx2t99_asimd_store1_1_mult): Delete.
1298 (thunderx2t99_asimd_store1_2_mult): Delete.
1299 (thunderx2t99_asimd_store2_mult): Delete.
1300 (thunderx2t99_asimd_store2_onelane): Delete.
1301 (thunderx2t99_asimd_store_stp): New.
1302 (thunderx2t99_asimd_store1): New.
1303 (thunderx2t99_asimd_store2): New.
1304 (thunderx2t99_asimd_store3): New.
1305 (thunderx2t99_asimd_store4): New.
1307 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1309 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1310 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1312 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1313 Segher Boessenkool <segher@kernel.crashing.org>
1316 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1319 2018-05-17 Richard Biener <rguenther@suse.de>
1321 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1322 for pruning loop and prune defs feeding only already visited PHIs.
1324 2018-05-17 Richard Biener <rguenther@suse.de>
1326 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1328 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1329 Richard Biener <rguenther@suse.de>
1331 PR tree-optimization/85793
1332 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1333 for VMAT_ELEMENTWISE.
1335 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1337 * internal-fn.h (lookup_internal_fn): Declare
1338 * internal-fn.c (lookup_internal_fn): New function.
1339 * gimple.c (gimple_build_call_from_tree): Handle calls to
1341 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1342 internal function names.
1343 * tree-pretty-print.c (dump_generic_node): Likewise.
1344 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1346 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1348 * gimple-fold.h (gimple_build): Make the function forms take
1349 combined_fn rather than built_in_function.
1350 (gimple_simplify): Likewise.
1351 * gimple-match-head.c (gimple_simplify): Likewise.
1352 * gimple-fold.c (gimple_build): Likewise.
1353 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1354 rather than gimple_build_call_internal.
1355 (get_initial_defs_for_reduction): Likewise.
1356 (vect_create_epilog_for_reduction): Likewise.
1357 (vectorizable_live_operation): Likewise.
1359 2018-05-17 Martin Liska <mliska@suse.cz>
1361 * gimple-ssa-sprintf.c (format_directive): Do not use
1362 space in between 'G_' and '('.
1364 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1367 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1368 even if the mask is not all ones.
1371 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1373 (ix86_gimple_fold_builtin): Likewise.
1376 * config/i386/i386.c: Include tree-vector-builder.h.
1377 (ix86_vector_shift_count): New function.
1378 (ix86_fold_builtin): Fold shift builtins by scalar count.
1379 (ix86_gimple_fold_builtin): Likewise.
1381 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1382 _mm512_setzero): New intrinsics.
1384 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1385 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1387 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1388 code generation for cases where splatting a value is not useful.
1389 * simplify-rtx.c (simplify_ternary_operation): Simplify
1390 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1391 mode to a vec_concat.
1393 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1395 * config.gcc: Support "goldmont-plus".
1396 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1398 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1399 PROCESSOR_GOLDMONT_PLUS.
1400 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1401 (processor_target_table): Add "goldmont-plus".
1402 (PTA_GOLDMONT_PLUS): Define.
1403 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1404 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1405 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1406 (fold_builtin_cpu): Add "goldmont-plus".
1407 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1408 (ix86_option_override_internal): Add "goldmont-plus".
1409 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1410 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1411 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1412 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1414 2018-05-17 Richard Biener <rguenther@suse.de>
1416 PR tree-optimization/85757
1417 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1418 remove defs that only feed that PHI from further processing.
1420 2018-05-16 Jim Wilson <jimw@sifive.com>
1422 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1424 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1426 2018-05-16 Mark Wielaard <mark@klomp.org>
1428 * dwarf2out.c (count_index_strings): New function.
1429 (output_indirect_strings): Call count_index_strings and generate
1430 header for dwarf_version >= 5.
1432 2018-05-16 Mark Wielaard <mark@klomp.org>
1434 * dwarf2out.c (dwarf_FORM): New function.
1435 (set_indirect_string): Use dwarf_FORM.
1436 (reset_indirect_string): Likewise.
1437 (size_of_die): Likewise.
1438 (value_format): Likewise.
1439 (output_die): Likewise.
1440 (add_skeleton_AT_string): Likewise.
1441 (output_macinfo_op): Likewise.
1442 (index_string): Likewise.
1443 (output_index_string_offset): Likewise.
1444 (output_index_string): Likewise.
1445 (count_index_strings): Likewise.
1447 2018-05-16 Carl Love <cel@us.ibm.com>
1449 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1450 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1452 2018-05-16 Martin Jambor <mjambor@suse.cz>
1454 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1455 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1457 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1459 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1460 (fnma<mode>4): Likewise.
1461 (fms<mode>4): Likewise.
1462 (fnms<mode>4): Likewise.
1463 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1464 (aarch64_fnma<mode>4): Likewise.
1465 (aarch64_fms<mode>4): Likewise.
1466 (aarch64_fnms<mode>4): Likewise.
1467 (aarch64_fnmadd<mode>4): Likewise.
1469 2018-05-16 Jason Merrill <jason@redhat.com>
1471 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1473 2018-05-16 Richard Biener <rguenther@suse.de>
1475 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1476 (dump_stmt_cost): Declare.
1477 (add_stmt_cost): Dump cost we add.
1478 (add_stmt_costs): New function.
1479 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1481 (vect_analyze_stmt): Adjust prototype.
1482 (vectorizable_condition): Likewise.
1483 (vectorizable_live_operation): Likewise.
1484 (vectorizable_reduction): Likewise.
1485 (vectorizable_induction): Likewise.
1486 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1487 cost vector to pass to vectorizable_ and record afterwards.
1488 (vect_model_reduction_cost): Take cost vector argument and adjust.
1489 (vect_model_induction_cost): Likewise.
1490 (vectorizable_reduction): Likewise.
1491 (vectorizable_induction): Likewise.
1492 (vectorizable_live_operation): Likewise.
1493 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1494 SLP_TREE_NUMBER_OF_VEC_STMTS.
1495 (vect_analyze_slp_cost_1): Remove.
1496 (vect_analyze_slp_cost): Likewise.
1497 (vect_slp_analyze_node_operations): Take visited args and
1498 a target cost vector. Avoid processing already visited stmt sets.
1499 (vect_slp_analyze_operations): Use a local cost vector to gather
1500 costs and register those of non-discarded instances.
1501 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1502 (vect_schedule_slp_instance): Remove copying of
1503 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1505 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1506 adding cost. Record cost entry location.
1507 (vect_prologue_cost_for_slp_op): Function to compute cost of
1508 a constant or invariant generated for SLP vect in the prologue,
1509 split out from vect_analyze_slp_cost_1.
1510 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1511 (vect_model_promotion_demotion_cost): Likewise.
1512 (vect_model_store_cost): Likewise, make static.
1513 (vect_model_load_cost): Likewise.
1514 (vectorizable_bswap): Add cost vector arg and adjust.
1515 (vectorizable_call): Likewise.
1516 (vectorizable_simd_clone_call): Likewise.
1517 (vectorizable_conversion): Likewise.
1518 (vectorizable_assignment): Likewise.
1519 (vectorizable_shift): Likewise.
1520 (vectorizable_operation): Likewise.
1521 (vectorizable_store): Likewise.
1522 (vectorizable_load): Likewise.
1523 (vectorizable_condition): Likewise.
1524 (vectorizable_comparison): Likewise.
1525 (can_vectorize_live_stmts): Likewise.
1526 (vect_analyze_stmt): Likewise.
1527 (vect_transform_stmt): Adjust calls to vectorizable_*.
1528 * tree-vectorizer.c: Include gimple-pretty-print.h.
1529 (dump_stmt_cost): New function.
1531 2018-05-16 Richard Biener <rguenther@suse.de>
1533 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1534 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1535 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1536 (check_name): New callback.
1537 (dse_classify_store): Track cycles via a visited bitmap of PHI
1538 defs and simplify handling of in-loop and across loop dead stores
1539 and properly fail for loop-variant refs. Handle byte-tracking with
1540 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1543 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1545 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1546 (vect_get_mask_type_for_stmt): Likewise.
1547 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1549 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1550 to determine the statement's vector type and the vector type that
1551 should be used for calculating nunits. Deal with cases in which
1552 the type has to be deferred.
1553 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1554 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1555 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1556 (vect_determine_vf_for_stmt): New functions, split out from...
1557 (vect_determine_vectorization_factor): ...here.
1558 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1559 (vect_get_mask_type_for_stmt): New functions, split out from
1560 vect_determine_vectorization_factor.
1562 2018-05-16 Richard Biener <rguenther@suse.de>
1564 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1565 verify the [VEC_]COND_EXPR embedded comparison.
1567 2018-05-15 Martin Sebor <msebor@redhat.com>
1569 PR tree-optimization/85753
1570 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1571 RECORD_TYPE in addition to ARRAY_TYPE.
1573 2018-05-15 Martin Sebor <msebor@redhat.com>
1576 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1578 2018-05-15 Richard Biener <rguenther@suse.de>
1580 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1581 add by_clobber_p one. Change algorithm to collect all defs
1582 representing uses we need to walk and try reducing them to
1583 a single one before failing.
1584 (dse_dom_walker::dse_optimize_stmt): Adjust.
1586 2018-05-13 Mark Wielaard <mark@klomp.org>
1588 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1589 (size_of_loc_descr): Likewise.
1590 (output_loc_operands): Likewise.
1591 (output_loc_operands_raw): Likewise.
1592 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1593 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1594 (hash_loc_operands): Likewise.
1595 (compare_loc_operands): Likewise.
1597 2018-05-14 Mark Wielaard <mark@klomp.org>
1599 * dwarf2out.c (count_index_addrs): New function.
1600 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1602 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1604 PR tree-optimization/83648
1605 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1606 return value as malloc candidate.
1608 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1611 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1612 as true in call to suggest_attribute.
1614 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1616 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1617 -mreadonly-in-sdata.
1619 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1621 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1623 (aarch64_crypto_aesd_fused): Likewise.
1625 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1627 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1628 (movsi_aarch64): Likewise.
1629 (load_pairsi): Likewise.
1630 (load_pairdi): Likewise.
1631 (store_pairsi): Likewise.
1632 (store_pairdi): Likewise.
1633 (load_pairsf): Likewise.
1634 (load_pairdf): Likewise.
1635 (store_pairsf): Likewise.
1636 (store_pairdf): Likewise.
1637 (zero_extend): Likewise.
1638 (trunc): Swap alternatives.
1639 (fcvt_target): Add '?' to prefer w over r.
1641 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1644 * config/i386/i386.md: Disallow non-commutative arithmetics in
1645 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1646 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1647 in the peephole2 before it.
1649 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1651 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1652 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1653 (ix86_handle_option): Handle -mcldemote.
1654 * config.gcc: New header.
1655 * config/i386/cldemoteintrin.h: New file.
1656 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1657 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1659 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1660 OPTION_MASK_ISA_CLDEMOTE.
1661 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1662 (ix86_valid_target_attribute_inner_p): Ditto.
1663 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1664 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1665 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1666 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1667 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1669 * config/i386/i386.opt: Add -mcldemote.
1670 * config/i386/x86intrin.h: New header.
1671 * doc/invoke.texi: Add -mcldemote.
1673 2018-05-14 Richard Biener <rguenther@suse.de>
1675 * doc/match-and-simplify.texi: Adjust :s documentation.
1677 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1679 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1680 intended memcpy size.
1681 (REORDER_45): Likewise.
1683 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1685 * sort.cc: New file.
1686 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1687 * vec.c (qsort_chk): Use gcc_qsort.
1688 * Makefile.in (OBJS-libcommon): Add sort.o.
1689 (build/sort.o): New target. Use it...
1690 (BUILD_RTL): ... here, and...
1691 (build/gencfn-macros): ... here, and...
1692 (build/genmatch): ... here.
1694 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1695 Chung-Ju Wu <jasonwucj@gmail.com>
1697 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1698 * config/nds32/nds32-graywolf.md: New file.
1699 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1700 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1702 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1703 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1704 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1705 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1706 * config/nds32/pipelines.md: Include n15 settings.
1708 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1709 Chung-Ju Wu <jasonwucj@gmail.com>
1711 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1712 * config/nds32/nds32-n13.md: New file.
1713 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1714 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1716 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1717 * config/nds32/nds32.md (pipeline_model): Add n13.
1718 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1719 * config/nds32/pipelines.md: Include n13 settings.
1721 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1722 Chung-Ju Wu <jasonwucj@gmail.com>
1724 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1725 * config/nds32/nds32-n10.md: New file.
1726 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1727 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1729 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1730 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1731 * config/nds32/nds32.md (pipeline_model): Add n10.
1732 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1733 * config/nds32/pipelines.md: Include n10 settings.
1735 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1736 Kito Cheng <kito.cheng@gmail.com>
1737 Chung-Ju Wu <jasonwucj@gmail.com>
1739 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1740 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1741 Add enum values for DSP extension instructions.
1742 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1744 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1745 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1747 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1748 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1749 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1750 * config/nds32/nds32-intrinsic.md: Likewise.
1751 * config/nds32/nds32_intrinsic.h: Likewise.
1752 * config/nds32/nds32-md-auxiliary.c: Likewise.
1753 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1754 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1755 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1756 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1757 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1758 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1759 TYPE_DMAC in switch statement.
1760 * config/nds32/nds32.c: New checking and implementation for DSP
1761 extension instructions.
1762 * config/nds32/nds32.h: Likewise.
1763 * config/nds32/nds32.md: Likewise.
1764 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1765 * config/nds32/predicates.md: Implement new predicates for DSP
1768 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1770 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1771 Reformat alternatives and attributes so it is easier to identify
1772 which constraints/attributes go with which instruction.
1773 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1774 (mov<mode>_softfloat32, FMOVE64): Likewise.
1775 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1776 (mov<mode>_softfloat64, FMOVE64): Likewise.
1778 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1780 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1782 (Basic PowerPC Built-in Functions): The new name of the
1783 subsection previously known as "PowerPC Built-in Functions".
1784 (Basic PowerPC Built-in Functions Available on all Configurations):
1786 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1787 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1788 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1789 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1791 2018-05-11 Martin Jambor <mjambor@suse.cz>
1794 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1797 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1800 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1802 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1804 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1805 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1806 (ix86_handle_option): Handle -mwaitpkg.
1807 * config.gcc: New header.
1808 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1809 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1810 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1812 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1813 OPTION_MASK_ISA_WAITPKG.
1814 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1815 (ix86_option_override_internal): Add PTA_WAITPKG.
1816 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1817 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1818 IX86_BUILTIN_TPAUSE.
1819 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1820 __builtin_ia32_umwait and __builtin_ia32_tpause.
1821 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1822 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1823 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1824 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1825 UNSPECV_TPAUSE): New.
1826 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1827 * config/i386/i386.opt: Add -mwaitpkg.
1828 * config/i386/waitpkgintrin.h: New file.
1829 * config/i386/x86intrin.h: New header.
1830 * doc/invoke.texi: Add -mwaitpkg.
1832 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1835 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1837 (cortex-m0): Use armv6s-m isa.
1838 (cortex-m0plus): Likewise.
1839 (cortex-m1): Likewise.
1840 (cortex-m0.small-multiply): Likewise.
1841 (cortex-m0plus.small-multiply): Likewise.
1842 (cortex-m1.small-multiply): Likewise.
1844 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1845 Jakub Jelinek <jakub@redhat.com>
1847 PR tree-optimization/85692
1848 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1849 source permute as well.
1851 2018-05-11 Martin Liska <mliska@suse.cz>
1854 * doc/extend.texi: Document LLVM style format for no_sanitize
1857 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1859 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1860 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1861 (mode_supports_vsx_dform_quad): Likewise.
1862 (mode_supports_vmx_dform): Move these functions to be next to the
1863 other mode_supports functions.
1864 (mode_supports_dq_form): Likewise.
1865 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1866 mode_supports_dq_form.
1867 (reg_offset_addressing_ok_p): Likewise.
1868 (offsettable_ok_by_alignment): Likewise.
1869 (rs6000_legitimate_offset_address_p): Likewise.
1870 (legitimate_lo_sum_address_p): Likewise.
1871 (rs6000_legitimize_address): Likewise.
1872 (rs6000_legitimize_reload_address): Likewise.
1873 (rs6000_secondary_reload_inner): Likewise.
1874 (rs6000_preferred_reload_class): Likewise.
1875 (rs6000_output_move_128bit): Likewise.
1877 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1879 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1880 Generate SImode target register for null target.
1881 <case IX86_BUILTIN_XGETBV>: Ditto.
1882 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1883 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1885 2018-05-10 Carl Love <cel@us.ibm.com>
1887 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1888 dcbtt and dcbtstt if operands[2] is 0.
1890 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1893 * config/i386/sse.md (usadv64qi): New expander.
1895 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1897 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1898 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1899 -maltivec=be support.
1900 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1901 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1902 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1903 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1904 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1905 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1906 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1907 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1908 altivec_vsumsws): Adjust.
1909 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1910 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1911 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1913 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1914 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1915 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1916 (altivec_lve<VI_char>x): Delete expand.
1917 (*altivec_lve<VI_char>x_internal): Rename to...
1918 (altivec_lve<VI_char>x): ... this.
1919 (altivec_lvxl_<mode>): Delete expand.
1920 (*altivec_lvxl_<mode>_internal): Rename to ...
1921 (altivec_lvxl_<mode>): ... this.
1922 (altivec_stvxl_<mode>): Delete expand.
1923 (*altivec_stvxl_<mode>_internal): Rename to ...
1924 (altivec_stvxl_<mode>): ... this.
1925 (altivec_stve<VI_char>x): Delete expand.
1926 (*altivec_stve<VI_char>x_internal): Rename to ...
1927 (altivec_stve<VI_char>x): ... this.
1928 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1929 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1930 reduc_plus_scal_<mode>): Adjust.
1931 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1933 (rs6000_cpu_cpp_builtins): Adjust.
1934 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1935 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1936 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1937 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1938 -maltivec=be support.
1939 (rs6000_split_vec_extract_var): Adjust.
1940 (rs6000_split_v4si_init): Adjust.
1941 (swap_selector_for_mode): Delete.
1942 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1943 altivec_expand_stvex_be): Delete.
1944 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1945 -maltivec=be support.
1946 (rs6000_gimple_fold_builtin): Ditto.
1947 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1949 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1950 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1951 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1952 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1953 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1954 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1955 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1956 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1957 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1958 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1959 anonymous split): Adjust.
1960 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1961 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1963 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1965 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1966 when --with-gxx-include-dir is also specified.
1967 * configure: Regenerate.
1969 2018-05-09 Jim Wilson <jimw@sifive.com>
1972 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1973 * config/riscv/t-withmultilib: New.
1974 * config/riscv/withmultilib.h: New.
1975 * doc/install.texi: Document RISC-V --with-multilib-list support.
1977 2018-05-09 Richard Biener <rguenther@suse.de>
1979 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1981 (vect_bb_vectorization_profitable_p): Adjust. Compute
1982 actual scalar cost using the cost vector and the add_stmt_cost
1985 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1987 PR rtl-optimization/85645
1988 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1989 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1991 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1993 PR rtl-optimization/85645
1994 * shrink-wrap.c (spread_components): Return a boolean saying if
1995 anything was changed.
1996 (try_shrink_wrapping_separate): Iterate spread_components until
1997 nothing changes anymore.
1999 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2001 PR rtl-optimization/85645
2002 * regrename.c (build_def_use): Also kill the chains that include the
2003 destination of a REG_CFA_REGISTER note.
2005 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2007 PR rtl-optimization/85645
2008 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2009 insn that has a REG_CFA_REGISTER note.
2011 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2013 * cfgexpand.c (expand_clobber): New function.
2014 (expand_gimple_stmt_1): Use it.
2015 * tree-vect-stmts.c (vect_clobber_variable): New function,
2017 (vectorizable_simd_clone_call): ...here.
2018 (vectorizable_store): Emit a clobber either side of an
2019 IFN_STORE_LANES sequence.
2020 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2022 2018-05-09 Tom de Vries <tom@codesourcery.com>
2025 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2026 (define_insn "trap_if_false"): Add exit after trap.
2028 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2030 PR rtl-optimization/85638
2031 * bb-reorder.c: Include common/common-target.h.
2032 (create_forwarder_block): New function extracted from...
2033 (fix_up_crossing_landing_pad): ...here. Rename into...
2034 (dw2_fix_up_crossing_landing_pad): ...this.
2035 (sjlj_fix_up_crossing_landing_pad): New function.
2036 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2037 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2038 from both partitions and exit the loop after one iteration.
2040 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2043 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2045 (Basic PowerPC Built-in Functions): The new name of the
2046 subsection previously known as "PowerPC Built-in Functions".
2047 (Basic PowerPC Built-in Functions Available on all Configurations):
2049 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2051 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2053 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2055 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2058 2018-05-08 Jim Wilson <jimw@sifive.com>
2060 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2061 (LD_EMUL_SUFFIX): New.
2062 (LINK_SPEC): Use it.
2064 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2066 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2068 (Basic PowerPC Built-in Functions): The new name of the
2069 subsection previously known as "PowerPC Built-in Functions".
2070 (Basic PowerPC Built-in Functions Available on all Configurations):
2072 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2074 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2076 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2078 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2081 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2084 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2085 after cmpelim optimization.
2087 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2089 * config.gcc: Support "goldmont".
2090 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2091 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2093 * config/i386/i386.c (m_GOLDMONT): Define.
2094 (processor_target_table): Add "goldmont".
2095 (PTA_GOLDMONT): Define.
2096 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2097 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2098 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2099 (fold_builtin_cpu): Add "goldmont".
2100 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2101 (ix86_option_override_internal): Add "goldmont".
2102 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2103 (processor_type): Add PROCESSOR_GOLDMONT.
2104 * config/i386/i386.md: Add CPU "glm".
2105 * config/i386/glm.md: New file.
2106 * config/i386/x86-tune.def: Add m_GOLDMONT.
2107 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2109 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2112 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2114 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2115 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2116 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2119 * config/i386/i386.c (ix86_fold_builtin): Handle
2120 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2123 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2124 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2126 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2129 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2130 (check_arch): Likewise.
2131 (check_fpu): Return the result rather than printing it.
2132 (end arch): Fix operator precedence.
2133 (end cpu): Likewise.
2134 (END): Print the result from check_fpu.
2136 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2137 Alan Hayward <alan.hayward@arm.com>
2138 David Sherwood <david.sherwood@arm.com>
2140 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2141 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2142 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2143 (*fcmuo<mode>_and): New patterns.
2145 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2147 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2148 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2149 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2150 (cmp_op, sve_imm_con): New code attributes.
2151 (SVE_COND_INT_CMP, imm_con): Delete.
2152 (cmp_op): Remove above unspecs from int attribute.
2153 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2155 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2156 comparison-specific unspecs.
2157 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2158 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2159 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2160 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2161 (*vec_fcm<cmp_op><mode>): Rename to...
2162 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2163 (*vec_fcmuo<mode>): Rename to...
2164 (*fcmuo<mode>): ...this and adjust likewise.
2165 (*pred_fcm<cmp_op><mode>): New pattern.
2166 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2167 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2169 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2171 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2172 (aarch64_emit_sve_predicated_cond): New function.
2173 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2174 (aarch64_emit_unspec_cond_or): Replace with...
2175 (aarch64_emit_sve_or_conds): ...this new function. Use
2176 aarch64_emit_sve_ptrue_op for the individual comparisons and
2177 aarch64_emit_binop to OR them together.
2178 (aarch64_emit_inverted_unspec_cond): Replace with...
2179 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2180 aarch64_emit_sve_ptrue_op for the comparison and
2181 aarch64_emit_unop to invert the result.
2182 (aarch64_expand_sve_vec_cmp_float): Update after the above
2183 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2185 2018-05-07 Nathan Sidwell <nathan@acm.org>
2187 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2188 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2189 (Backwards Compatibility): Likewise.
2191 2018-05-07 Luis Machado <luis.machado@linaro.org>
2195 2018-05-07 Luis Machado <luis.machado@linaro.org>
2197 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2198 <prefetch_dynamic_strides>: New const bool field.
2199 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2200 prefetch_dynamic_strides.
2201 (exynosm1_prefetch_tune): Likewise.
2202 (thunderxt88_prefetch_tune): Likewise.
2203 (thunderx_prefetch_tune): Likewise.
2204 (thunderx2t99_prefetch_tune): Likewise.
2205 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2206 (aarch64_override_options_internal): Update to set
2207 PARAM_PREFETCH_DYNAMIC_STRIDES.
2208 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2209 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2210 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2211 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2212 prefetch-dynamic-strides setting.
2214 2018-05-07 Luis Machado <luis.machado@linaro.org>
2216 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2217 <minimum_stride>: New const int field.
2218 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2219 minimum_stride field.
2220 (exynosm1_prefetch_tune): Likewise.
2221 (thunderxt88_prefetch_tune): Likewise.
2222 (thunderx_prefetch_tune): Likewise.
2223 (thunderx2t99_prefetch_tune): Likewise.
2224 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2225 (aarch64_override_options_internal): Update to set
2226 PARAM_PREFETCH_MINIMUM_STRIDE.
2227 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2228 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2229 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2230 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2231 stride is constant and is below the minimum stride threshold.
2233 2018-05-07 Luis Machado <luis.machado@linaro.org>
2235 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2238 2018-05-07 Luis Machado <luis.machado@linaro.org>
2240 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2241 <prefetch_dynamic_strides>: New const bool field.
2242 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2243 prefetch_dynamic_strides.
2244 (exynosm1_prefetch_tune): Likewise.
2245 (thunderxt88_prefetch_tune): Likewise.
2246 (thunderx_prefetch_tune): Likewise.
2247 (thunderx2t99_prefetch_tune): Likewise.
2248 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2249 (aarch64_override_options_internal): Update to set
2250 PARAM_PREFETCH_DYNAMIC_STRIDES.
2251 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2252 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2253 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2254 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2255 prefetch-dynamic-strides setting.
2257 2018-05-07 Luis Machado <luis.machado@linaro.org>
2259 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2260 <minimum_stride>: New const int field.
2261 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2262 minimum_stride field.
2263 (exynosm1_prefetch_tune): Likewise.
2264 (thunderxt88_prefetch_tune): Likewise.
2265 (thunderx_prefetch_tune): Likewise.
2266 (thunderx2t99_prefetch_tune): Likewise.
2267 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2268 (aarch64_override_options_internal): Update to set
2269 PARAM_PREFETCH_MINIMUM_STRIDE.
2270 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2271 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2272 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2273 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2274 stride is constant and is below the minimum stride threshold.
2276 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2279 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2280 the type is addressable. Don't force op into register if it has
2283 2018-05-05 Roland McGrath <mcgrathr@google.com>
2286 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2287 any section for which we don't know a specific type it should have,
2288 regardless of name. Previously this was done only for the exact
2289 names ".init_array", ".fini_array", and ".preinit_array".
2290 (default_elf_asm_named_section): Add comment about
2291 relationship with default_section_type_flags and SECTION_NOTYPE.
2292 (get_section): Don't consider it a type conflict if one side has
2293 SECTION_NOTYPE and the other doesn't, as long as neither has the
2294 SECTION_BSS et al used in the default_section_type_flags logic.
2296 2018-05-05 Tom de Vries <tom@codesourcery.com>
2299 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2300 (workaround_barsyncs): New function.
2301 (nvptx_reorg): Use workaround_barsyncs.
2302 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2303 (define_expand "nvptx_membar_cta"): New define_expand.
2304 (define_insn "*nvptx_membar_cta"): New insn.
2306 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2308 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2309 To improve optimization opportunities.
2310 * builtin-types.def: The new needed builtin types for the above.
2312 2018-05-04 Richard Biener <rguenther@suse.de>
2314 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2315 * gimple-ssa-store-merging.c
2316 (imm_store_chain_info::output_merged_store): Remove redundant create,
2317 release split_store vector contents on failure.
2318 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2319 scalar stmt vector on cache hit.
2321 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2323 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2325 * config.gcc (powerpc-xilinx-eabi*): Remove.
2326 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2328 (fusion_addis_mem_combo_load): Ditto.
2329 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2331 (rs6000_cpu_cpp_builtins): Ditto.
2332 * config/rs6000/rs6000-linux.c
2333 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2334 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2335 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2337 (rs6000_setup_reg_addr_masks): Ditto.
2338 (rs6000_init_hard_regno_mode_ok): Ditto.
2339 (rs6000_option_override_internal): Ditto.
2340 (legitimate_lo_sum_address_p): Ditto.
2341 (rs6000_legitimize_address): Ditto.
2342 (rs6000_legitimize_reload_address): Ditto.
2343 (rs6000_legitimate_address_p): Ditto.
2344 (abi_v4_pass_in_fpr): Ditto.
2345 (setup_incoming_varargs): Ditto.
2346 (rs6000_gimplify_va_arg): Ditto.
2347 (rs6000_split_multireg_move): Ditto.
2348 (rs6000_savres_strategy): Ditto.
2349 (rs6000_emit_prologue_components): Ditto.
2350 (rs6000_emit_epilogue_components): Ditto.
2351 (rs6000_emit_prologue): Ditto.
2352 (rs6000_emit_epilogue): Ditto.
2353 (rs6000_elf_file_end): Ditto.
2354 (rs6000_function_value): Ditto.
2355 (rs6000_libcall_value): Ditto.
2356 * config/rs6000/rs6000.h: Ditto.
2357 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2358 (TARGET_MINMAX): ... this. New.
2359 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2360 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2361 (*movsi_internal1_single): Delete.
2362 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2363 mfpu=, mxilinx-fpu): Delete.
2364 * config/rs6000/singlefp.h: Delete.
2365 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2366 * config/rs6000/t-rs6000: Ditto.
2367 * config/rs6000/t-xilinx: Delete.
2368 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2369 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2370 (VStype_simple): Delete.
2371 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2372 * config/rs6000/xfpu.h: Delete.
2373 * config/rs6000/xfpu.md: Delete.
2374 * config/rs6000/xilinx.h: Delete.
2375 * config/rs6000/xilinx.opt: Delete.
2376 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2377 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2379 2018-05-04 Tom de Vries <tom@codesourcery.com>
2382 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2385 2018-05-04 Richard Biener <rguenther@suse.de>
2388 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2389 (expand_complex_div_wide): Likewise.
2390 (expand_complex_operations_1): Likewise.
2391 (expand_complex_libcall): Preserve EH info of the original stmt.
2392 (tree_lower_complex): Handle removed blocks.
2393 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2394 on complex multiplication and division libcall builtins.
2396 2018-05-04 Richard Biener <rguenther@suse.de>
2399 * fold-const.c (negate_expr_p): Restrict negation of operand
2400 zero of a division to when we know that can happen without
2402 (fold_negate_expr_1): Likewise.
2404 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2407 * real.h (real_nextafter): Declare.
2408 * real.c (real_nextafter): New function.
2409 * fold-const-call.c (fold_const_nextafter): New function.
2410 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2411 CASE_CFN_NEXTTOWARD.
2412 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2413 even when arg1_mode is different from arg0_mode.
2415 2018-05-03 Nathan Sidwell <nathan@acm.org>
2417 * doc/extend.texi (Deprecated Features): Remove
2419 (Backwards Compatibility): Likewise.
2420 * doc/invoke.texi (C++ Language Options): Likewise.
2421 (C++ Dialect Options): Likewise.
2423 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2426 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2427 _mm512_mask_mullox_epi64): New intrinsics.
2429 2018-05-03 Tom de Vries <tom@codesourcery.com>
2432 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2433 dump files): Add offload-tree.
2435 2018-05-03 Richard Biener <rguenther@suse.de>
2437 PR tree-optimization/85615
2438 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2439 to loops not nested in BBs loop father to avoid creating multi-entry
2442 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2444 PR tree-optimization/70291
2445 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2446 arguments. Change return type to tree. Emit libcall as a new
2447 statement rather than replacing existing one when inplace_p is true.
2448 (expand_complex_multiplication_components): New function.
2449 (expand_complex_multiplication): Expand floating-point complex
2450 multiplication using the above.
2451 (expand_complex_division): Rename inner_type parameter to type.
2452 Update expand_complex_libcall call-site.
2453 (expand_complex_operations_1): Update expand_complex_multiplication
2454 and expand_complex_division call-sites.
2456 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2459 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2460 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2461 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2462 the highest significant bit of the shift count mask is clear. In
2463 check whether and[sq]i3 is needed verify that all significant bits
2464 of the shift count other than the highest are set.
2466 2018-05-02 Tom de Vries <tom@codesourcery.com>
2469 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2470 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2471 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2472 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2473 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2474 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2475 __builtin_goacc_parlevel_size.
2477 2018-05-02 Richard Biener <rguenther@suse.de>
2479 PR tree-optimization/85597
2480 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2481 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2484 2018-05-02 Tom de Vries <tom@codesourcery.com>
2487 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2488 dump files): Add ltrans-tree.
2490 2018-05-02 Tom de Vries <tom@codesourcery.com>
2493 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2494 dump files): Add wpa-ipa.
2496 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2498 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2499 powerpc*-*-linux*paired* target.
2500 * config/rs6000/750cl.h: Delete.
2501 * config/rs6000/paired.h: Delete.
2502 * config/rs6000/paired.md: Delete.
2503 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2505 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2506 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2507 comment. Remove paired float support.
2508 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2509 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2511 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2512 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2514 * config/rs6000/rs6000.c: Remove paired float support.
2515 (paired_expand_vector_init, paired_expand_vector_move,
2516 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2517 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2518 paired_expand_builtin, paired_expand_predicate_builtin,
2519 paired_init_builtins): Delete.
2520 * config/rs6000/rs6000.h: Remove paired float support.
2521 * config/rs6000/rs6000.md: Remove paired float support.
2522 (move_from_CR_ov_bit): Delete.
2523 * config/rs6000/rs6000.opt (mpaired): Delete.
2524 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2525 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2527 2018-05-02 Richard Biener <rguenther@suse.de>
2530 * gimplify.c (gimplify_save_expr): When in SSA form allow
2531 SAVE_EXPRs to compute to SSA vars.
2533 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2536 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2537 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2538 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2539 clobber operands[2], instead use a new pseudo. Formatting fixes.
2541 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2543 PR tree-optimization/85586
2544 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2545 exit early for statements in the same group if the accesses are
2548 2018-05-02 Tom de Vries <tom@codesourcery.com>
2551 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2554 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2556 PR tree-optimization/85143
2557 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2559 2018-05-01 Tom de Vries <tom@codesourcery.com>
2562 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2563 not found" error message.
2565 2018-05-01 Tom de Vries <tom@codesourcery.com>
2568 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2569 * vec.c (test_ordered_remove_if): New function.
2570 (vec_c_tests): Call test_ordered_remove_if.
2571 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2572 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2573 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2574 VEC_ORDERED_REMOVE_IF.
2576 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2578 PR tree-optimization/82665
2579 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2580 pointer subtraction where arguments come from a memchr call.
2582 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2584 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2585 --push-state --as-needed and --pop-state instead of --as-needed and
2586 --no-as-needed if ld supports it.
2587 * configure: Regenerated.
2590 * doc/install.texi2html: Replace _002d with - and _002a with * in
2591 generated html files using sed.
2593 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2596 * gcc-rich-location.c (blank_line_before_p): New function.
2597 (use_new_line): New function.
2598 (gcc_rich_location::add_fixit_insert_formatted): New function.
2599 * gcc-rich-location.h
2600 (gcc_rich_location::add_fixit_insert_formatted): New function.
2602 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2604 * selftest.c (assert_streq): Rename "expected" and "actual" to
2605 "val1" and "val2". Extend NULL-handling to cover both inputs
2606 symmetrically, while still requiring both to be non-NULL for a pass.
2607 * selftest.h (assert_streq): Rename "expected" and "actual" to
2609 (ASSERT_EQ): Likewise.
2610 (ASSERT_EQ_AT): Likewise.
2611 (ASSERT_KNOWN_EQ): Likewise.
2612 (ASSERT_KNOWN_EQ_AT): Likewise.
2613 (ASSERT_NE): Likewise.
2614 (ASSERT_MAYBE_NE): Likewise.
2615 (ASSERT_MAYBE_NE_AT): Likewise.
2616 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2617 the assertion to pass.
2618 (ASSERT_STREQ_AT): Likewise.
2620 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2622 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2623 interaction with -pie.
2625 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2627 * selftest.h: Fix alphabetization of per-source-file selftest
2630 2018-04-30 Jason Merrill <jason@redhat.com>
2632 PR c++/61982 - dead stores to destroyed objects.
2633 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2636 2018-04-30 Jason Merrill <jason@redhat.com>
2638 * tree.c (build_clobber): New.
2639 * tree.h: Declare it.
2640 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2642 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2644 * diagnostic-show-locus.c (layout::layout): Update for
2645 location_get_source_line returning a char_span.
2646 (struct char_span): Move to input.h.
2647 (struct correction): Update for fields in char_span becoming
2649 (struct source_line): Update for location_get_source_line
2650 returning a char_span.
2651 (layout::print_line): Likewise.
2652 * edit-context.c (edited_file::print_content): Likewise.
2653 (edited_file::print_diff_hunk): Likewise.
2654 (edited_file::print_run_of_changed_lines): Likewise.
2655 (edited_file::get_num_lines): Likewise.
2656 (edited_line::edited_line): Likewise.
2657 * final.c (asm_show_source): Likewise.
2658 * input.c (location_get_source_line): Convert return type
2659 from const char * to char_span, losing the final "line_len"
2661 (dump_location_info): Update for the above.
2662 (get_substring_ranges_for_loc): Likewise. Use a char_span
2663 when handling the literal within the line.
2664 (test_reading_source_line): Update for location_get_source_line
2665 returning a char_span.
2666 * input.h (class char_span): Move here from
2667 diagnostic-show-locus.c, converting from a struct to a class.
2668 Make data members private.
2669 (char_span::operator bool): New.
2670 (char_span::length): New.
2671 (char_span::get_buffer): New.
2672 (char_span::operator[]): New.
2673 (char_span::subspan): Make const.
2674 (char_span::xstrdup): New.
2675 (location_get_source_line): Convert return type from const char *
2676 to char_span, losing the final "line_size" param.
2678 2018-04-30 Jan Hubicka <jh@suse.cz>
2680 * lto-wrapper.c (ltrans_priorities): New static var.
2681 (cmp_priority): New.
2682 (run_gcc): Read priorities and if doing parallel build order
2683 the Makefile by them.
2685 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2687 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2689 2018-04-30 Richard Biener <rguenther@suse.de>
2691 * tree-cfg.c (verify_address): Remove base argument, add
2692 flag whether to check TREE_ADDRESSABLE and do that.
2693 (verify_expr): Remove.
2694 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2695 (verify_gimple_assign_single): Likewise.
2696 (verify_gimple_switch): Likewise.
2697 (verify_expr_location_1): Dereference tp once. Add (disabled)
2698 piece from verify_expr.
2699 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2701 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2703 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2705 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2707 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2708 (small_data_pattern): Likewise.
2709 (arc_rewrite_small_data): Likewise.
2710 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2711 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2712 (get_symbol_alignment): New function.
2713 (legitimate_small_data_address_p): Likewise.
2714 (legitimate_scaled_address): Update, call
2715 legitimate_small_data_address_p.
2716 (output_sdata): New static variable.
2717 (arc_print_operand): Update how we handle small data operands.
2718 (arc_print_operand_address): Likewise.
2719 (arc_legitimate_address_p): Update, use
2720 legitimate_small_data_address_p.
2721 (arc_rewrite_small_data_p): Remove.
2722 (arc_rewrite_small_data_1): Likewise.
2723 (arc_rewrite_small_data): Likewise.
2724 (small_data_pattern): Likewise.
2725 (compact_sda_memory_operand): Update to use
2726 legitimate_small_data_address_p and get_symbol_alignment.
2727 (prepare_move_operands): Don't rewite sdata pattern.
2728 (prepare_extend_operands): Remove.
2729 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2731 (zero_extendqisi2): Likewise.
2732 (zero_extendhisi2): Likewise.
2733 (extendqihi2): Likewise.
2734 (extendqisi2): Likewise.
2735 (extendhisi2): Likewise.
2739 * config/arc/constraints.md (Usd): Change it to memory constraint.
2741 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2743 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2744 as source of std instructions.
2745 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2746 allow 6-bit constants as source for store instructions.
2747 (movdi_insn): Update instruction pattern to allow 6-bit constants
2748 as source for store instructions.
2750 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2752 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2754 2018-04-30 Nathan Sidwell <nathan@acm.org>
2755 Sandra Loosemore <sandra@codesourcery.com>
2757 * dumpfile.c (dump_open): Allow '-' for stdout.
2758 * doc/invoke.texi (Developer Options): Document dump filename
2759 determination early. Document stdin/stdout selection.
2761 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2763 Microblaze Target: PIC data text relative
2765 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2766 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2768 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2769 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2770 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2771 New addressing mode for data-text relative position indepenedent code.
2772 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2773 'ADDRESS_SYMBOLIC_TXT_REL'.
2774 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2775 (microblaze_legitimate_pic_operand): Exclude function calls from
2776 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2777 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2779 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2780 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2781 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2782 for 'address + offset'.
2783 (microblaze_expand_prologue): Add new function prologue call for
2785 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2786 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2787 table in case of TARGET_PIC_DATA_TEXT_REL.
2788 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2789 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2790 Add new macros 'UNSPEC_TEXT',
2791 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2792 + exclude function calls from 'UNSPEC_PLT' in case of data text
2794 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2795 new target hook for generating address diff vector tables in case of
2797 * doc/tm.texi : Regenerate.
2798 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2799 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2800 of addr diff vector generation.
2801 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2802 target hook definition.
2803 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2804 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2805 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2806 option for data text relative.
2808 2018-04-30 Richard Biener <rguenther@suse.de>
2810 * tree-chrec.h (evolution_function_is_constant_p): Remove
2812 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2814 2018-04-30 Richard Biener <rguenther@suse.de>
2817 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2819 2018-04-30 Richard Biener <rguenther@suse.de>
2821 PR tree-optimization/28364
2822 PR tree-optimization/85275
2823 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2824 copying first exit test.
2826 2018-04-28 Mark Wielaard <mark@klomp.org>
2828 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2830 (dwarf_AT): Handle DW_AT_addr_base.
2831 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2833 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2836 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2837 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2838 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2839 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2841 2018-04-28 Richard Biener <rguenther@suse.de>
2843 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2844 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2845 to reflect use. Only add interesting stmts.
2847 2018-04-27 Martin Jambor <mjambor@suse.cz>
2850 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2851 the jump function allows for passing through aggregate values.
2853 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2855 * input.h (in_system_header_at): Convert from macro to inline
2857 (from_macro_expansion_at): Likewise.
2858 (from_macro_definition_at): Likewise.
2860 2018-04-27 Jeff Law <law@redhat.com>
2862 * config.gcc: Mark tile* targets as deprecated/obsolete.
2864 2018-04-27 Richard Biener <rguenther@suse.de>
2866 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2869 2018-04-27 Richard Biener <rguenther@suse.de>
2871 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2873 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2875 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2876 with Yd constraint. Set "preferred_for_speed" attribute from
2877 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2879 (*movdi_internal): Ditto.
2880 (movti_interunit splitters): Remove
2881 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2882 (movdi_interunit splitters): Ditto.
2883 * config/i386/constraints.md (Ye): Remove.
2884 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2886 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2889 * config/aarch64/constraints.md (Usg): Limit to 31.
2892 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2894 PR tree-optimization/85529
2895 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2896 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2897 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2898 zero extension or masking of the MSB bit.
2899 (optimize_range_tests): Add FIRST_BB argument, pass it through
2900 to optimize_range_tests_var_bound.
2901 (maybe_optimize_range_tests, reassociate_bb): Adjust
2902 optimize_range_tests callers.
2904 2018-04-26 Richard Biener <rguenther@suse.de>
2905 Jakub Jelinek <jakub@redhat.com>
2907 * cgraph.h (symbol_table): Just declare debug method here.
2908 * symtab.c (symbol_table::debug): Define.
2910 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2912 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2914 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2916 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2917 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2918 (*movdi_internal): Substitute Yi and Yj constraint with x
2919 and Ym and Yn constraint with y constraint. Update "isa"
2920 attribute and set "preferred_for_speed" attribute from
2921 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2922 (*movsi_internal): Ditto.
2923 (*movdf_internal): Ditto.
2924 (*movsf_internal): Ditto.
2925 (*zero_extendsidi2): Ditto.
2926 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2927 (sse2_loadld): Ditto.
2928 (*vec_extract<ssevecmodelower>_0): Ditto.
2929 (*vec_extractv4si_0_zext_sse4): Ditto.
2930 (vec_concatv2di): Ditto.
2931 (*vec_dup<mode>): Ditto.
2932 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2933 * config/i386/constraints.md (Yi): Remove.
2938 2018-04-26 Nathan Sidwell <nathan@acm.org>
2940 * dumpfile.c (dump_open): New.
2941 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2942 (dump_finish): Detect stdio/stderr by value not name.
2944 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2946 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2948 2018-04-26 Tom de Vries <tom@codesourcery.com>
2951 * config/nvptx/nvptx.c (verify_neutering_jumps)
2952 (verify_neutering_labels): New function
2953 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2955 2018-04-26 Tom de Vries <tom@codesourcery.com>
2958 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2959 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2962 2018-04-26 Richard Biener <rguenther@suse.de>
2963 Tom de Vries <tom@codesourcery.com>
2966 * lto-streamer-out.c (output_function): Fixup loops if required to match
2967 discovery done in the reader.
2969 2018-04-26 Richard Biener <rguenther@suse.de>
2971 PR tree-optimization/85116
2972 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2973 have a loop exit from the single latch predecessor. Remove
2974 case of header with just condition.
2975 (ch_base::copy_headers): Exclude infinite loops from any
2977 (pass_ch::execute): Record exits.
2979 2018-04-26 Richard Biener <rguenther@suse.de>
2981 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2982 prologue cost vector and pass it to vect_get_load_cost.
2983 (vect_get_peeling_costs_all_drs): Likewise.
2984 (vect_peeling_hash_get_lowest_cost): Likewise.
2985 (vect_enhance_data_refs_alignment): Likewise.
2987 2018-04-26 Richard Biener <rguenther@suse.de>
2990 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2991 checking of integer<->pointer conversions.
2992 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2993 sign-/zero-extending pointer types.
2994 (expand_omp_for_static_chunk): Likewise.
2996 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2997 Jean Lee <xiaoyur347@gmail.com>
2999 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3000 (TARGET_ASAN_SHADOW_OFFSET): Define.
3001 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3002 true for -fsanitize=address.
3004 2018-04-25 Mark Wielaard <mark@klomp.org>
3006 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3009 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3011 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3012 than "alu", remove explicit "memory" and "imm_disp" attributes.
3013 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3016 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3017 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3018 gen_lowpart_no_emit.
3020 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3023 * config/i386/i386.c (ix86_expand_builtin): Change memory
3024 operand to XI, extend p0 to Pmode.
3025 * config/i386/i386.md: Change unspec volatile and operand
3026 1 mode to XI, change operand 0 mode to P.
3028 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3030 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3031 GET_MODE_MASK before any checking.
3032 (nds32_can_use_bset_p): Likewise.
3033 (nds32_can_use_btgl_p): Likewise.
3035 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3037 * config/nds32/nds32-doubleword.md: New define_split pattern for
3038 illegal register number.
3040 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3042 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3044 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3046 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3048 2018-04-25 Richard Biener <rguenther@suse.de>
3050 * lto-streamer.h (LTO_major_version): Bump to 8.
3052 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3054 * BASE-VER: Set to 9.0.0.
3056 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3058 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3059 in __abskf2 and __powikf2.
3061 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3064 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3065 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3066 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3067 Use the above on operand 2. Reindent.
3068 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3070 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3073 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3075 * config/i386/i386.opt (mcet): Removed.
3076 * doc/install.texi: Remove -mcet documentation.
3077 * doc/invoke.texi: Likewise.
3079 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3082 * doc/install.texi: Remove -mcet from bootstrap-cet.
3084 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3087 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3088 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3092 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3093 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3094 containing a CONST_VECTOR.
3096 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3098 * doc/install.texi: Update newlib dependency for nvptx.
3100 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3103 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3104 instead of INTVAL when shifting x left.
3106 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3108 PR tree-optimization/85478
3109 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3110 vect_grouped_store_supported for single element vectors.
3112 2018-04-24 Richard Biener <rguenther@suse.de>
3115 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3116 load cost increase to the case of non-constant step.
3118 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3121 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3122 destination if any_malformed_asm.
3124 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3127 * expr.c (store_field): In the bitfield case, if the value comes from
3128 a function call and is returned in registers by means of a PARALLEL,
3129 do not change the mode of the temporary unless BLKmode and VOIDmode.
3131 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3133 PR rtl-optimization/85423
3134 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3135 dependencies to debug insns when the previous insn is non-debug.
3137 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3139 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3140 enums into a single definition.
3141 (fls): Fix predicates and printing.
3144 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3146 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3147 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3148 and short u6 immediate.
3149 (check_if_valid_sleep_operand): Remove.
3150 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3152 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3154 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3155 flag_always_save_lp condition.
3156 * config/nds32/nds32.opt (malways-save-lp): New option.
3158 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3160 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3161 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3162 * config/nds32/nds32.h
3163 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3164 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3166 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3168 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3169 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3171 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3172 Chung-Ju Wu <jasonwucj@gmail.com>
3174 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3176 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3177 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3179 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3181 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3183 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3185 * config/nds32/nds32-protos.h (nds32_data_alignment,
3186 nds32_local_alignment): Declare.
3187 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3188 nds32_local_alignment): New functions.
3189 (TARGET_CONSTANT_ALIGNMENT): Define.
3190 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3192 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3194 * config/nds32/nds32.c
3195 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3196 (TARGET_MODES_TIEABLE_P): Likewise.
3198 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3200 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3203 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3204 Chung-Ju Wu <jasonwucj@gmail.com>
3206 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3207 for unaligned access.
3208 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3210 * config/nds32/nds32-intrinsic.md: Likewise.
3211 * config/nds32/nds32_intrinsic.h: Likewise.
3212 * config/nds32/nds32.h (nds32_builtins): Likewise.
3213 * config/nds32/nds32.opt (munaligned-access): New option.
3214 * config/nds32/nds32.c (nds32_asm_file_start): Display
3215 flag_unaligned_access status.
3217 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3219 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3220 -mno-relax is present.
3221 * config/riscv/linux.h (LINK_SPEC): Ditto.
3223 2018-04-20 Martin Sebor <msebor@redhat.com>
3226 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3228 (gimple_fold_builtin_stxcpy_chk): Same.
3229 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3231 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3234 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3235 __powikf2 when long double is IEEE 128-bit.
3237 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3239 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3240 step to make sure stack always aligned.
3242 2018-04-20 Carl Love <cel@us.ibm.com>
3245 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3246 size check for arg0.
3248 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3249 Tom de Vries <tom@codesourcery.com>
3252 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3253 Emit insns for calls too.
3254 (nvptx_find_par): Always look for worker-level predecessor insn.
3255 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3257 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3258 (nvptx_process_pars): Propagate frames for calls.
3260 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3263 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3265 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3266 (ix86_handle_option): Don't handle OPT_mibt.
3267 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3269 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3271 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3272 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3273 (ix86_target_macros): Define __CET__ with flag_cf_protection
3274 for -fcf-protection.
3275 * config/i386/i386.c (isa2_opts): Remove -mibt.
3276 * config/i386/i386.h (TARGET_IBT): Removed.
3277 (TARGET_IBT_P): Likewise.
3278 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3279 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3280 * config/i386/i386.opt (mcet): Update help message.
3283 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3284 -mcet as an alias for -mshstk.
3286 2018-04-20 Richard Biener <rguenther@suse.de>
3289 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3290 complexity by forcing a single use of the multiply operand.
3292 2018-04-20 Martin Jambor <mjambor@suse.cz>
3295 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3296 recursion dependency to only apply to non-clones.
3298 2018-04-20 Martin Jambor <mjambor@suse.cz>
3301 * ipa-cp.c (create_specialized_node): Check that clones of
3302 self-recursive edges exist during IPA-CP.
3304 2018-04-19 Toon Moene <toon@moene.org>
3306 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3309 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3311 PR tree-optimization/85467
3312 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3313 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3314 VECTOR_CST element to type.
3316 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3319 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3320 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3321 (builtin_longjmp): Likewise.
3322 (save_stack_nonlocal): New pattern.
3323 (restore_stack_nonlocal): Likewise.
3325 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3328 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3329 Replace ASM_OUTPUT_LABEL with fprintf.
3331 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3334 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3335 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3336 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3337 define __IBT__ and __SHSTK__ for -fcf-protection.
3338 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3340 (ix86_trampoline_init): Likewise.
3341 (x86_output_mi_thunk): Likewise.
3342 (ix86_notrack_prefixed_insn_p): Likewise.
3343 (ix86_option_override_internal): Don't disallow -fcf-protection.
3344 * config/i386/i386.md (rdssp<mode>): Also enable for
3346 (incssp<mode>): Likewise.
3347 (nop_endbr): Likewise.
3348 * config/i386/i386.opt (mcet): Change help message to built-in
3352 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3353 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3354 enable CET built-in functions.
3356 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3358 * common/config/i386/i386-common.c
3359 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3360 OPTION_MASK_ISA_MOVDIRI_UNSET,
3361 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3362 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3363 * config.gcc (movdirintrin.h): New header.
3364 * config/i386/cpuid.h (bit_MOVDIRI,
3365 bit_MOVDIR64B): New bits.
3366 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3368 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3369 (VOID, PVOID, PCVOID)): New function types.
3370 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3371 __builtin_ia32_directstoreu_u64,
3372 __builtin_ia32_movdir64b): New builtins.
3373 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3374 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3376 (ix86_valid_target_attribute_inner_p): Ditto.
3377 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3378 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3379 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3380 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3381 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3382 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3383 (movdiri<mode>, movdir64b_<mode>): New.
3384 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3385 * config/i386/immintrin.h: Include movdirintrin.h.
3386 * config/i386/movdirintrin.h: New file.
3387 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3389 2018-04-19 Richard Biener <rguenther@suse.de>
3392 * cfg.c (clear_bb_flags): When loop state says we have
3393 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3395 2018-04-19 Richard Biener <rguenther@suse.de>
3397 PR tree-optimization/84737
3398 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3399 copying restrict info.
3400 (vect_setup_realignment): Use it.
3401 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3402 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3403 the first DR to all generated stores.
3404 (vectorizable_load): Likewise for loads.
3406 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3408 PR tree-optimization/85446
3409 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3410 the integral and pointer types to have the same precision.
3412 * doc/install.texi: Document --disable-cet being the default and
3415 2018-04-18 Martin Liska <mliska@suse.cz>
3417 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3420 2018-04-18 Martin Liska <mliska@suse.cz>
3423 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3426 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3427 arguments if they are comparable.
3429 2018-04-18 Martin Liska <mliska@suse.cz>
3432 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3435 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3438 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3441 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3442 ENDBR after calling __morestack.
3444 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3447 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3448 by using gcc_base_ver to generate a gcc_driver_version, and use
3449 it when generating GCC_DRIVER_NAME.
3450 * configure: Regenerate.
3452 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3455 * config.gcc: Obsolete powerpc*-*-*spe*.
3457 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3460 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3461 (stabstr_D): Change type of unum from unsigned int to
3462 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3465 2018-04-17 Jim Wilson <jimw@sifive.com>
3468 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3469 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3470 Set arg_pointer_offset after using pretend_args_size.
3472 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3474 PR rtl-optimization/85431
3475 * dse.c (record_store): Ignore zero width stores.
3478 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3479 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3480 __builtin_stack_restore rather than after it.
3481 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3482 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3483 argument instead of virtual_dynamic_stack_rtx.
3485 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3487 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3489 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3490 Add note to error message to explain internal mapping of overloaded
3491 built-in function name to non-overloaded built-in function name.
3492 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3495 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3498 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3499 where the inputs overlap with the output.
3501 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3503 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3504 (=v, v) alternative and explicit "memory" attribute.
3505 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3506 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3508 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3509 "sselog1" type instead of "sselog".
3510 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3511 "sselog". Remove explicit "memory" attribute.
3512 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3513 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3515 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3516 "sselog1" type instead of "sselog". Remove explicit "memory"
3518 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3519 use "sselog1" type instead of "sselog". Remove explicit "memory"
3521 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3522 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3524 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3525 "sselog1" type instead of "sselog". Remove explicit "memory"
3527 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3528 use "sselog1" type instead of "sselog". Remove explicit "memory"
3532 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3535 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3538 2018-04-17 Martin Jambor <mjambor@suse.cz>
3541 * ipa-cp.c (create_specialized_node): Call
3542 expand_all_artificial_thunks if necessary.
3544 2018-04-17 Martin Liska <mliska@suse.cz>
3547 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3548 in message, remote space in between '_G' and '('.
3550 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3553 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3554 avx512f_vmcmp<mode>3<round_saeonly_name>,
3555 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3556 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3557 avx512f_rndscale<mode><round_saeonly_name>,
3558 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3559 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3560 Use %<iptr>2 instead of %2 for -masm=intel.
3561 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3562 avx512f_vcvttss2usi<round_saeonly_name>,
3563 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3565 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3566 avx512f_vcvttsd2usi<round_saeonly_name>,
3567 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3568 Use %q1 instead of %1 for -masm=intel.
3569 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3570 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3571 of %3 for -masm=intel.
3572 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3574 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3576 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3578 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3579 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3581 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3583 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3584 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3585 %g1 and one with %0 and %1.
3586 (avx512er_vmrcp28<mode><round_saeonly_name>,
3587 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3589 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3590 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3591 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3592 of %0 and %{%4%} for -masm=intel.
3593 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3594 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3595 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3596 order of %0 and %{%5%}%{z%} for -masm=intel.
3598 2018-04-17 Jan Hubicka <jh@suse.cz>
3601 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3603 2018-04-17 Martin Liska <mliska@suse.cz>
3606 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3607 for target_clone error message. Make default implementation
3608 clone to be a local declaration.
3609 (separate_attrs): Add new argument and check for an empty
3611 (expand_target_clones): Handle it.
3612 (ipa_target_clone): Make redirection just for target_clones
3615 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3616 Tom de Vries <tom@codesourcery.com>
3619 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3620 tiled basic blocks without omp continue statements.
3622 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3625 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3626 vec_extract expression as having side effects to make sure it gets
3629 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3632 * config/i386/i386.c (get_builtin_code_for_version): Check
3635 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3638 * gcc/config.gcc: Support "skylake".
3639 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3641 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3642 (processor_target_table): Add "skylake".
3643 (ix86_option_override_internal): Add "skylake".
3644 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3645 PROCESSOR_CANNONLAKE.
3646 (get_builtin_code_for_version): Fix priority for
3647 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3648 PROCESSOR_SKYLAKE-AVX512.
3649 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3650 (processor_type): Add PROCESSOR_SKYLAKE.
3652 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3653 Jason Merrill <jason@redhat.com>
3656 * convert.c (convert_to_integer_1): Use direct recursion for
3657 enumeral types and types with a precision less than the number
3658 of bits in their mode.
3660 2018-04-16 Julia Koval <julia.koval@intel.com>
3663 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3664 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3666 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3669 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3670 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3671 and -mno-direct-move.
3673 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3676 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3677 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3678 Ensure negative shifts result in {0}.
3680 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3682 PR rtl-optimization/79916
3683 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3684 regs (if any) to define how to gnerate SD moves when LRA is in
3687 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3689 PR rtl-optimization/85393
3690 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3691 * except.c (expand_dw2_landing_pad_for_region): Make static.
3692 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3693 a label and unconditional jump to old_bb, rather than
3694 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3697 PR rtl-optimization/85376
3698 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3699 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3700 instead of a specific value.
3702 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3703 Bin Cheng <bin.cheng@arm.com>
3705 PR tree-optimization/82965
3706 PR tree-optimization/83991
3707 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3708 by_profile_only parameter.
3709 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3710 information if the loop was predicted to iterate too many times.
3711 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3713 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3716 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3719 2018-04-13 Martin Liska <mliska@suse.cz>
3720 Jakub Jelinek <jakub@redhat.com>
3723 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3724 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3725 * builtins.c (expand_builtin_memory_copy_args): Use
3726 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3727 handle dest_addr == pc_rtx.
3729 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3732 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3733 asked to not generate direct moves.
3734 (fix_trunc<mode>si2_stfiwx): Similar.
3735 (fix_trunc<mode>si2_internal): Similar.
3737 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3740 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3741 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3742 lookup if dest in some wider mode is known to be const0_rtx and
3743 if so, record permanent equivalence for it to be ZERO_EXTEND of
3744 the narrower mode destination.
3746 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3748 * lto-streamer-out.c (output_function): Revert 259346.
3749 * omp-expand.c (expand_oacc_for): Likewise.
3751 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3753 PR rtl-optimization/85354
3754 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3755 * sel-sched.c (sel_global_init): ... here.
3757 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3760 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3761 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3762 mode for PE-COFF targets.
3763 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3764 (i386_pe_asm_lto_end): Likewise.
3765 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3766 (TARGET_ASM_LTO_END): Likewise.
3767 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3768 (i386_pe_asm_lto_start): New function.
3769 (i386_pe_asm_lto_end): Likewise.
3771 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3772 Richard Biener <rguenther@suse.de>
3775 * lto-streamer-out.c (output_function): Fix CFG loop state before
3777 * omp-expand.c (expand_oacc_for): Handle calls to internal
3778 functions like regular functions.
3780 2018-04-12 Richard Biener <rguenther@suse.de>
3783 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3784 for the early LTO debug to properly generate references to it
3785 during DIE emission. Do not re-use that for the skeleton for
3787 (dwarf2out_early_finish): Likewise.
3789 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3792 * config/i386/sse.md
3793 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3794 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3795 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3796 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3797 and output is a reg, avoid creating invalid lowpart subreg, but
3798 instead split into a 512-bit move. Don't split if not AVX512VL,
3799 input is xmm16+ reg and output is a mem.
3800 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3801 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3802 xmm16+ reg and output is a mem.
3804 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3806 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3807 also for flag_dwarf2_cfi_asm.
3809 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3811 PR rtl-optimization/85342
3812 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3813 a bool scalar var inside of the loop instead. Don't try to update
3814 recog_data.operand after failed apply_change_group.
3816 2018-04-12 Tom de Vries <tom@codesourcery.com>
3819 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3820 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3821 array with flexible array member as array without given dimension.
3822 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3823 argument for undefined param to true.
3825 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3828 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3829 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3830 from PowerPC section.
3831 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3832 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3833 help text that is too long.
3834 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3835 help text that is too long.
3836 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3837 help text that is too long.
3839 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3841 * config/alpha/alpha.md (stack_probe_internal): Rename
3842 from "probe_stack". Update all callers.
3844 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3846 PR rtl-optimization/84566
3847 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3848 sched_macro_fuse_insns.
3850 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3853 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3854 (compute_block_dependences): ... from here.
3856 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3858 PR tree-optimization/85331
3859 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3860 from int to HOST_WIDE_INT.
3862 2018-04-11 Martin Jambor <mjambor@suse.cz>
3865 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3866 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3867 not the same as the source val.
3868 (cgraph_edge_brings_value_p): New parameter.
3869 (gather_edges_for_value): Pass destination value to
3870 cgraph_edge_brings_value_p.
3871 (perhaps_add_new_callers): Likewise.
3872 (get_info_about_necessary_edges): Likewise and exclude values brought
3873 only by self-recursive edges.
3874 (create_specialized_node): Redirect only clones of self-calling edges.
3875 (+self_recursive_pass_through_p): New function.
3876 (find_more_scalar_values_for_callers_subset): Use it.
3877 (find_aggregate_values_for_callers_subset): Likewise.
3878 (known_aggs_to_agg_replacement_list): Removed.
3879 (decide_whether_version_node): Re-calculate known constants for all
3880 remaining context clones.
3882 2018-04-11 Richard Biener <rguenther@suse.de>
3885 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3886 from early DWARF output.
3887 (dwarf2out_early_finish): Output line info unconditionally into
3888 early DWARF and add reference to it.
3890 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3893 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3894 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3895 other than V2DFmode using iptr mode attribute.
3896 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3898 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3900 PR rtl-optimization/84659
3901 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3903 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3906 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3908 (output_loc_list): Pass address of a dummy size variable even in the
3909 locview handling loop.
3910 (index_location_lists): Add comment on why skip_loc_list_entry can't
3913 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3916 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3919 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3922 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3923 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3924 and -mstring-compare-inline-limit.
3926 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3929 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3930 for stack clash protection in a register whenever we need it to be in
3933 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3935 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3936 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3938 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3941 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3943 (mlong-double-): Ditto.
3944 * config/rs6000/sysv4.opt (msdata=): Ditto.
3945 (mtls-size=): Ditto.
3947 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3949 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3950 erroneous entries for
3951 "vector int vec_ldl (int, long int *)", and
3952 "vector unsigned int vec_ldl (int, unsigned long int *)".
3953 Add comments and entries for
3954 "vector bool char vec_ldl (int, bool char *)",
3955 "vector bool short vec_ldl (int, bool short *)",
3956 "vector bool int vec_ldl (int, bool int *)",
3957 "vector bool long long vec_ldl (int, bool long long *)",
3958 "vector pixel vec_ldl (int, pixel *)",
3959 "vector long long vec_ldl (int, long long *)",
3960 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3961 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3962 type tree bool_long_long_type_node and correct definition of
3963 bool_V2DI_type_node to make reference to this new type tree.
3964 (rs6000_mangle_type): Replace erroneous reference to
3965 bool_long_type_node with bool_long_long_type_node.
3966 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3967 comments to emphasize sign distinctions for char and int types and
3968 replace RS6000_BTI_bool_long constant with
3969 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3970 use of RS6000_BTI_pixel.
3971 (bool_long_type_node): Remove this macro definition.
3972 (bool_long_long_type_node): New macro definition
3974 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3976 PR rtl-optimization/85300
3977 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3978 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3979 simplify_unary_operation fails.
3981 2018-04-10 Martin Liska <mliska@suse.cz>
3983 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3984 cgraph_edge and ipa_ref.
3986 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3990 * config/i386/sse.md
3991 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3992 computation of the VEC_MERGE selector from mask.
3993 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3994 Fix decoding of the VEC_MERGE selector into mask.
3996 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3998 PR tree-optimization/85286
3999 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4001 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4003 * final.c (final_1): Set insn_last_address as well as
4004 insn_current_address.
4006 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4009 * explow.c (emit_stack_probe): Call validize_mem on memory location
4010 before passing it to gen_probe_stack. Create address operand and
4011 legitimize it for the probe_stack_address case.
4013 2018-04-09 Jan Hubicka <jh@suse.cz>
4016 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4017 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4018 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4020 (free_lang_data): Rebuild type inheritance graph.
4022 2018-04-09 Martin Sebor <msebor@redhat.com>
4024 * invoke.texi (-finline-small-functions): Mention other optimization
4026 (-findirect-inlining, -fpartial-inlining): Same.
4027 (-finline-functions-called-once): Same.
4028 (-freorder-blocks-and-partition): Same.
4030 2018-04-09 Jan Hubicka <jh@suse.cz>
4033 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4034 jumps; choose last target that matches the criteria (i.e.
4035 no partition changes for non-crossing jumps).
4036 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4037 support for redirecting crossing jumps to non-crossing.
4039 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4041 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4042 also for naked functions.
4044 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4046 * config/arc/arc.md (add_shift): New pattern.
4047 (add_shift2): Likewise.
4048 (sub_shift): Likewise.
4049 (sub_shift_cmp0_noout): Likewise.
4050 (compare_si_ashiftsi): Likewise.
4051 (xbfu_cmp0_noout): New combine pattern.
4052 (xbfu_cmp0"): Likewise.
4053 (movsi_set_cc_insn): Place the predicable variant first.
4054 (commutative_binary_cmp0_noout): Remove clobber.
4055 (commutative_binary_cmp0): New pattern.
4056 (noncommutative_binary_cmp0): Likewise.
4057 (noncommutative_binary_cmp0_noout): Likewise.
4058 (noncommutative_binary_comparison_result_used): Removed.
4059 (rsub_cmp0): New pattern.
4060 (rsub_cmp0_noout): Likewise.
4061 (extzvsi): Changed, keep only meaningful variants.
4062 (SQH, SEZ): New iterators.
4063 (SQH_postfix): New mode attribute.
4064 (SEZ_prefix): New code attribute.
4065 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4066 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4067 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4069 (noncommutative_operator): Check the availability of barrel
4072 2018-04-09 Richard Biener <rguenther@suse.de>
4074 PR tree-optimization/85284
4075 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4076 Only use the niter constraining form of simple_iv when the exit
4079 2018-04-09 Tom de Vries <tom@codesourcery.com>
4082 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4083 (define_expand "*memory_barrier"): New define_expand.
4084 (define_insn "memory_barrier"): New insn.
4086 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4088 PR rtl-optimization/80463
4089 PR rtl-optimization/83972
4090 PR rtl-optimization/83480
4092 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4093 correct producer for the insn.
4094 (tidy_control_flow): Fixup seqnos in case of debug insns.
4096 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4098 PR rtl-optimization/83913
4100 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4101 different sched-times when merging exprs.
4103 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4105 PR rtl-optimization/83962
4107 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4108 tidy_fallthru_edge and tidy_control_flow.
4110 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4112 PR rtl-optimization/83530
4114 * sel-sched.c (force_next_insn): New global variable.
4115 (remove_insn_for_debug): When force_next_insn is true, also leave only
4116 next insn in the ready list.
4117 (sel_sched_region): When the region wasn't scheduled, make another pass
4118 over it with force_next_insn set to 1.
4120 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4122 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4124 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4125 for interrupt control.
4126 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4127 functions for interrupt control.
4128 * config/nds32/nds32-intrinsic.md: Likewise.
4129 * config/nds32/nds32_intrinsic.h: Likewise.
4130 * config/nds32/nds32.h (nds32_builtins): Likewise.
4132 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4134 * config/nds32/nds32.c (nds32_init_machine_status,
4135 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4136 strict_aligned_p field.
4137 (nds32_expand_to_rtl_hook): New function.
4138 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4139 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4141 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4142 Chung-Ju Wu <jasonwucj@gmail.com>
4144 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4145 * config/nds32/nds32-n7.md: New file.
4146 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4147 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4149 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4150 * config/nds32/nds32.md (pipeline_model): Add n7.
4151 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4152 * config/nds32/pipelines.md: Include n7 settings.
4154 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4155 Chung-Ju Wu <jasonwucj@gmail.com>
4157 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4158 * config/nds32/nds32-e8.md: New file.
4159 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4160 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4162 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4163 * config/nds32/nds32.md (pipeline_model): Add e8.
4164 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4165 * config/nds32/pipelines.md: Include e8 settings.
4167 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4168 Chung-Ju Wu <jasonwucj@gmail.com>
4170 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4171 * config/nds32/nds32-n8.md: New file.
4172 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4173 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4175 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4176 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4177 * config/nds32/nds32.md (pipeline_model): Add n8.
4178 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4179 * config/nds32/pipelines.md: Include n8 settings.
4181 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4182 Chung-Ju Wu <jasonwucj@gmail.com>
4184 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4185 * config/nds32/nds32-n9-2r1w.md: New file.
4186 * config/nds32/nds32-n9-3r2w.md: New file.
4187 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4188 nds32_register_ports): New or modify for cpu n9.
4189 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4191 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4192 * config/nds32/nds32-utils.c: New file.
4193 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4194 TARGET_MUL_SLOW): Define.
4195 * config/nds32/nds32.md (pipeline_model): New attribute.
4196 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4197 New options that support cpu n9.
4198 * config/nds32/pipelines.md: Include n9 settings.
4199 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4201 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4203 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4204 information if necessary.
4205 (output_cond_branch_compare_zero): Likewise.
4206 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4207 (nds32_target_alignment): Refine for alignment.
4208 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4209 (FUNCTION_BOUNDARY): Modify.
4210 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4212 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4214 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4216 * config/nds32/constants.md (unspec_volatile_element): Add values for
4217 TLB operation and data prefetch.
4218 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4219 functions for TLB operation and data prefetch.
4220 * config/nds32/nds32-intrinsic.md: Likewise.
4221 * config/nds32/nds32_intrinsic.h: Likewise.
4222 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4223 (nds32_print_operand): Likewise.
4224 * config/nds32/nds32.h (nds32_builtins): Likewise.
4226 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4227 Andrew Pinski <pinsika@gcc.gnu.org>
4230 * match.pd: Use constant_boolean_node of correct type instead of
4231 boolean_true_node or boolean_false_node for simplifying
4232 pointer comparisons to zero.
4234 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4236 PR tree-optimization/80021
4237 * tree.c (verify_type_variant): Make error call in verify_variant_match
4238 translatable and remove final full stop.
4240 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4242 * config/nds32/constants.md (unspec_volatile_element): Add
4243 UNSPEC_VOLATILE_EH_RETURN.
4244 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4245 nds32_output_stack_pop): Support dwarf exception handling process.
4246 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4247 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4248 exception handling process.
4249 (nds32_compute_stack_frame): Likewise.
4250 (nds32_return_addr_rtx): Likewise.
4251 (nds32_initial_elimination_offset): Likewise.
4252 (nds32_expand_prologue): Likewise.
4253 (nds32_expand_epilogue): Likewise.
4254 (nds32_dynamic_chain_address): New function.
4255 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4257 (DYNAMIC_CHAIN_ADDRESS): Define.
4258 (EH_RETURN_DATA_REGNO): Define.
4259 (EH_RETURN_STACKADJ_RTX): Define.
4260 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4261 patterns for dwarf exception handling.
4263 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4265 * config/nds32/nds32.h: Clean up obsolete macros.
4267 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4269 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4270 Add enum values for particular instructions.
4271 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4272 particular intrinsic functions.
4273 * config/nds32/nds32-intrinsic.md: Likewise.
4274 * config/nds32/nds32_intrinsic.h: Likewise.
4275 * config/nds32/nds32.h (nds32_builtins): Likewise.
4276 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4277 (btst, ave): New patterns for particular instructions.
4279 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4281 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4282 Add enum values for atomic load/store and memory sync.
4283 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4285 * config/nds32/nds32-intrinsic.md: Likewise.
4286 * config/nds32/nds32_intrinsic.h: Likewise.
4287 * config/nds32/nds32.h (nds32_builtins): Likewise.
4289 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4291 PR tree-optimization/85257
4292 * fold-const.c (native_encode_vector): If not all elts could fit
4293 and off is -1, return 0 rather than offset.
4294 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4295 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4296 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4297 adjust buffer in native_interpret_expr call.
4299 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4301 * config/nds32/constants.md (unspec_volatile_element): Add cache
4302 control enum values.
4303 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4304 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4305 * config/nds32/nds32.c (nds32_cctl_names): New.
4306 (nds32_print_operand): Handle cache control register names.
4307 * config/nds32/nds32.h (nds32_builtins): New enum values.
4308 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4310 * config/nds32/nds32.md (type): Add mmu.
4311 * config/nds32/pipelines.md (simple_insn): Add mmu.
4313 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4315 * config/nds32/nds32.md (type): Remove call.
4316 * config/nds32/pipelines.md (simple_insn): Likewise.
4318 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4320 * config/nds32/constants.md (unspec_volatile_element): Add
4321 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4322 UNSPEC_VOLATILE_FMFCFG.
4323 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4324 description for fmfcfg and fmfcsr.
4325 (bdesc_1arg): Add fmtcsr.
4326 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4327 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4328 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4329 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4330 unspec_fmfcfg): New patterns.
4331 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4332 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4333 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4334 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4335 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4336 __nds32__fmfcfg): Define.
4338 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4340 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4341 intrinsic register names.
4342 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4343 intrinsic register enum values and macros.
4345 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4347 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4348 for load/store addressing form.
4349 (nds32_print_operand_address): Likewise.
4351 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4354 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4355 based on LABEL_REF. Remove useless assertion.
4356 (pic_address_needs_scratch): Fix formatting.
4357 (sparc_legitimize_pic_address): Minor tweaks.
4358 (sparc_delegitimize_address): Adjust assertion accordingly.
4359 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4360 into symbolic_operand.
4361 (movsi_high_pic_label_ref): Likewise.
4362 (movsi_lo_sum_pic_label_ref): Likewise.
4363 (movdi_pic_label_ref): Likewise.
4364 (movdi_high_pic_label_ref): Likewise.
4365 (movdi_lo_sum_pic_label_ref): Likewise.
4367 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4369 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4370 custom LIB_SPEC setup.
4372 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4373 Kito Cheng <kito.cheng@gmail.com>
4375 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4376 * config/riscv/freebsd.h: New.
4378 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4380 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4381 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4384 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4385 Kito Cheng <kito.cheng@gmail.com>
4387 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4388 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4389 * config/nds32/nds32-protos.h (nds32_output_call,
4390 nds32_output_return): Declare.
4391 * config/nds32/nds32.md: Refine all the call and return patterns.
4393 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4396 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4397 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4399 PR rtl-optimization/84872
4400 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4401 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4404 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4406 * expr.c (copy_blkmode_to_reg): Revert 254862.
4407 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4409 2018-04-06 Richard Biener <rguenther@suse.de>
4412 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4413 after seeing a component reference with an adjacent field. Treat
4414 refs to arrays at struct end of external decls similar to
4415 refs to unconstrained commons.
4417 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4420 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4421 look through SAVE_EXPRs with non-side-effects argument. Adjust
4423 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4426 2018-04-06 Richard Biener <rguenther@suse.de>
4429 * alias.c (find_base_term): New wrapper around find_base_term
4430 unwinding CSELIB_VAL_PTR changes.
4431 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4434 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4436 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4438 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4439 constant definitions.
4440 ("nop"): lr 0,0 -> nopr r0
4441 ("nop_lr0", "nop_lr1"): New insn definitions.
4443 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4445 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4446 NDS32_V3PUSH_AVAILABLE_P macro.
4448 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4449 Chung-Ju Wu <jasonwucj@gmail.com>
4451 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4452 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4453 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4454 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4455 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4456 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4457 * config/nds32/constraints.md: New constraints and checking for hard
4458 float configuration.
4459 * config/nds32/iterators.md: New mode iterator and attribute for hard
4460 float configuration.
4461 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4463 * config/nds32/nds32-fpu.md: New file.
4464 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4465 deal with hard float code generation.
4466 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4468 (abi_type, float_reg_number): New enum type.
4469 * config/nds32/nds32-predicates.c: New predicates for hard float.
4470 * config/nds32/nds32-protos.h: Declare functions for hard float.
4471 * config/nds32/nds32.c: Implementation for hard float configuration.
4472 * config/nds32/nds32.h: Definitions for hard float configuration.
4473 * config/nds32/nds32.md: Include hard float machine description and
4474 modify patterns for hard float configuration.
4475 * config/nds32/nds32.opt: New options for hard float configuration.
4476 * config/nds32/predicates.md: New predicates for hard float
4479 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4481 * common/config/nds32/nds32-common.c
4482 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4484 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4487 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4488 CONSTRUCTOR_ELT (ctor, ...)->value.
4490 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4493 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4495 2018-04-05 Tom de Vries <tom@codesourcery.com>
4498 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4501 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4502 Kito Cheng <kito.cheng@gmail.com>
4504 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4505 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4506 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4507 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4509 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4510 Kito Cheng <kito.cheng@gmail.com>
4512 * config/nds32/constraints.md (Ufe): New memory constraint.
4513 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4514 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4515 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4517 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4518 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4520 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4522 * config/nds32/nds32.md: Use optimize_size in the condition for
4523 alu-shift instructions.
4525 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4527 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4529 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4531 * config/nds32/nds32.md (negsi2): Refine pattern.
4533 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4534 Chung-Ju Wu <jasonwucj@gmail.com>
4536 * config/nds32/iterators.md (shift_rotate): New code iterator.
4537 (shift): New code attribute.
4538 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4539 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4540 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4541 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4542 bit-wise operations.
4543 (andsi3, *andsi3): Ditto.
4544 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4545 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4546 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4547 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4548 nds32_ior_operand, nds32_xor_operand): New predicates.
4550 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4552 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4553 (addsi3, subsi3): ... this.
4555 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4557 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4559 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4561 * config/nds32/nds32.md: Adjust indention.
4563 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4565 * config/nds32/nds32.md (feature): New attribute.
4567 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4569 * config/nds32/nds32.md (subtype): New attribute.
4571 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4574 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4575 expansion to perform a bitwise AND of the argument followed by a
4576 boolean negation of the result.
4578 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4580 PR rtl-optimization/84878
4581 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4582 the basic block. Assert the use reference is not artificial and that
4583 it has an associated insn.
4585 2018-04-04 Michael Matz <matz@suse.de>
4587 * builtins.c (compute_objsize): Pass correct operand
4588 to array_at_struct_end_p.
4590 2018-04-04 Richard Biener <rguenther@suse.de>
4593 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4594 from contexts for DINFO_LEVEL_TERSE and below.
4596 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4598 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4599 resiter_operand condition.
4600 * config/nds32/nds32.md (*move<mode>): Ditto.
4602 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4603 Monk Chiang <sh.chiang04@gmail.com>
4605 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4607 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4609 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4611 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4612 Kito Cheng <kito.cheng@gmail.com>
4614 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4615 nds32_cond_code_str, output_cond_branch,
4616 output_cond_branch_compare_zero, nds32_expand_cbranch,
4617 nds32_expand_cstore, nds32_expand_movcc,
4618 nds32_output_cbranchsi4_equality_zero,
4619 nds32_output_cbranchsi4_equality_reg,
4620 nds32_output_cbranchsi4_equality_reg_or_const_int,
4621 nds32_output_cbranchsi4_greater_less_zero: New functions.
4622 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4623 nds32_expand_cstore, nds32_expand_movcc,
4624 nds32_output_cbranchsi4_equality_zero,
4625 nds32_output_cbranchsi4_equality_reg,
4626 nds32_output_cbranchsi4_equality_reg_or_const_int,
4627 nds32_output_cbranchsi4_greater_less_zero): Declare.
4628 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4629 nds32_rimm11s_operand): New predicates.
4630 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4631 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4634 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4636 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4637 * config/nds32/nds32.md: Ditto.
4638 * config/nds32/pipelines.md: Ditto.
4640 2018-04-04 Richard Biener <rguenther@suse.de>
4642 PR tree-optimization/85168
4643 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4644 propagating abnormals.
4646 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4648 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4650 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4651 Kito Cheng <kito.cheng@gmail.com>
4653 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4654 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4655 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4656 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4657 * config/nds32/nds32.md (sibcall_internal): New.
4658 (sibcall_register): Remove.
4659 (sibcall_immediate): Remove.
4660 (sibcall_value_internal): New.
4661 (sibcall_value_register): Remove.
4662 (sibcall_value_immediate): Remove.
4663 * config/nds32/predicates.md (nds32_general_register_operand): New.
4664 (nds32_call_address_operand): New.
4666 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4668 PR rtl-optimization/85167
4669 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4670 bb_defs if *split_p, instead preinitialize it to NULL.
4672 PR tree-optimization/85156
4673 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4674 evaluating the argument multiple times.
4676 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4678 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4680 (_mm_cvtpd_ps): Likewise.
4681 (_mm_cvttpd_epi32): Likewise.
4682 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4683 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4684 vector, pixel, and bool following altivec.h include.
4686 2018-04-03 Martin Sebor <msebor@redhat.com>
4688 * doc/extend.texi (Common Function Attributes): Clarify.
4689 (const attribute): Likewise.
4690 (pure attribute): Likewise.
4692 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4695 * config/i386/i386.c (ix86_expand_vector_set): Use
4696 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4698 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4700 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4701 instructions when changing rounding bits to preserve precision bits
4702 in the x87 control word.
4704 2018-04-03 Martin Liska <mliska@suse.cz>
4706 PR tree-optimization/82491
4707 * rtl.h (strip_offset_and_add): Replace += suboffset with
4708 poly_uint64 () + suboffset.
4710 2018-03-29 Martin Liska <mliska@suse.cz>
4711 Martin Jambor <mjambor@suse.cz>
4714 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4715 param_type is not an integral or pointer type.
4717 2018-04-03 Richard Biener <rguenther@suse.de>
4719 * sese.h (recompute_all_dominators): Remove.
4721 2018-04-02 Martin Sebor <msebor@redhat.com>
4723 * doc/invoke.texi (-Wrestrict): Fix typos.
4725 2018-04-02 Jim Wilson <jimw@sifive.com>
4727 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4728 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4729 (<optab>di3, <optab>si3_extend): Likewise.
4730 (<optab>si3_mask, <optab>si3_mask_1): New.
4731 (<optab>di3_mask, <optab>di3_mask_1): New.
4732 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4733 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4734 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4736 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4738 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4741 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4743 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4744 (nds32_canonicalize_comparison): New function.
4746 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4747 Kito Cheng <kito.cheng@gmail.com>
4748 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4750 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4751 * config/nds32/constants.md (unspec_volatile_element): Add
4752 UNSPEC_VOLATILE_RELAX_GROUP.
4753 * config/nds32/nds32-relax-opt.c: New file.
4754 * config/nds32/nds32-predicates.c
4755 (nds32_symbol_load_store_p): New function.
4756 * config/nds32/nds32-protos.h
4757 (nds32_symbol_load_store_p): Declare function.
4758 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4759 * config/nds32/nds32.c
4760 (nds32_register_pass): New function to register pass.
4761 (nds32_register_passes): New function to register passes.
4762 * config/nds32/nds32.md (relax_group): New pattern.
4763 * config/nds32/nds32.opt (mrelax-hint): New option.
4764 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4766 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4768 * config/nds32/t-nds32: Modify files dependency.
4770 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4772 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4773 (PROFILE_HOOK): Define its implementation.
4775 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4777 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4778 type and 32-bit size.
4780 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4783 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4784 (V_128_256): New mode iterator.
4785 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4786 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4787 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4789 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4790 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4792 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4795 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4796 NaN inputs correctly.
4798 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4801 * config/rs6000/vsx.md (??r): New mode attribute.
4802 (*vsx_mov<mode>_64bit): Use it.
4803 (*vsx_mov<mode>_32bit): Likewise.
4805 2018-03-30 Martin Sebor <msebor@redhat.com>
4807 PR tree-optimization/84818
4808 * builtins.c (check_access): Use warning_n.
4810 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4813 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4815 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4818 2018-03-30 Julia Koval <julia.koval@intel.com>
4821 * x86-tune.def (movx, partial_reg_dependency): Enable for
4824 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4827 * lra-constraints.c (process_alt_operands): Move setting
4828 this_alternative_matches below.
4830 2018-03-29 Martin Liska <mliska@suse.cz>
4833 * doc/invoke.texi: Document how LTO works with debug info.
4834 Describe auto-load support of binutils. Mention 'x86-64'
4835 as valid option value of -march option.
4837 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4839 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4842 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4843 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4844 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4847 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4850 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4851 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4852 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4853 for RS6000_BTM_POWERPC64.
4854 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4855 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4856 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4861 2018-03-28 Carl Love <cel@us.ibm.com>
4864 2017-09-27 Carl Love <cel@us.ibm.com>
4866 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4867 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4868 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4871 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4873 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4874 instead of __vector bool.
4875 (_mm_max_pu8): Likewise.
4876 (_mm_min_pi16): Likewise.
4878 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4881 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4882 (DIVWEUO): Likewise.
4884 (DIVDEUO): Likewise.
4885 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4886 DIVWEUO and DIVDEUO.
4887 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4888 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4889 (div_extend): Likewise.
4890 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4892 (__builtin_divweuo): Likewise.
4893 (__builtin_divdeo): Likewise.
4894 (__builtin_divdeuo): Likewise.
4896 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4899 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4900 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4902 PR tree-optimization/82004
4903 * gimple-match-head.c (optimize_pow_to_exp): New function.
4904 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4905 Don't fold to exp if optimize_pow_to_exp is false.
4907 2018-03-28 Martin Liska <mliska@suse.cz>
4910 * calls.c (initialize_argument_information): Fix trailing space.
4911 * common.opt: Fix typo and provide better explanation for
4912 -fsanitize-coverage option.
4913 * config/i386/i386.opt: Fix typo.
4915 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4916 Martin Liska <mliska@suse.cz>
4919 * gimplify.c (asan_poison_variable): Don't do the check for
4920 gimplify_omp_ctxp here.
4921 (gimplify_decl_expr): Do it here.
4922 (gimplify_target_expr): Likewise.
4924 2018-03-28 Martin Liska <mliska@suse.cz>
4927 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4928 chkp_type_bounds_count if MPX is not enabled.
4930 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4932 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4934 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4937 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4938 function to create the function decl for complex long double
4939 multiply and divide for -mabi=ieeelongdouble.
4940 (init_float128_ieee): Call it.
4942 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4945 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4946 -fcf-protection=branch -mibt.
4947 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4949 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4952 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4954 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4957 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4958 extern array declarations.
4960 2018-03-27 Richard Biener <rguenther@suse.de>
4963 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4964 explicit single_use checks.
4966 2018-03-27 Richard Biener <rguenther@suse.de>
4968 PR tree-optimization/85082
4969 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4972 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4974 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4975 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4976 Turn on fasynchronous-unwind-tables and funwind-tables.
4978 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4981 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4982 (*bmi_blsr_<mode>_ccz): Ditto.
4984 2018-03-26 Tom de Vries <tom@codesourcery.com>
4986 PR tree-optimization/85063
4987 * omp-general.c (offloading_function_p): New function. Factor out
4989 * omp-offload.c (pass_omp_target_link::gate): ... here.
4990 * omp-general.h (offloading_function_p): Declare.
4991 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4992 with attribute omp declare target for offloading functions.
4994 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4996 PR tree-optimization/84005
4997 * tree-data-ref.h (get_base_for_alignment): Declare.
4998 * tree-data-ref.c (get_base_for_alignment_1): New function.
4999 (get_base_for_alignment): Likewise.
5000 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5001 get_base_for_alignment to find a suitable base object, instead
5002 of always using drb->base_address.
5004 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5007 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5008 known size by default.
5010 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5013 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5014 and non BLKmode operands.
5016 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5019 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5020 Clean up attributes.
5022 2018-03-23 Richard Biener <rguenther@suse.de>
5025 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5026 we are going to emit early debug for LTO.
5028 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5031 * function.c (match_asm_constraints_1): Don't optimize if input
5032 doesn't satisfy general_operand predicate for output's mode.
5035 * alias.c (write_dependence_p): Don't require for x_canonicalized
5036 non-VOIDmode if x has VOIDmode.
5039 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5040 just don't try to optimize it rather than assert it never happens.
5042 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5044 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5045 macro expansions for definition of ST_INTERNAL_<mode> and
5046 LD_INTERNAL_<mode> builtins.
5047 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5049 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5051 (altivec_expand_st_builtin): Likewise.
5052 (altivec_expand_builtin): Remove calls to deleted functions.
5053 (rs6000_address_for_altivec): Delete this function.
5054 * config/rs6000/vector.md: Remove expands for
5055 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5057 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5060 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5061 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5062 re-computing once computed.
5063 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5064 (arm_init_machine_status): Initialize
5065 machine->static_chain_stack_bytes.
5067 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5070 * doc/extend.texi: Add four new prototypes for vec_ld.
5071 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5072 definitions for more logical presentation.
5073 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5074 entries for V1TI variants of __builtin_altivec_ld builtin.
5075 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5076 handling of V1TI variant of LVX icode pattern.
5077 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5078 (rs6000_gimple_fold_builtin): Likewise.
5079 (altivec_init_builtins): Add code to define
5080 __builtin_altivec_lvx_v1ti function.
5082 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5085 * function.c (match_asm_constraints_1): Don't do the optimization
5086 if input isn't a REG, SUBREG, MEM or constant.
5088 2018-03-22 Tom de Vries <tom@codesourcery.com>
5090 PR tree-optimization/84956
5091 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5092 bb_has_abnormal_pred.
5094 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5097 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5098 DECL_INITIAL (decl) to decl at the end.
5099 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5102 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5104 * doc/extend.texi (__builtin_tgmath): Document when complex
5105 integer types are treated as _Complex _Float64.
5107 2018-03-21 Tom de Vries <tom@codesourcery.com>
5109 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5111 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5113 PR tree-optimization/84960
5114 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5115 if it is ENTRY block, move them into single succ of ENTRY in that case.
5117 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5119 PR tree-optimization/84811
5120 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5121 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5122 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5123 (known_subrange_p): Update accordingly. Cast each value involved
5124 in the size comparison, rather than casting the result of the
5127 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5129 PR tree-optimization/84982
5130 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5131 by flipping the least significant bit rather than all bits from
5132 bitpos to bitpos + bitsize - 1.
5134 2018-03-21 Nathan Sidwell <nathan@acm.org>
5136 * doc/extend.texi (Deprecated Features): Remove mention of
5137 long-deleted deprecations.
5139 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5142 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5143 * configure: Regenerate.
5145 2018-03-21 Tom de Vries <tom@codesourcery.com>
5147 PR tree-optimization/83126
5148 * tree-parloops.c (num_phis): New function.
5149 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5151 2018-03-21 Nathan Sidwell <nathan@acm.org>
5153 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5154 mention anon-struct/union members and trailing attributes.
5156 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5158 PR tree-optimization/84969
5159 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5160 builtin memset partitions if they set different rhs values.
5162 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5164 PR rtl-optimization/84989
5165 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5166 VEC_DUPLICATE with scalar result mode.
5168 2018-03-21 Martin Liska <mliska@suse.cz>
5171 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5172 not intended return statement.
5174 2018-03-21 Martin Liska <mliska@suse.cz>
5177 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5178 (chkp_find_bound_slots_1): Limit number of iterations.
5180 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5183 * Minor grammar fixes for x86 options.
5185 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5188 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5189 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5191 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5194 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5195 (altivec_lvx_<mode>_1op): Likewise.
5196 (altivec_stvx_<mode>_2op): Likewise.
5197 (altivec_stvx_<mode>_1op): Likewise.
5198 (altivec_lvx_<VM2:mode>): New define_expand.
5199 (altivec_stvx_<VM2:mode>): Likewise.
5200 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5201 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5202 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5203 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5204 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5205 (rs6000_gen_lvx): Likewise.
5206 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5207 (altivec_expand_stv_builtin): Likewise.
5208 (altivec_expand_builtin): Likewise.
5209 * config/rs6000/vector.md: Likewise.
5211 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5214 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5217 2018-03-20 Richard Biener <rguenther@suse.de>
5220 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5221 sign-conversions as zero, fall back to standard scalar_stmt
5224 2018-03-20 Martin Liska <mliska@suse.cz>
5227 * predict.c (rebuild_frequencies): Handle case when we have
5228 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5230 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5233 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5234 flag_section_anchors.
5235 * varasm.c (use_blocks_for_decl_p): Remove hack for
5236 dw2_force_const_mem.
5239 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5241 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5242 be created, use lowpart_subreg of operands[0] rather than operands[0]
5244 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5245 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5246 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5247 and n constraint instead of aarch64_shift_imm_di and Usd.
5248 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5249 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5251 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5254 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5255 to favor GPR over NEON registers.
5256 (<shift>di3_neon): Likewise.
5258 2018-03-20 Tom de Vries <tom@codesourcery.com>
5261 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5262 (nvptx_process_pars): Emit bar.sync asap and alap.
5264 2018-03-20 Tom de Vries <tom@codesourcery.com>
5267 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5268 seen_label if seen_label is already set.
5270 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5273 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5274 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5275 Use 1U instead of 1. Formatting fixes.
5278 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5279 instead of TREE_TYPE (s1) for the return value.
5281 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5283 PR tree-optimization/84946
5284 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5285 bitsize + bitsize in poly_uint64 rather than poly_int64.
5288 * dwarf2asm.c: Include fold-const.c.
5289 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5290 of decl rather than decl itself.
5292 PR rtl-optimization/84643
5293 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5295 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5298 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5299 calling assemble_variable.
5301 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5304 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5305 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5307 2018-03-19 Jim Wilson <jimw@sifive.com>
5310 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5311 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5312 (riscv_first_stack_step): Likewise.
5313 (riscv_option_override): Use STACK_BOUNDARY instead of
5315 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5317 (BIGGEST_ALIGNMENT): Set to 128.
5318 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5319 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5322 2018-03-19 Richard Biener <rguenther@suse.de>
5324 PR tree-optimization/84933
5325 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5326 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5328 2018-03-19 Richard Biener <rguenther@suse.de>
5330 PR tree-optimization/84859
5331 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5332 (cond_if_else_store_replacement): Perform sinking operation on
5333 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5334 Generalize what a BB with a single eligible store is.
5336 2018-03-19 Richard Biener <rguenther@suse.de>
5338 PR tree-optimization/84929
5339 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5340 chrec_is_positive against non-chrec arg.
5342 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5345 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5347 2018-03-18 Martin Liska <mliska@suse.cz>
5349 PR rtl-optimization/84635
5350 * regrename.c (build_def_use): Use matches_mode only when
5353 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5355 PR tree-optimization/84913
5356 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5357 vectorize chains of COND_EXPRs.
5359 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5361 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5363 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5365 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5367 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5369 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5371 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5372 Kito Cheng <kito.cheng@gmail.com>
5374 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5375 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5376 (nds32_adjust_reg_alloc_order): New function.
5377 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5379 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5381 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5382 nds32_print_operand, nds32_print_operand_address): Use
5383 HOST_WIDE_INT_PRINT_DEC instead.
5385 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5387 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5389 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5392 * config/i386/i386.c (initial_ix86_tune_features,
5393 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5395 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5396 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5397 rather than 1u << ix86_tune. Formatting fix.
5398 (ix86_option_override_internal): Change ix86_arch_mask from
5399 unsigned int to unsigned HOST_WIDE_INT, initialize to
5400 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5401 (ix86_function_specific_restore): Likewise.
5403 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5406 * postreload.c (reload_combine_recognize_pattern): Perform
5407 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5408 truncate_int_for_mode the result for the destination's mode.
5411 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5412 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5414 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5417 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5420 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5421 regno_allocno_class_array and sorted_pseudos.
5422 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5423 insns where regno is used.
5425 2018-03-16 Martin Liska <mliska@suse.cz>
5428 * multiple_target.c (create_dispatcher_calls): Redirect
5429 reference in the symbol table.
5431 2018-03-16 Martin Liska <mliska@suse.cz>
5434 * multiple_target.c (create_dispatcher_calls): Redirect also
5437 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5441 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5442 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5443 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5445 2018-03-16 Julia Koval <julia.koval@intel.com>
5447 * doc/invoke.texi (Skylake Server): Add CLWB.
5448 Cannonlake): Remove CLWB.
5450 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5452 PR tree-optimization/84841
5453 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5455 (FLOAT_ONE_CONST_TYPE): Define.
5456 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5457 (sort_by_operand_rank): Put entries with higher constant_type last
5458 rather than first to match comments.
5460 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5462 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5465 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5468 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5469 check and use address of target always.
5471 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5474 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5475 (indirect_thunk_bnd_needed): Likewise.
5476 (indirect_thunks_used): Likewise.
5477 (indirect_thunks_bnd_used): Likewise.
5478 (indirect_return_needed): New.
5479 (indirect_return_bnd_needed): Likewise.
5480 (output_indirect_thunk_function): Add a bool argument for
5482 (output_indirect_thunk_function): Don't generate alias for
5483 function return thunk.
5484 (ix86_code_end): Call output_indirect_thunk_function to generate
5485 function return thunks.
5486 (ix86_output_function_return): Set indirect_return_bnd_needed
5487 and indirect_return_needed instead of indirect_thunk_bnd_needed
5488 and indirect_thunk_needed.
5490 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5492 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5493 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5494 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5496 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5497 Paul Hua <paul.hua.gm@gmail.com>
5500 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5502 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5504 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5505 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5508 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5511 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5512 instead of GET_MODE_SIZE when comparing Units.
5514 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5517 * varasm.c (hash_section): Return an unchangeble hash value
5518 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5519 Return !aarch64_can_use_per_function_literal_pools_p ().
5521 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5524 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5525 rather than address of cmode as last argument to prepare_cmp_insn.
5527 2018-03-15 Julia Koval <julia.koval@intel.com>
5529 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5530 F_AVX512VNNI, F_AVX512BITALG): New.
5532 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5535 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5536 insn for floating-point loads and stores.
5538 2018-03-14 Carl Love <cel@us.ibm.com>
5540 * config/rs6000/rs6000-c.c: Add macro definitions for
5541 ALTIVEC_BUILTIN_VEC_PERMXOR.
5542 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5543 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5544 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5545 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5547 * config/doc/extend.texi: Add prototypes for vec_permxor.
5549 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5552 * diagnostic-show-locus.c (class layout_point): Convert m_line
5553 from int to linenum_type.
5554 (line_span::comparator): Use linenum "compare" function when
5555 comparing line numbers.
5556 (test_line_span): New function.
5557 (layout_range::contains_point): Convert param "row" from int to
5559 (layout_range::intersects_line_p): Likewise.
5560 (layout::will_show_line_p): Likewise.
5561 (layout::print_source_line): Likewise.
5562 (layout::should_print_annotation_line_p): Likewise.
5563 (layout::print_annotation_line): Likewise.
5564 (layout::print_leading_fixits): Likewise.
5565 (layout::annotation_line_showed_range_p): Likewise.
5566 (struct line_corrections): Likewise for field m_row.
5567 (line_corrections::line_corrections): Likewise for param "row".
5568 (layout::print_trailing_fixits): Likewise.
5569 (layout::get_state_at_point): Likewise.
5570 (layout::get_x_bound_for_row): Likewise.
5571 (layout::print_line): Likewise.
5572 (diagnostic_show_locus): Likewise for locals "last_line" and
5574 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5575 * input.c (selftest::test_linenum_comparisons): New function.
5576 (selftest::input_c_tests): Call it.
5577 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5578 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5579 * selftest.h (ASSERT_GT): New macro.
5580 (ASSERT_GT_AT): New macro.
5581 (ASSERT_LT): New macro.
5582 (ASSERT_LT_AT): New macro.
5584 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5586 PR rtl-optimization/84780
5587 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5589 2018-03-14 Martin Liska <mliska@suse.cz>
5591 * tree.c (record_node_allocation_statistics): Use
5592 get_stats_node_kind.
5593 (get_stats_node_kind): New function extracted from
5594 record_node_allocation_statistics.
5595 (free_node): Use get_stats_node_kind.
5597 2018-03-14 Richard Biener <rguenther@suse.de>
5599 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5600 that the value-set of ANTIC_IN doesn't grow.
5603 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5605 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5606 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5608 2018-03-14 Julia Koval <julia.koval@intel.com>
5610 * config.gcc (icelake-client, icelake-server): New.
5612 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5613 (initial_ix86_arch_features): Ditto.
5614 (PTA_SKYLAKE): Add SGX.
5615 (PTA_ICELAKE): Remove.
5616 (PTA_ICELAKE_CLIENT): New.
5617 (PTA_ICELAKE_SERVER): New.
5618 (ix86_option_override_internal): Split up icelake on icelake client and
5620 (get_builtin_code_for_version): Ditto.
5621 (fold_builtin_cpu): Ditto.
5622 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5623 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5624 * config/i386/i386.h (processor_type): Ditto.
5625 * doc/invoke.texi: Ditto.
5627 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5630 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5631 INTEGER_CST offset, add it together with bitpos / 8 and
5632 sign extend based on POINTER_SIZE.
5636 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5639 * config/i386/constraints.md (Yc): New register constraint.
5640 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5641 Use Yc constraint for alternative 2 of operand 0. Remove
5642 preferred_for_speed attribute.
5644 2018-03-14 Richard Biener <rguenther@suse.de>
5646 PR tree-optimization/84830
5647 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5648 with the old one to avoid oscillations.
5650 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5653 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5655 (assign_by_spills): Return a flag of reload assignment failure.
5656 Do not process the reload assignment failures. Do not spill other
5657 reload pseudos if they has the same reg class. Update n if
5659 (lra_assign): Add a return arg. Set up from the result of
5660 assign_by_spills call.
5661 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5662 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5663 usage_insns if it is not NULL.
5664 (spill_hard_reg_in_range): New function.
5665 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5666 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5667 function prototypes.
5668 (lra_assign): Change prototype.
5669 * lra.c (lra): Add code to deal with fails by splitting hard reg
5672 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5674 * config/riscv/riscv.opt (mrelax): New option.
5675 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5676 "norelax" when riscv_mrelax is disabled.
5677 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5679 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5682 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5683 reassociation for int modes.
5685 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5687 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5688 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5690 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5691 * config/aarch64/aarch64-sve.md
5692 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5693 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5694 (*extend<mode><Vwide>2): Rename to...
5695 (aarch64_sve_extend<mode><Vwide>2): ...this.
5696 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5697 renaming the old pattern to...
5698 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5700 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5701 define_expand, renaming the old pattern to...
5702 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5703 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5704 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5705 account when deciding which SVE instruction the optab should use.
5706 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5708 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5710 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5711 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5712 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5713 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5714 (tlsdesc_small_<mode>): Turn a define_expand and use
5715 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5716 (tlsdesc_small_advsimd_<mode>): ...this.
5717 (tlsdesc_small_sve_<mode>): New pattern.
5719 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5721 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5722 (UNSPEC_UMUL_HIGHPART): New constants.
5723 (MUL_HIGHPART): New int iteraor.
5724 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5725 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5727 (*<su>mul<mode>3_highpart): New define_insn.
5729 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5732 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5735 2018-03-13 Martin Liska <mliska@suse.cz>
5738 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5740 (sem_item_optimizer::~sem_item_optimizer): Release it.
5741 (sem_item_optimizer::merge_classes): Register variable aliases.
5742 (sem_item_optimizer::fixup_pt_set): New function.
5743 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5744 * ipa-icf.h: Declare new variables and functions.
5746 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5749 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5750 integer_pow2p@2 and test integer_pow2p in condition.
5751 (A < 0 ? C : 0): Similarly for @1.
5754 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5755 characters starting at p contain '\0' character, don't look beyond
5759 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5760 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5763 * reg-stack.c (change_stack): Change update_end var from int to
5764 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5765 also call set_block_for_insn on the newly added insns and rescan.
5768 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5769 on the last operand.
5772 * tree.c (stabilize_reference_1): Return save_expr (e) for
5773 STATEMENT_LIST even if it doesn't have side-effects.
5775 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5777 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5779 2018-03-12 Renlin Li <renlin.li@arm.com>
5781 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5782 aarch64_output_scalar_simd_mov_immediate.
5784 2018-03-12 Martin Sebor <msebor@redhat.com>
5786 PR tree-optimization/83456
5787 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5788 for perfectly overlapping calls to memcpy.
5789 (gimple_fold_builtin_memory_chk): Same.
5790 (gimple_fold_builtin_strcpy): Handle no-warning.
5791 (gimple_fold_builtin_stxcpy_chk): Same.
5792 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5794 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5796 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5797 parameter. Use it for SFmode.
5798 (rs6000_function_arg_advance_1): Adjust.
5799 (rs6000_function_arg): Adjust.
5800 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5802 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5804 PR rtl-optimization/84169
5805 PR rtl-optimization/84780
5806 * combine.c (can_combine_p): Check for a 2-insn combination whether
5807 the destination register is used between the two insns, too.
5809 2018-03-12 Richard Biener <rguenther@suse.de>
5811 PR tree-optimization/84803
5812 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5813 for refs DR analysis didn't process.
5815 2018-03-12 Richard Biener <rguenther@suse.de>
5817 PR tree-optimization/84777
5818 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5819 force-vectorize loops ignore whether we are optimizing for size.
5821 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5823 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5824 (TARGET_MD_ASM_ADJUST): Define.
5826 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5827 Kito Cheng <kito.cheng@gmail.com>
5828 Chung-Ju Wu <jasonwucj@gmail.com>
5830 * config/nds32/nds32.c (nds32_compute_stack_frame,
5831 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5832 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5833 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5834 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5835 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5836 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5837 * config/nds32/nds32.md (prologue, epilogue): Use macro
5838 NDS32_V3PUSH_AVAILABLE_P to do checking.
5840 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5843 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5844 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5845 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5846 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5847 addition of most attributes on !orig_type_die or the attribute not
5848 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5850 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5851 Chung-Ju Wu <jasonwucj@gmail.com>
5853 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5855 * config/nds32/nds32.opt (mvh): New option.
5857 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5858 Chung-Ju Wu <jasonwucj@gmail.com>
5860 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5862 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5863 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5866 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5867 Chung-Ju Wu <jasonwucj@gmail.com>
5869 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5871 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5872 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5874 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5875 Kito Cheng <kito.cheng@gmail.com>
5876 Chung-Ju Wu <jasonwucj@gmail.com>
5878 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5879 UNSPEC_FFMISM and UNSPEC_FLMISM.
5880 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5881 for ffb, ffmism and flmism.
5882 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5883 (unspec_ffmism): Ditto.
5884 (unspec_flmism): Ditto.
5885 (nds32_expand_builtin_impl): Check if string extension is available.
5886 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5887 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5889 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5892 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5895 * lra-assigns.c (assign_by_spills): Return a flag of reload
5896 assignment failure. Do not process the reload assignment
5897 failures. Do not spill other reload pseudos if they has the same
5899 (lra_assign): Add a return arg. Set up from the result of
5900 assign_by_spills call.
5901 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5902 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5903 usage_insns if it is not NULL.
5904 (spill_hard_reg_in_range): New function.
5905 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5906 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5907 function prototypes.
5908 (lra_assign): Change prototype.
5909 * lra.c (lra): Add code to deal with fails by splitting hard reg
5912 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5915 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5917 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5920 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5921 (dw_val_node): Add val_symbolic_view.
5922 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5923 (symview_upper_bound): New.
5924 (new_line_info_table): Initialize symviews_since_reset.
5925 (dwarf2out_source_line): Count symviews_since_reset and set
5926 symview_upper_bound.
5927 (dw_val_equal_p): Handle symview.
5928 (add_AT_symview): New.
5929 (print_dw_val): Handle symview.
5930 (attr_checksum, attr_checksum_ordered): Likewise.
5931 (same_dw_val_p, size_of_die): Likewise.
5932 (value_format, output_die): Likewise.
5933 (add_high_low_attributes): Use add_AT_symview for entry_view.
5934 (dwarf2out_finish): Reset symview_upper_bound, clear
5937 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5940 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5941 Add strict argument and use it.
5942 (rs6000_split_multireg_move): Update for new strict argument.
5943 (mem_operand_gpr): Disallow all non-offsettable addresses.
5944 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5946 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5949 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5950 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5951 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5954 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5955 decl, use remap_type if we want to use the type.
5957 2018-03-09 Martin Sebor <msebor@redhat.com>
5959 PR tree-optimization/84526
5960 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5962 (builtin_access::generic_overlap): Be prepared to handle non-array
5965 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5967 PR rtl-optimization/84682
5968 * lra-constraints.c (process_address_1): Check is_address flag
5969 for address constraints.
5970 (process_alt_operands): Likewise.
5971 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5972 preprocess_constraints.
5973 * recog.h (preprocess_constraints): Add oploc parameter.
5975 * recog.c (preprocess_constraints): Test address_operand for
5976 CT_ADDRESS constraints.
5978 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5981 * lra-assigns.c (assign_by_spills): Return a flag of reload
5982 assignment failure. Do not process the reload assignment
5983 failures. Do not spill other reload pseudos if they has the same
5985 (lra_assign): Add a return arg. Set up from the result of
5986 assign_by_spills call.
5987 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5988 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5989 usage_insns if it is not NULL.
5990 (spill_hard_reg_in_range): New function.
5991 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5992 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5993 function prototypes.
5994 (lra_assign): Change prototype.
5995 * lra.c (lra): Add code to deal with fails by splitting hard reg
5998 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6001 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6002 Accept complain bool parameter. Only emit errors if it is true.
6003 (arm_parse_cpu_option_name): Likewise.
6004 (arm_target_thumb_only): Adjust callers of the above.
6005 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6006 prototype to take a default true bool parameter.
6007 (arm_parse_arch_option_name): Likewise.
6009 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6010 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6014 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6015 * configure: Regenerate.
6016 * configure.ac ("linker --version-script option"): New.
6017 ("linker soname option"): New.
6019 2018-03-09 Richard Biener <rguenther@suse.de>
6021 PR tree-optimization/84775
6022 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6023 immediate uses of predicate stmts and mark them modified.
6026 PR tree-optimization/84178
6027 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6029 (version_loop_for_if_conversion): Delay update_ssa call.
6030 (tree_if_conversion): Delay update_ssa until after predicate
6033 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6036 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6037 when the function accesses prior frames.
6039 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6042 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6043 gen_llsym, otherwise call maybe_gen_llsym.
6046 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6047 has ',' character inside of it.
6049 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6052 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6053 as clobbering CC_REGNUM.
6055 2018-03-08 Richard Biener <rguenther@suse.de>
6058 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6059 (follow_copies_to_constant): Do not follow SSA names registered
6062 2018-03-08 Richard Biener <rguenther@suse.de>
6064 PR tree-optimization/84178
6065 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6067 (version_loop_for_if_conversion): Delay update_ssa call.
6068 (tree_if_conversion): Delay update_ssa until after predicate
6071 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6073 PR tree-optimization/84178
6074 * tree-if-conv.c (release_bb_predicate): Remove the
6075 the assertion that the stmts have NULL use_ops.
6076 Discard the statements, asserting that they haven't
6077 yet been added to a BB.
6079 2018-03-08 Richard Biener <rguenther@suse.de>
6081 PR tree-optimization/84746
6082 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6083 (phi_translate): Pass in destination ANTIC_OUT set.
6084 (phi_translate_1): Likewise. For a simplified result lookup
6085 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6086 (phi_translate_set): Adjust.
6087 (do_pre_regular_insertion): Likewise.
6088 (do_pre_partial_partial_insertion): Likewise.
6090 2018-03-08 Martin Liska <mliska@suse.cz>
6092 PR gcov-profile/84735
6093 * doc/gcov.texi: Document usage of profile files.
6094 * gcov-io.h: Document changes in the format.
6096 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6100 * dwarf2out.c (struct dw_line_info_table): Update comments for
6102 (FORCE_RESET_NEXT_VIEW): New.
6103 (FORCE_RESETTING_VIEW_P): New.
6104 (RESETTING_VIEW_P): Check for -1 too.
6105 (ZERO_VIEW_P): Likewise.
6106 (new_line_info_table): Force-reset next view.
6107 (dwarf2out_begin_function): Likewise.
6108 (dwarf2out_source_line): Simplify zero_view_p initialization.
6109 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6110 view directly. Omit view when omitting .loc at line 0.
6112 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6114 PR tree-optimization/84740
6115 * tree-switch-conversion.c (process_switch): Call build_constructors
6116 only if info.phi_count is non-zero.
6118 PR tree-optimization/84739
6119 * tree-tailcall.c (find_tail_calls): Check call arguments against
6120 DECL_ARGUMENTS (current_function_decl) rather than
6121 DECL_ARGUMENTS (func) when checking for tail recursion.
6123 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6125 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6126 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6127 Volker Reichelt's entry and add entries for people that perform
6128 GCC fuzzy testing and report numerous bugs.
6130 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6133 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6134 readonly data in sdata, if that is disabled.
6135 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6136 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6137 -mreadonly-in-sdata option.
6139 2018-03-07 Martin Sebor <msebor@redhat.com>
6141 PR tree-optimization/84468
6142 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6143 basic block when looking for nul assignment.
6145 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6148 * except.h (output_function_exception_table): Adjust prototype.
6149 * except.c (output_function_exception_table): Remove FNNAME parameter
6150 and add SECTION parameter. Ouput one part of the table at a time.
6151 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6152 the first part of the exception table and emit unwind directives.
6153 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6154 (i386_pe_seh_cold_init): Likewise.
6155 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6156 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6157 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6158 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6159 * config/i386/winnt.c: Include except.h.
6160 (struct seh_frame_state): Add reg_offset, after_prologue and
6161 in_cold_section fields.
6162 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6163 (i386_pe_seh_cold_init): New function.
6164 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6165 to seh->in_cold_section.
6166 (seh_emit_push): Record the offset of the push.
6167 (seh_emit_save): Record the offet of the save.
6168 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6169 Test seh->after_prologue to disregard the epilogue.
6170 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6171 (i386_pe_end_cold_function): New function.
6173 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6176 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6177 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6180 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6182 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6183 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6186 * multiple_target.c: Include tree-inline.h and intl.h.
6187 (expand_target_clones): Diagnose and fail if node->definition and
6188 !tree_versionable_function_p (node->decl).
6190 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6192 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6194 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6195 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6196 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6198 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6201 * combine.c (try_combine): Use reg_or_subregno instead of handling
6202 just paradoxical SUBREGs and REGs.
6204 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6206 * config/arc/arc.c (arc_finalize_pic): Remove function.
6207 (arc_must_save_register): We use single base PIC register, remove
6208 checks to save/restore the PIC register.
6209 (arc_expand_prologue): Likewise.
6210 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6212 (arc_verify_short): Likewise.
6213 (arc_attr_type): Likewise.
6214 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6215 (walk_stores): Likewise.
6216 (arc_address_cost): Make it static.
6217 (arc_verify_short): Likewise.
6218 (branch_dest): Likewise.
6219 (arc_attr_type): Likewise.
6220 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6221 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6222 (arc_final_prescan_insn): Remove inserting the nops due to
6223 hardware hazards. It is done in reorg step.
6224 (insn_length_variant_t): Remove.
6225 (insn_length_parameters_t): Likewise.
6226 (arc_insn_length_parameters): Likewise.
6227 (arc_get_insn_variants): Likewise.
6228 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6230 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6233 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6236 PR tree-optimization/84687
6237 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6239 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6241 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6243 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6244 Rename to ppc_speculation_barrier.
6245 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6246 __builtin_ppc_speculation_barrier.
6248 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6251 * combine.c (combine_simplify_rtx): Don't try to simplify if
6252 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6255 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6257 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6258 to 32 bytes when compiling for POWER9.
6260 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6263 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6264 regparm >= 3 with no arg reg available also for calls with
6265 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6268 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6270 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6272 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6275 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6277 2018-03-05 Richard Biener <rguenther@suse.de>
6279 PR tree-optimization/84486
6280 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6281 When inserting a __builtin_assume_aligned call set the LHS
6282 SSA name alignment info accordingly.
6284 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6286 PR tree-optimization/84114
6287 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6288 Avoid reassociation of FLOAT_MODE addition.
6290 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6292 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6293 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6294 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6295 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6296 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6297 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6298 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6300 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6301 __builtin_ia32_wbinvd): New builtins.
6302 (SPECIAL_ARGS2): New.
6303 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6304 (SPECIAL_ARGS2): New.
6305 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6306 (ix86_valid_target_attribute_inner_p): Ditto.
6307 (ix86_init_mmx_sse_builtins): Add special_args2.
6308 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6309 TARGET_WBNOINVD_P): New.
6310 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6311 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6312 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6313 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6314 * config/i386/pconfigintrin.h: New file.
6315 * config/i386/wbnoinvdintrin.h: Ditto.
6316 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6317 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6319 2018-03-05 Richard Biener <rguenther@suse.de>
6321 PR tree-optimization/84670
6322 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6324 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6325 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6326 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6327 doesn't grow if all successors (recursively) were visited at least
6330 2018-03-05 Richard Biener <rguenther@suse.de>
6332 PR tree-optimization/84650
6333 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6334 if executed in the loop pipeline.
6336 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6338 * doc/configfiles.texi (Configuration Files): Move info about
6339 conditionalizing $target-protos.h to...
6340 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6341 differs from $target-protos.h.
6343 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6344 Chung-Ju Wu <jasonwucj@gmail.com>
6346 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6347 * config/nds32/nds32-multiple.md (setmemsi): Define.
6348 * config/nds32/nds32-memory-manipulation.c
6349 (nds32_gen_dup_4_byte_to_word_value): New.
6350 (emit_setmem_word_loop): New.
6351 (emit_setmem_byte_loop): New.
6352 (nds32_expand_setmem_loop): New.
6353 (nds32_expand_setmem_loop_v3m): New.
6354 (nds32_expand_setmem_unroll): New.
6355 (nds32_expand_setmem): New.
6357 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6358 Chung-Ju Wu <jasonwucj@gmail.com>
6360 * config/nds32/nds32-memory-manipulation.c
6361 (nds32_emit_load_store): New.
6362 (nds32_emit_post_inc_load_store): New.
6363 (nds32_emit_mem_move): New.
6364 (nds32_emit_mem_move_block): New.
6365 (nds32_expand_movmemsi_loop_unknown_size): New.
6366 (nds32_expand_movmemsi_loop_known_size): New.
6367 (nds32_expand_movmemsi_loop): New.
6368 (nds32_expand_movmemsi_unroll): New.
6369 (nds32_expand_movmemqi): Rename ...
6370 (nds32_expand_movmemsi): ... to this.
6371 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6372 (movmemsi): ... to this.
6373 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6374 (nds32_expand_movmemsi): ... to this.
6376 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6377 Monk Chiang <sh.chiang04@gmail.com>
6378 Chung-Ju Wu <jasonwucj@gmail.com>
6380 * config/nds32/nds32-protos.h
6381 (nds32_expand_load_multiple): New arguments.
6382 (nds32_expand_store_multiple): Ditto.
6383 (nds32_valid_multiple_load_store): Rename ...
6384 (nds32_valid_multiple_load_store_p): ... to this.
6385 * config/nds32/nds32-memory-manipulation.c
6386 (nds32_expand_load_multiple): Refine implementation.
6387 (nds32_expand_store_multiple): Ditto.
6388 * config/nds32/nds32-multiple.md
6389 (load_multiple): Update nds32_expand_load_multiple interface.
6390 (store_multiple): Update nds32_expand_store_multiple interface.
6391 * config/nds32/nds32-predicates.c
6392 (nds32_valid_multiple_load_store): Rename ...
6393 (nds32_valid_multiple_load_store_p): ... to this and refine
6395 * config/nds32/predicates.md
6396 (nds32_load_multiple_and_update_address_operation): New predicate.
6397 (nds32_store_multiple_and_update_address_operation): New predicate.
6399 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6400 Chung-Ju Wu <jasonwucj@gmail.com>
6402 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6403 (combo): New attribute.
6404 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6406 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6408 * config/nds32/nds32.opt: Change -mcmodel= default value.
6410 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6411 Monk Chiang <sh.chiang04@gmail.com>
6412 Chung-Ju Wu <jasonwucj@gmail.com>
6414 * config/nds32/constants.md (unspec_element): New enum.
6415 * config/nds32/constraints.md (Umw): New constraint.
6416 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6417 * config/nds32/nds32-intrinsic.md: Likewise.
6418 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6419 (nds32_valid_smw_lwm_base_p): New.
6420 (nds32_output_smw_single_word): New.
6421 (nds32_output_lmw_single_word): New.
6422 (nds32_expand_unaligned_load): New.
6423 (nds32_expand_unaligned_store): New.
6424 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6425 (nds32_output_smw_single_word): Declare.
6426 (nds32_output_lmw_single_word): Declare.
6427 (nds32_expand_unaligned_load): Declare.
6428 (nds32_expand_unaligned_store): Declare.
6429 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6430 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6431 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6432 NDS32_BUILTIN_UASTORE_DW.
6433 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6436 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6437 Kito Cheng <kito.cheng@gmail.com>
6438 Chung-Ju Wu <jasonwucj@gmail.com>
6440 * config/nds32/nds32-intrinsic.c
6441 (nds32_expand_builtin_null_ftype_reg): Delete.
6442 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6443 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6444 (nds32_read_argument): New.
6445 (nds32_legitimize_target): Ditto.
6446 (nds32_legitimize_argument): Ditto.
6447 (nds32_check_constant_argument): Ditto.
6448 (nds32_expand_unop_builtin): Ditto.
6449 (nds32_expand_unopimm_builtin): Ditto.
6450 (nds32_expand_binop_builtin): Ditto.
6451 (nds32_builtin_decl_impl): Ditto.
6452 (builtin_description): Ditto.
6453 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6454 (nds32_init_builtins_impl): Ditto.
6455 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6456 (nds32_builtin_decl): New.
6457 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6458 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6460 2018-03-02 Jeff Law <law@redhat.com>
6462 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6463 (redundant_insn, fill_simple_delay_slots): Likewise.
6464 (fill_slots_from_thread): Likewise.
6465 * resource.c (mark_referenced_resources): Likewise.
6466 (mark_set_resources, find_dead_or_set_registers): Likewise.
6468 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6470 * substring-locations.h (format_warning_va): Formatting fix for
6472 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6474 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6475 * substring-locations.c: Include intl.h.
6476 (format_warning_va): Turned into small wrapper around
6477 format_warning_n_va, renamed to ...
6478 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6479 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6481 (format_warning_at_substring_n): New function.
6482 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6483 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6484 format_warning_at_substring with just a shorter name instead of
6485 const function pointer.
6486 (fmtwarn_n): New function.
6487 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6488 appropriate, get rid of all the fmtstr temporaries, move conditionals
6489 with G_() wrapped string literals directly into fmtwarn arguments,
6490 cast dir.len to (int), formatting fixes.
6492 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6494 * doc/invoke.texi: Remove "Cilk Plus" references.
6496 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6497 Richard Biener <rguenther@suse.de>
6500 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6501 for error or warning attributes if CALL_FROM_THUNK_P is set.
6504 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6507 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6508 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6511 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6512 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6514 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6515 instead of -1U in last predictors element's probability member.
6517 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6520 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6521 arguments if they are comparable.
6523 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6525 PR tree-optimization/84634
6526 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6527 masks and masked_loop_p with a single loop_masks, making sure it's
6528 null for bb vectorization.
6530 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6532 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6533 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6534 loop->force_vectorize to check whether there is no alias.
6536 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6539 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6541 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6543 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6544 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6545 instead of a loop around prev_real_insn.
6546 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6550 * config/i386/i386.c (ix86_print_operand): Use conditional
6551 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6554 2018-03-02 Richard Biener <rguenther@suse.de>
6556 PR tree-optimization/84427
6557 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6558 (bitmap_set_subtract_values): Rewrite to handle multiple
6561 (prune_clobbered_mems): Likewise.
6562 (phi_translate): Take edge instead of pred/phiblock.
6563 (phi_translate_1): Likewise.
6564 (phi_translate_set): Likewise. Insert all translated
6565 exprs for a value into the set, keeping possibly multiple
6566 expressions per value.
6567 (compute_antic_aux): Adjust for phi_translate changes.
6568 When intersecting union the expressions and prune those
6569 not in the final value set, keeping possibly multiple
6570 expressions per value. Do not use value-insertion
6571 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6572 all expressions. Add verification that the value-sets
6573 only shrink during iteration.
6574 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6575 (do_pre_regular_insertion): Likewise.
6576 (do_pre_partial_partial_insertion): Likewise.
6578 2018-03-02 Richard Biener <rguenther@suse.de>
6581 * config/darwin.c (saved_debug_info_level): New static global.
6582 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6583 (darwin_asm_lto_end): Restore debug info generation settings.
6585 2018-03-01 Martin Liska <mliska@suse.cz>
6588 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6591 2018-03-01 Richard Biener <rguenther@suse.de>
6594 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6595 type annotation with LTO when debug was disabled at compile-time.
6597 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6599 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6601 (mips_final_postscan_insn): Likewise.
6603 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6605 PR rtl-optimization/84528
6606 * alias.c (init_alias_target): Add commentary.
6607 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6608 a unique base value if the frame pointer is not eliminated
6609 to the stack pointer.
6611 2018-03-01 Tom de Vries <tom@codesourcery.com>
6613 PR rtl-optimization/83327
6614 * lra-int.h (hard_regs_spilled_into): Declare.
6615 * lra.c (hard_regs_spilled_into): Define.
6616 (init_reg_info): Init hard_regs_spilled_into.
6617 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6618 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6619 (process_bb_lives): Handle hard_regs_spilled_into.
6620 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6621 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6623 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6625 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6626 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6627 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6628 * config/rs6000/aix72.h: New file.
6630 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6632 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6633 instead of warning_at with conditional singular and plural messages
6637 * stor-layout.c (update_alignment_for_field): For
6638 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6639 && !DECL_PACKED (field), do the alignment update, just use
6640 only desired_align instead of MAX (type_align, desired_align)
6642 (place_field): Don't do known_align < desired_align handling
6643 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6644 is non-NULL, instead do it after rli->prev_field handling and
6645 only if not within a bitfield word. For DECL_PACKED (field)
6646 use type_align of BITS_PER_UNIT.
6648 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6650 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6651 superfluous parentheses and trailing spaces.
6653 2018-02-28 Richard Biener <rguenther@suse.de>
6655 PR tree-optimization/84584
6656 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6657 SCoPs with fake exit edge.
6659 2018-02-28 Martin Liska <mliska@suse.cz>
6662 * timevar.c (timer::print): Fix format to properly print 100%
6665 2018-02-28 Richard Biener <rguenther@suse.de>
6668 * genmatch.c (capture_info::walk_match): Do not mark
6669 captured expressions without operands as expr_p given
6670 they act more like predicates and should be subject to
6671 "lost tail" side-effect preserving.
6673 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6675 PR rtl-optimization/81611
6676 * auto-inc-dec.c (attempt_change): Move dead note from
6677 mem_insn if it's the next use of regno
6678 (find_address): Take address use of reg holding
6679 non-incremented value. Add parm to limit search to the named
6681 (merge_in_block): Attempt to use a mem insn that is the next
6682 use of the original regno.
6684 2018-02-27 Martin Sebor <msebor@redhat.com>
6687 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6688 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6690 2018-02-27 Martin Sebor <msebor@redhat.com>
6692 PR translation/84207
6693 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6694 n argument to unsigned HOST_WIDE_INT.
6695 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6696 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6697 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6698 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6700 2018-02-27 Richard Biener <rguenther@suse.de>
6702 PR tree-optimization/84512
6703 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6704 Do not use the estimate returned from record_stmt_cost for
6705 the scalar iteration cost but sum properly using add_stmt_cost.
6707 2018-02-27 Richard Biener <rguenther@suse.de>
6709 PR tree-optimization/84466
6710 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6711 Adjust last change to less strictly validate use operands.
6713 2018-02-27 Martin Liska <mliska@suse.cz>
6715 PR gcov-profile/84548
6716 * gcov.c (process_file): Allow partial overlap and consider it
6717 also as group functions.
6718 (output_lines): Properly calculate range of lines for a group.
6720 2018-02-27 Martin Liska <mliska@suse.cz>
6722 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6723 'ggc' suffixes. Change first column width.
6724 (timer::print): Fix formatting of the column.
6726 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6728 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6729 preserve inline entry blocks for the sake of debug inline
6730 entry point markers alone.
6731 (remove_unused_locals): Suggest in comments a better place to
6732 force the preservation of inline entry blocks that are
6733 otherwise unused, but do not preserve them.
6735 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6737 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6739 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6742 * config/i386/constraints.md (Bs): Replace
6743 ix86_indirect_branch_register with
6744 TARGET_INDIRECT_BRANCH_REGISTER.
6746 * config/i386/i386.md (indirect_jump): Likewise.
6747 (tablejump): Likewise.
6748 (*sibcall_memory): Likewise.
6749 (*sibcall_value_memory): Likewise.
6750 Peepholes of indirect call and jump via memory: Likewise.
6751 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6752 (*sibcall_value_GOT_32): Likewise.
6753 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6754 (GOT_memory_operand): Likewise.
6755 (call_insn_operand): Likewise.
6756 (sibcall_insn_operand): Likewise.
6757 (GOT32_symbol_operand): Likewise.
6758 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6760 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6762 PR rtl-optimization/83496
6763 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6764 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6765 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6766 redundant insn, if any.
6767 (relax_delay_slots): Likewise.
6768 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6770 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6772 PR tree-optimization/83965
6773 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6774 that grouped statements are part of a reduction chain. Return
6775 true if the statement is not marked as a reduction itself but
6777 (vect_recog_dot_prod_pattern): Don't check whether the statement
6778 is part of a group here.
6779 (vect_recog_sad_pattern): Likewise.
6780 (vect_recog_widen_sum_pattern): Likewise.
6782 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6785 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6787 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6790 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6792 (ix86_output_indirect_function_return): New prototype.
6793 (ix86_split_simple_return_pop_internal): Likewise.
6794 * config/i386/i386.c (indirect_return_via_cx): New.
6795 (indirect_return_via_cx_bnd): Likewise.
6796 (indirect_thunk_name): Handle return va CX_REG.
6797 (output_indirect_thunk_function): Create alias for
6798 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6799 (ix86_output_indirect_jmp): Remove the bool argument.
6800 (ix86_output_indirect_function_return): New function.
6801 (ix86_split_simple_return_pop_internal): Likewise.
6802 * config/i386/i386.md (*indirect_jump): Don't pass false
6803 to ix86_output_indirect_jmp.
6804 (*tablejump_1): Likewise.
6805 (simple_return_pop_internal): Change it to define_insn_and_split.
6806 Call ix86_split_simple_return_pop_internal to split it for
6808 (simple_return_indirect_internal): Call
6809 ix86_output_indirect_function_return instead of
6810 ix86_output_indirect_jmp.
6812 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6815 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6816 memset and value initialization afterwards.
6818 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6820 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6822 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6825 * common/config/aarch64/aarch64-common.c
6826 (aarch_option_optimization_table[]): Switch
6827 off fomit-frame-pointer
6829 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6830 Chung-Ju Wu <jasonwucj@gmail.com>
6832 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6834 (store_multiple): Ditto.
6836 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6838 * config.gcc: Add --with-cpu support for nds32 target.
6839 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6840 * config/nds32/nds32.opt: Add -mcpu= option.
6842 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6844 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6845 isel=yes): Warn for these deprecated options.
6847 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6849 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6850 ISA_2_5_MASKS_EMBEDDED.
6852 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6854 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6855 p->max as pointers rather than using iterative_hash_expr.
6857 2018-02-23 Carl Love <cel@us.ibm.com>
6859 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6860 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6862 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6863 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6864 P8V_BUILTIN_VEC_VUNSIGNED2.
6866 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6869 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6870 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6872 * lra-constraints.c (curr_insn_transform): Set up
6873 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6875 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6876 * lra-eliminations.c (spill_pseudos): Ditto.
6877 (process_insn_for_elimination): Ditto.
6878 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6879 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6880 LRA_NON_CLOBBERED_ALT.
6882 2018-02-22 Martin Sebor <msebor@redhat.com>
6884 PR tree-optimization/84480
6885 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6886 to maybe_diag_stxncpy_trunc. Call it.
6887 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6888 from gimple_fold_builtin_strcpy. Print inlining stack.
6889 (handle_builtin_stxncpy): Print inlining stack.
6890 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6892 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6895 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6896 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6897 and -fcheck-pointer-bounds are used together.
6898 (indirect_thunk_prefix): New enum.
6899 (indirect_thunk_need_prefix): New function.
6900 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6901 "_nt" instead of "_bnd" for NOTRACK prefix.
6902 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6903 (output_indirect_thunk_function): Likewise.
6905 (ix86_code_end): Update output_indirect_thunk_function calls.
6906 (ix86_output_indirect_branch_via_reg): Replace
6907 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6908 (ix86_output_indirect_branch_via_push): Likewise.
6909 (ix86_output_function_return): Likewise.
6910 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6911 incompatible with -fcf-protection=branch and
6912 -fcheck-pointer-bounds.
6914 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6917 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6918 Change gcc_assert call to output_operand_lossage.
6920 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6922 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6924 2018-02-22 DJ Delorie <dj@redhat.com>
6925 Sebastian Perta <sebastian.perta@renesas.com>
6926 Oleg Endo <olegendo@gcc.gnu.org>
6928 * config/rx/rx.c (rx_rtx_costs): New function.
6929 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6931 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6933 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6935 2018-02-22 Martin Liska <mliska@suse.cz>
6938 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6939 Add "native" as a possible value.
6941 2018-02-22 Martin Liska <mliska@suse.cz>
6944 * config/i386/i386.c (ix86_option_override_internal):
6945 Add "native" as a possible value for -march and -mtune.
6947 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6950 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6951 to all type variants.
6953 PR tree-optimization/84503
6954 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6955 width as info->bitpos + info->bitsize - start.
6956 (merged_store_group::merge_overlapping): Simplify width computation.
6957 (check_no_overlap): New function.
6958 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6959 start + width and last_order of the group, fail if check_no_overlap
6961 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6962 to group if check_no_overlap fails.
6964 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6966 * config/rs6000/altivec.md: Delete contraint arguments to
6967 define_expand, define_split, and define_peephole2, and in
6968 define_insn_and_split if always unused.
6969 * config/rs6000/darwin.md: Ditto.
6970 * config/rs6000/dfp.md: Ditto.
6971 * config/rs6000/rs6000.md: Ditto.
6972 * config/rs6000/sync.md: Ditto.
6973 * config/rs6000/vector.md: Ditto.
6974 * config/rs6000/vsx.md: Ditto.
6976 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6978 * config/rs6000/altivec.md: Write output control strings as braced
6979 blocks instead of double-quoted strings.
6980 * config/rs6000/darwin.md: Ditto.
6981 * config/rs6000/rs6000.md: Ditto.
6982 * config/rs6000/vector.md: Ditto.
6983 * config/rs6000/vsx.md: Ditto.
6985 2018-02-21 Jason Merrill <jason@redhat.com>
6987 PR c++/84314 - ICE with templates and fastcall attribute.
6988 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6990 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6992 * ipa-cp.c (determine_versionability): Fix comment typos.
6994 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6997 * ipa-cp.c (determine_versionability): Do not version functions caling
7000 2018-02-21 Martin Liska <mliska@suse.cz>
7003 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7004 Add "native" as a possible value.
7005 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7006 the macro when native cpu detection is available.
7008 2018-02-21 Martin Liska <mliska@suse.cz>
7011 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7012 Add "native" as a possible value.
7013 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7014 when native cpu detection is available.
7016 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7017 Martin Sebor <msebor@redhat.com>
7019 PR tree-optimization/84478
7020 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7022 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7023 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7024 support which is conservatively correct, for 2 only stay conservative
7025 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7026 argument to the 2 argument get_range_strlen, adjust 6 arg
7027 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7029 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7030 (gimple_fold_builtin_strlen): Pass true as last argument to
7033 2018-02-20 Martin Sebor <msebor@redhat.com>
7036 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7037 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7038 (builtin_memref::builtin_memref): Factor out parts into
7039 set_base_and_offset and call it.
7041 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7044 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7045 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7046 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7047 search at the associated MODE_INT.
7049 2018-02-20 Jeff Law <law@redhat.com>
7052 PR tree-optimization/81592
7054 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7055 for range data rather than using global data.
7056 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7057 range data rather than using global data.
7058 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7059 pass it to children as needed.
7060 (struct directive::fmtresult): Similarly.
7061 (struct directive::set_width): Similarly.
7062 (struct directive::set_precision): Similarly.
7063 (format_integer, format_directive, parse_directive): Similarly.
7064 (format_none): Accept unnamed vr_values parameter.
7065 (format_percent, format_floating, format_character): Similarly.
7066 (format_string, format_plain): Similarly.
7067 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7068 the EVRP range analyzer for range data rather than using global data.
7069 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7070 gimple-ssa-evrp-analyze.h
7071 (class sprintf_dom_walker): Add after_dom_children member function.
7072 Add evrp_range_analyzer member.
7073 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7074 range analyzer as needed.
7075 (sprintf_dom_walker::after_dom_children): New member function.
7076 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7078 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7079 (evrp_range_analyzer::pop_to_marker): Likewise.
7081 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7083 PR tree-optimization/84419
7084 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7085 with the required type if its current type is compatible but
7088 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7091 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7092 after vectorization.
7094 2018-02-20 Martin Liska <mliska@suse.cz>
7097 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7098 possible values if we don't have a hint.
7100 2018-02-20 Martin Liska <mliska@suse.cz>
7104 * final.c (shorten_branches): Build align_tab array with one
7106 * opts.c (finish_options): Add alignment option limit check.
7107 (MAX_CODE_ALIGN): Likewise.
7108 (MAX_CODE_ALIGN_VALUE): Likewise.
7109 * doc/invoke.texi: Document maximum allowed option value for
7110 all -falign-* options.
7112 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7115 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7116 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7117 * var-tracking.c (emit_note_insn_var_location): Remove all references
7118 to NOTE_INSN_CALL_ARG_LOCATION.
7119 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7120 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7121 Use copy_rtx_if_shared.
7122 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7123 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7124 (dwarf2out_var_location): Remove handling of
7125 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7127 * final.c (final_scan_insn): Remove all references to
7128 NOTE_INSN_CALL_ARG_LOCATION.
7129 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7130 before dumping final insns.
7131 * except.c (emit_note_eh_region_end): Remove all references to
7132 NOTE_INSN_CALL_ARG_LOCATION.
7133 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7134 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7135 * config/arc/arc.c (hwloop_optimize): Likewise.
7136 * config/arm/arm.c (create_fix_barrier): Likewise.
7137 * config/s390/s390.c (s390_chunkify_start): Likewise.
7138 * config/sh/sh.c (find_barrier): Likewise.
7139 * config/i386/i386.c (rest_of_insert_endbranch,
7140 ix86_seh_fixup_eh_fallthru): Likewise.
7141 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7142 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7143 * config/frv/frv.c (frv_function_prologue): Likewise.
7144 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7146 (note_outside_basic_block_p): Remove all references to
7147 NOTE_INSN_CALL_ARG_LOCATION.
7148 * gengtype.c (adjust_field_rtx_def): Likewise.
7149 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7151 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7152 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7155 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7158 PR tree-optimization/84452
7159 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7160 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7163 2018-02-19 Martin Liska <mliska@suse.cz>
7166 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7168 2018-02-19 Martin Liska <mliska@suse.cz>
7169 Richard Sandiford <richard.sandiford@linaro.org>
7171 PR tree-optimization/82491
7172 * gimple-fold.c (get_base_constructor): Make earlier bail out
7175 2018-02-19 Carl Love <cel@us.ibm.com>
7177 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7178 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7180 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7181 P8V_BUILTIN_VEC_NEG.
7183 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7185 * config/rl78/rl78.md (movdf): New define expand.
7187 2018-02-19 Martin Liska <mliska@suse.cz>
7190 * doc/invoke.texi: Fix typo.
7191 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7193 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7195 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7196 handle rs6000_single_float and rs6000_double_float specially for
7199 2018-02-16 Jeff Law <law@redhat.com>
7201 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7204 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7207 * ipa-inline.c (inline_small_functions): Fix a typo.
7209 2018-02-16 Nathan Sidwell <nathan@acm.org>
7211 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7213 2018-02-16 Carl Love <cel@us.ibm.com>
7215 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7216 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7217 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7218 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7219 expansion to P8V_BUILTIN_VEC_FLOAT2.
7221 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7223 PR rtl-optimization/70023
7224 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7225 src_regno into account.
7227 2018-02-16 Carl Love <cel@us.ibm.com>
7229 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7230 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7231 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7232 * config/rs6000/rs6000.c: Remove case statements for
7233 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7234 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7235 and P9V_BUILTIN_VEC_VINSERT4B.
7236 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7237 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7238 * config/rs6000/vsx.md:
7239 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7242 2018-02-16 Carl Love <cel@us.ibm.com>
7244 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7246 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7248 * config/rs6000/rs6000-c.c: Add the definitions for
7249 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7250 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7251 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7252 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7253 definition for insert4b and define insn *insert3b_internal.
7254 * doc/extend.texi: Add documentation for vec_extract4b.
7256 2018-02-16 Nathan Sidwell <nathan@acm.org>
7258 * doc/extend.texi (Backwards Compatibility): Mention friend
7259 injection. Note for-scope is deprecated.
7260 * doc/invoke.texi (-ffriend-injection): Deprecate.
7262 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7264 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7265 that moved to I2, also allow destinations that are a paradoxical
7266 subreg (instead of a normal reg).
7268 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7271 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7274 2018-02-16 Richard Biener <rguenther@suse.de>
7276 PR tree-optimization/84037
7277 PR tree-optimization/84016
7279 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7280 Adjust vec_construct for the fact we need additional higher latency
7281 128bit inserts for AVX256 and AVX512 vector builds.
7282 (ix86_add_stmt_cost): Scale vector construction cost for
7285 2018-02-16 Richard Biener <rguenther@suse.de>
7287 PR tree-optimization/84417
7288 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7289 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7290 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7292 2018-02-16 Martin Liska <mliska@suse.cz>
7295 * internal-fn.def (ASAN_CHECK): Set proper flags.
7296 (ASAN_MARK): Likewise.
7298 2018-02-16 Julia Koval <julia.koval@intel.com>
7300 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7301 from PTA_CANNONLAKE.
7303 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7306 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7307 Use ++iter rather than iter++ for std::list iterators.
7308 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7309 defer deleting them until all nodes in the forest are processed. Do
7310 free even leaf nodes. Change to_process into auto_vec.
7313 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7314 * vec.h (vec_default_construct): Use memset instead of placement new
7315 if BROKEN_VALUE_INITIALIZATION is defined.
7316 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7317 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7320 PR rtl-optimization/83723
7321 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7322 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7323 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7325 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7327 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7329 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7331 PR rtl-optimization/81443
7332 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7333 from inner REGs to paradoxical SUBREGs.
7335 2018-02-16 Richard Biener <rguenther@suse.de>
7337 PR tree-optimization/84399
7338 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7339 For operands we can analyze at their definition make sure we can
7340 analyze them at each use as well.
7342 2018-02-16 Richard Biener <rguenther@suse.de>
7344 PR tree-optimization/84190
7345 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7346 volatile accesses if the decl isn't volatile.
7348 2018-02-15 Jason Merrill <jason@redhat.com>
7350 PR c++/84314 - ICE with templates and fastcall attribute.
7351 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7352 TYPE_CANONICAL on an existing type.
7354 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7356 PR tree-optimization/84383
7357 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7358 dstoff nor call operand_equal_p if dstbase is NULL.
7360 PR tree-optimization/84334
7361 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7362 also a CONSTANT_CLASS_P, punt.
7364 2018-02-14 Jim Wilson <jimw@sifive.com>
7366 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7367 first SMALL_OPERAND check. New local min_second_step. Move assert
7368 to where locals are set. Add TARGET_RVC support.
7369 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7371 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7373 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7375 2018-02-14 Martin Sebor <msebor@redhat.com>
7377 PR tree-optimization/83698
7378 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7379 arrays constrain the offset range to their bounds.
7380 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7381 (builtin_access::overlap): Avoid setting the size of overlap if it's
7383 (maybe_diag_overlap): Also consider arrays when deciding what values
7384 of offsets to include in diagnostics.
7386 2018-02-14 Martin Sebor <msebor@redhat.com>
7389 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7390 that correspond to the kind of a declaration.
7392 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7395 * config/pa/pa.md: Load address of PIC label using the linkage table
7396 if the label is nonlocal.
7398 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7400 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7401 warning message if user requests -maltivec=be.
7402 * doc/invoke.texi: Document deprecation of -maltivec=be.
7404 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7407 * config/rs6000/rs6000-c.c: Update definitions for
7408 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7409 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7411 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7414 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7415 add _get_ssp intrinsics. Remove argument from
7416 __builtin_ia32_rdssp[d|q].
7417 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7418 * config/i386/i386-builtin.def: Remove argument from
7419 __builtin_ia32_rdssp[d|q].
7420 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7421 ix86_expand_special_args_builtin for _rdssp[d|q].
7422 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7423 Clear register before usage.
7424 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7425 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7427 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7429 PR tree-optimization/84357
7430 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7431 operand 1 of an ARRAY_REF too.
7433 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7436 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7437 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7439 (set_of_reg): New struct.
7440 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7441 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7442 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7444 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7445 Split into bitclr, bitset, bitinvert patterns if appropriate.
7446 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7447 use rx_fuse_in_memory_bitop.
7448 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7449 to named insn, correct maximum insn length.
7451 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7454 * machmode.def: Define a complex mode for PARTIAL_INT.
7455 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7457 * doc/rtl.texi: Document CSPImode.
7458 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7460 (msp430_hard_regno_nregs_with_padding): Likewise.
7462 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7465 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7467 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7469 PR rtl-optimization/84169
7470 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7471 we generated a parallel as new i3 and we split that to new i2 and i3
7472 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7473 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7474 those to i2, not i1. Partially rewrite this scan code.
7476 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7479 * stor-layout.c (place_field): For variable length fields, adjust
7480 offset_align afterwards not just based on the field's alignment,
7481 but also on the size.
7484 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7485 of exps and logs in the use_exp2 case.
7487 2018-02-13 Jeff Law <law@redhat.com>
7489 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7492 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7495 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7499 * common.opt (gas-loc-support, gas-locview-support): New.
7500 (ginline-points, ginternal-reset-location-views): New.
7501 * doc/invoke.texi: Document them. Use @itemx where intended.
7502 (gvariable-location-views): Adjust.
7503 * target.def (reset_location_view): New.
7504 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7505 (TARGET_RESET_LOCATION_VIEW): New.
7506 * doc/tm.texi: Rebuilt.
7507 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7508 (dwarf2out_default_as_locview_support): New.
7509 (output_asm_line_debug_info): Use option variables.
7510 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7511 (output_loc_list): Likewise.
7512 (add_high_low_attributes): Check option variables.
7513 Don't output entry view attribute in strict mode.
7514 (gen_inlined_subroutine_die): Check option variables.
7515 (dwarf2out_inline_entry): Likewise.
7516 (init_sections_and_labels): Likewise.
7517 (dwarf2out_early_finish): Likewise.
7518 (maybe_reset_location_view): New, from...
7519 (dwarf2out_var_location): ... here. Call it.
7520 * debug.h (dwarf2out_default_as_loc_support): Declare.
7521 (dwarf2out_default_as_locview_support): Declare.
7522 * hooks.c (hook_int_rtx_insn_0): New.
7523 * hooks.h (hook_int_rtx_insn_0): Declare.
7524 * toplev.c (process_options): Take -gas-loc-support and
7525 -gas-locview-support from dwarf2out. Enable
7526 -gvariable-location-views by default only with locview
7527 assembler support. Enable -ginternal-reset-location-views by
7528 default only if the target defines the corresponding hook.
7529 Enable -ginline-points by default if location views are
7530 enabled; force it disabled if statement frontiers are
7532 * tree-inline.c (expand_call_inline): Check option variables.
7533 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7535 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7537 PR tree-optimization/84321
7538 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7539 handling. Also check whether the anti-range contains any values
7540 that satisfy the mask; switch to a VR_RANGE if not.
7542 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7545 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7547 2018-02-13 Martin Jambor <mjambor@suse.cz>
7550 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7551 of call statements, also set location of a load to a temporary.
7553 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7555 * config/rl78/rl78.c (add_vector_labels): New function.
7556 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7557 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7558 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7559 which checks that no arguments are passed.
7560 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7561 * doc/extend.texi: Documentation for the new attribute.
7563 2018-02-13 Andreas Schwab <schwab@suse.de>
7565 * config/riscv/linux.h (CPP_SPEC): Define.
7567 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7570 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7571 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7572 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7573 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7574 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7575 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7576 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7577 temporarily for AES and PCLMUL builtins.
7579 PR tree-optimization/84339
7580 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7581 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7585 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7586 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7587 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7589 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7591 * omp-simd-clone.h: New file.
7592 * omp-simd-clone.c: Include omp-simd-clone.h.
7593 (expand_simd_clones): No longer static.
7594 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7595 cgraph.h and omp-simd-clone.h.
7596 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7597 (vect_recog_widen_shift_pattern): Formatting fix.
7598 (vect_pattern_recog_1): Don't check optab for calls.
7601 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7602 operands[2] into a REG before using gen_lowpart on it.
7604 2018-02-12 Jeff Law <law@redhat.com>
7607 * config/sh/sh.c (find_barrier): Consider a sibling call
7610 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7611 successfully back substituting a reg.
7613 2018-02-12 Richard Biener <rguenther@suse.de>
7615 PR tree-optimization/84037
7616 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7617 parameter, move visited init to caller.
7618 (vect_slp_analyze_operations): Separate cost from validity
7619 check, initialize visited once for all instances.
7620 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7622 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7624 (vect_model_promotion_demotion_cost): Likewise.
7625 (vectorizable_bswap): Guard cost modeling with !slp_node
7626 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7628 (vectorizable_call): Likewise.
7629 (vectorizable_conversion): Likewise.
7630 (vectorizable_assignment): Likewise.
7631 (vectorizable_shift): Likewise.
7632 (vectorizable_operation): Likewise.
7633 (vectorizable_store): Likewise.
7634 (vectorizable_load): Likewise.
7635 (vectorizable_condition): Likewise.
7636 (vectorizable_comparison): Likewise.
7638 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7641 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7642 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7643 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7645 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7648 * params.def (inline-min-speedup): Increase from 8 to 15.
7649 (max-inline-insns-auto): Decrease from 40 to 30.
7650 * ipa-split.c (consider_split): Add some buffer for function to
7651 be considered inlining candidate.
7652 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7655 2018-02-12 Richard Biener <rguenther@suse.de>
7657 PR tree-optimization/84037
7658 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7659 matched stmts if we cannot swap the non-matched ones.
7661 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7663 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7664 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7665 _mm_maskz_scalef_round_ss): New intrinsics.
7666 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7667 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7668 __builtin_ia32_scalefss_round): Remove.
7669 (__builtin_ia32_scalefsd_mask_round,
7670 __builtin_ia32_scalefss_mask_round): New intrinsics.
7671 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7672 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7673 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7674 "<round_constraint>")): Changed to ...
7675 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7676 "<round_scalar_constraint>")): ... this.
7677 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7678 %0, %1, %2<round_op3>}"): Changed to ...
7679 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7680 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7681 %2<round_scalar_mask_op3>}"): ... this.
7682 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7684 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7686 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7687 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7688 (_mm_maskz_sqrt_round_ss): New intrinsics.
7689 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7690 (__builtin_ia32_sqrtsd_mask_round)
7691 (__builtin_ia32_sqrtss_mask_round): New builtins.
7692 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7693 (__builtin_ia32_sqrtss_round): Remove.
7694 (__builtin_ia32_sqrtsd_mask_round)
7695 (__builtin_ia32_sqrtss_mask_round): New builtins.
7696 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7697 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7698 ((match_operand:VF_128 1 "vector_operand"
7699 "xBm,<round_constraint>")): Changed to ...
7700 ((match_operand:VF_128 1 "vector_operand"
7701 "xBm,<round_scalar_constraint>")): ... this.
7702 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7703 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7704 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7705 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7706 %<iptr>1<round_scalar_mask_op3>}): ... this.
7707 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7708 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7710 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7713 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7714 Cast vec_cmpeq result to correct type.
7715 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7716 Cast vec_cmpgt result to correct type.
7718 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7720 * final.c (final_scan_insn_1): Renamed from...
7721 (final_scan_insn): ... this. New wrapper, to recover
7722 seen from the outermost call in recursive ones.
7723 * config/sparc/sparc.c (output_return): Drop seen from call.
7724 (output_sibcall): Likewise.
7725 * config/visium/visium.c (output_branch): Likewise.
7727 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7729 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7732 2018-02-10 Alan Modra <amodra@gmail.com>
7735 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7736 Specify LR as an input.
7738 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7741 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7742 remove_member_access_dummy_vars): New functions.
7743 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7744 lower_omp_1, execute_lower_omp): Use them.
7746 PR rtl-optimization/84308
7747 * shrink-wrap.c (spread_components): Release todo vector.
7749 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7751 PR rtl-optimization/57193
7752 * ira-color.c (struct allocno_color_data): Add member
7753 conflict_allocno_hard_prefs.
7754 (update_conflict_allocno_hard_prefs): New.
7755 (bucket_allocno_compare_func): Add a preference based on
7756 conflict_allocno_hard_prefs.
7757 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7758 (color_allocnos): Remove a dead code. Initiate
7759 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7761 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7764 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7765 constraint from =wa to wa. Avoid a subreg on the output operand,
7766 instead use a pseudo and subreg it in a move.
7767 (p9_xxbrd_<mode>): Changed to ...
7768 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7769 (p9_xxbrd_v2df): New expander.
7770 (p9_xxbrw_<mode>): Changed to ...
7771 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7772 (p9_xxbrw_v4sf): New expander.
7774 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7776 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7778 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7781 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7782 multiply in 32-bit mode.
7783 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7784 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7787 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7789 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7790 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7791 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7792 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7794 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7797 * dwarf2out.c (is_trivial_indirect_ref): New function.
7798 (dwarf2out_late_global_decl): Do not generate a location
7799 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7800 and that are not defined in the current unit.
7802 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7804 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7805 instead of a libcall for UNORDERED.
7807 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7810 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7811 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7813 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7816 * config/s390/s390.c (s390_set_current_function): Invoke
7817 s390_indirect_branch_settings also if fndecl didn't change.
7819 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7821 * config/rs6000/rs6000.md (blockage): Set length to zero.
7823 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7825 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7827 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7830 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7831 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7835 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7836 PARALLEL incoming that failed vt_get_decl_and_offset check.
7839 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7840 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7842 (get_variable_section): For decls in named .bss* sections pass true as
7843 second argument to bss_initializer_p.
7845 2018-02-09 Marek Polacek <polacek@redhat.com>
7846 Jakub Jelinek <jakub@redhat.com>
7849 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7850 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7851 Sync some changes from cxx_fold_indirect_ref.
7853 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7855 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7857 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7858 (BLOCK_INLINE_ENTRY_LABEL): New.
7859 (dwarf2out_var_location): Disregard inline entry markers.
7860 (inline_entry_data): New struct.
7861 (inline_entry_data_hasher): New hashtable type.
7862 (inline_entry_data_hasher::hash): New.
7863 (inline_entry_data_hasher::equal): New.
7864 (inline_entry_data_table): New variable.
7865 (add_high_low_attributes): Add DW_AT_entry_pc and
7866 DW_AT_GNU_entry_view attributes if a pending entry is found
7867 in inline_entry_data_table. Add old entry_pc attribute only
7868 if debug nonbinding markers are disabled.
7869 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7870 markers are enabled.
7871 (block_within_block_p, dwarf2out_inline_entry): New.
7872 (dwarf2out_finish): Check that no entries remained in
7873 inline_entry_data_table.
7874 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7875 (final_scan_insn, notice_source_line): Likewise.
7876 (rest_of_clean_state): Skip inline entry markers.
7877 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7879 * gimple.c (gimple_build_debug_inline_entry): New.
7880 * gimple.h (enum gimple_debug_subcode): Add
7881 GIMPLE_DEBUG_INLINE_ENTRY.
7882 (gimple_build_debug_inline_entry): Declare.
7883 (gimple_debug_inline_entry_p): New.
7884 (gimple_debug_nonbind_marker_p): Adjust.
7885 * insn-notes.def (INLINE_ENTRY): New.
7886 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7887 inline entry marker notes.
7888 (print_insn): Likewise.
7889 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7890 (INSN_DEBUG_MARKER_KIND): Likewise.
7891 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7892 * tree-inline.c (expand_call_inline): Build and insert
7893 debug_inline_entry stmt.
7894 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7895 inline entry blocks early, if nonbind markers are enabled.
7896 (dump_scope_block): Dump fragment info.
7897 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7898 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7899 (gimple_build_debug_inline_entry): New.
7900 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7901 Enable/disable inline entry points too.
7902 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7903 (DEBUG_INSN): Describe inline entry markers.
7905 * common.opt (gvariable-location-views): New.
7906 (gvariable-location-views=incompat5): New.
7907 * config.in: Rebuilt.
7908 * configure: Rebuilt.
7909 * configure.ac: Test assembler for view support.
7910 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7911 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7912 * dwarf2out.c (var_loc_view): New typedef.
7913 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7914 (dwarf2out_locviews_in_attribute): New.
7915 (dwarf2out_locviews_in_loclist): New.
7916 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7917 (enum dw_line_info_opcode): Add LI_adv_address.
7918 (struct dw_line_info_table): Add view.
7919 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7920 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7921 (zero_view_p): New variable.
7922 (ZERO_VIEW_P): New macro.
7923 (output_asm_line_debug_info): New.
7924 (struct var_loc_node): Add view.
7925 (add_AT_view_list, AT_loc_list): New.
7926 (add_var_loc_to_decl): Add view param. Test it against last.
7927 (new_loc_list): Add view params. Record them.
7928 (AT_loc_list_ptr): Handle loc and view lists.
7929 (view_list_to_loc_list_val_node): New.
7930 (print_dw_val): Handle dw_val_class_view_list.
7931 (size_of_die): Likewise.
7932 (value_format): Likewise.
7933 (loc_list_has_views): New.
7934 (gen_llsym): Set vl_symbol too.
7935 (maybe_gen_llsym, skip_loc_list_entry): New.
7936 (dwarf2out_maybe_output_loclist_view_pair): New.
7937 (output_loc_list): Output view list or entries too.
7938 (output_view_list_offset): New.
7939 (output_die): Handle dw_val_class_view_list.
7940 (output_dwarf_version): New.
7941 (output_compilation_unit_header): Use it.
7942 (output_skeleton_debug_sections): Likewise.
7943 (output_rnglists, output_line_info): Likewise.
7944 (output_pubnames, output_aranges): Update version comments.
7945 (output_one_line_info_table): Output view numbers in asm comments.
7946 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7947 Call maybe_gen_llsym.
7948 (loc_list_from_tree_1): Adjust.
7949 (add_AT_location_description): Create view list attribute if
7950 needed, check it's absent otherwise.
7951 (convert_cfa_to_fb_loc_list): Adjust.
7952 (maybe_emit_file): Call output_asm_line_debug_info for test.
7953 (dwarf2out_var_location): Reset views as needed. Precompute
7954 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7955 attribute. Set view.
7956 (new_line_info_table): Reset next view.
7957 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7958 (dwarf2out_source_line): Likewise. Output view resets and labels to
7959 the assembler, or select appropriate line info opcodes.
7960 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7961 (optimize_string_length): Catch it. Adjust.
7962 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7963 dw_val_class_view_list, and remove it if no longer needed.
7964 (hash_loc_list): Hash view numbers.
7965 (loc_list_hasher::equal): Compare them.
7966 (optimize_location_lists): Check whether a view list symbol is
7967 needed, and whether the locview attribute is present, and
7968 whether they match. Remove the locview attribute if no longer
7970 (index_location_lists): Call skip_loc_list_entry for test.
7971 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7972 Use output_dwarf_version.
7973 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7974 (struct dw_val_node): Add val_view_list.
7975 * final.c (SEEN_NEXT_VIEW): New.
7976 (set_next_view_needed): New.
7977 (clear_next_view_needed): New.
7978 (maybe_output_next_view): New.
7979 (final_start_function): Rename to...
7980 (final_start_function_1): ... this. Take pointer to FIRST,
7981 add SEEN parameter. Emit param bindings in the initial view.
7982 (final_start_function): Reintroduce SEEN-less interface.
7983 (final): Rename to...
7984 (final_1): ... this. Take SEEN parameter. Output final pending
7985 next view at the end.
7986 (final): Reintroduce seen-less interface.
7987 (final_scan_insn): Output pending next view before switching
7988 sections or ending a block. Mark the next view as needed when
7989 outputting variable locations. Notify debug backend of section
7990 changes, and of location view changes.
7991 (rest_of_handle_final): Adjust.
7992 * toplev.c (process_options): Autodetect value for debug variable
7993 location views option. Warn on incompat5 without -gdwarf-5.
7994 * doc/invoke.texi (gvariable-location-views): New.
7995 (gvariable-location-views=incompat5): New.
7996 (gno-variable-location-views): New.
7998 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8000 PR tree-optimization/84136
8001 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8002 that the result of find_edge is non-NULL.
8004 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8007 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8008 storing integer register in SImode. Fix cost of 256 and 512
8009 byte aligned SSE register store.
8011 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8013 * config/i386/i386.c (ix86_multiplication_cost): Fix
8014 multiplication cost for TARGET_AVX512DQ.
8016 2018-02-08 Marek Polacek <polacek@redhat.com>
8018 PR tree-optimization/84238
8019 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8022 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8024 PR tree-optimization/84265
8025 * tree-vect-stmts.c (vectorizable_store): Don't treat
8026 VMAT_CONTIGUOUS accesses as grouped.
8027 (vectorizable_load): Likewise.
8029 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8031 PR tree-optimization/81635
8032 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8033 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8034 (test_round_for_mask): New functions.
8035 (wide_int_cc_tests): Call test_round_for_mask.
8036 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8037 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8038 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8039 range returned by get_range_info.
8041 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8044 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8045 * symtab.c: Include builtins.h
8046 (symtab_node::output_to_lto_symbol_table_p): Move here
8047 from lto-streamer-out.c:output_symbol_p.
8048 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8049 (output_symbol_p): Move all logic to symtab.c
8050 (produce_symtab): Update.
8052 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8054 * config/s390/s390-opts.h (enum indirect_branch): Define.
8055 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8056 (s390_indirect_branch_via_thunk)
8057 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8058 (enum s390_indirect_branch_type): Define.
8059 * config/s390/s390.c (struct s390_frame_layout, struct
8060 machine_function): Remove.
8061 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8062 (indirect_branch_table_label_no, indirect_branch_table_name):
8064 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8065 (enum s390_indirect_branch_option): Define.
8066 (s390_return_addr_from_memory): New function.
8067 (s390_handle_string_attribute): New function.
8068 (s390_attribute_table): Add new attribute handler.
8069 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8070 (s390_indirect_branch_via_thunk): New function.
8071 (s390_indirect_branch_via_inline_thunk): New function.
8072 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8073 sibling call optimization for non z10 compiles.
8074 (s390_emit_call): Force indirect branch target to be a single
8075 register. Add r1 clobber for non-z10 compiles.
8076 (s390_emit_epilogue): Emit return jump via return_use expander.
8077 (s390_reorg): Handle JUMP_INSNs as execute targets.
8078 (s390_option_override_internal): Perform validity checks for the
8079 new command line options.
8080 (s390_indirect_branch_attrvalue): New function.
8081 (s390_indirect_branch_settings): New function.
8082 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8083 (s390_output_indirect_thunk_function): New function.
8084 (s390_code_end): Implement target hook.
8085 (s390_case_values_threshold): Implement target hook.
8086 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8088 * config/s390/s390.h (struct s390_frame_layout)
8089 (struct machine_function): Move here from s390.c.
8090 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8091 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8092 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8093 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8094 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8095 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8096 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8097 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8098 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8099 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8100 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8101 (mnemonic attribute): Add values which aren't recognized
8103 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8104 pattern for branch conversion. Fix mnemonic attribute.
8105 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8106 indirect branch via thunk if requested.
8107 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8108 ("*indirect_jump"): Disable for branch conversion using out of
8110 ("indirect_jump_via_thunk<mode>_z10")
8111 ("indirect_jump_via_thunk<mode>")
8112 ("indirect_jump_via_inlinethunk<mode>_z10")
8113 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8114 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8115 ("casesi_jump_via_inlinethunk<mode>_z10")
8116 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8117 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8118 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8119 ("*indirect2_jump"): Disable for branch conversion.
8120 ("casesi_jump"): Turn into expander and expand patterns for branch
8122 ("return_use"): New expander.
8123 ("*return"): Emit return via thunk and rename it to ...
8124 ("*return<mode>"): ... this one.
8125 * config/s390/s390.opt: Add new options and and enum for the
8128 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8130 * lra-constraints.c (match_reload): Unconditionally use
8131 gen_lowpart_SUBREG, rather than selecting between that
8132 and equivalent gen_rtx_SUBREG code.
8134 2018-02-08 Richard Biener <rguenther@suse.de>
8136 PR tree-optimization/84233
8137 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8138 changed flag instead of boguously re-using phi_inserted.
8140 2018-02-08 Martin Jambor <mjambor@suse.cz>
8142 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8143 static local variables.
8145 2018-02-08 Richard Biener <rguenther@suse.de>
8147 PR tree-optimization/84278
8148 * tree-vect-stmts.c (vectorizable_store): When looking for
8149 smaller vector types to perform grouped strided loads/stores
8150 make sure the mode is supported by the target.
8151 (vectorizable_load): Likewise.
8153 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8155 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8156 Increase LDP/STP opportunities by adding adjacent callee-saves.
8158 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8160 PR rtl-optimization/84068
8161 PR rtl-optimization/83459
8162 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8164 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8166 PR tree-optimization/84224
8167 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8168 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8171 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8174 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8175 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8176 count, remove one USE.
8178 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8180 * doc/install.texi (Configuration): Document the
8181 --with-long-double-format={ibm,ieee} PowerPC configuration
8185 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8186 Convert from define_expand to be define_insn_and_split. Rework
8187 float/double/_Float128 conversions to QI/HI/SImode to work with
8188 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8189 conversions to QI/HImode types did a store and then a load to
8190 truncate the value. For conversions to VSX registers, don't split
8191 the insn, instead emit the code directly. Use the code iterator
8192 any_fix to combine signed and unsigned conversions.
8193 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8194 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8195 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8196 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8197 (fix_<mode>di2_hw): Likewise.
8198 (fixuns_<mode>di2_hw): Likewise.
8199 (fix_<mode>si2_hw): Likewise.
8200 (fixuns_<mode>si2_hw): Likewise.
8201 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8202 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8203 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8204 fix<uns>_trunc<SFDF:mode>si2_p8.
8205 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8207 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8208 (fix<uns>_<mode>_mem): Likewise.
8209 (fctiw<u>z_<mode>_mem): Likewise.
8210 (fix<uns>_<mode>_mem): Likewise.
8211 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8212 the register allocator from doing a direct move to the GPRs to do
8213 a store, and instead use the ISA 3.0 store byte/half-word from
8214 vector register instruction. For IEEE 128-bit floating point,
8215 also optimize stores of 32-bit ints.
8216 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8218 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8220 * genextract.c (push_pathstr_operand): New function to support
8222 (walk_rtx): Call push_pathstr_operand.
8223 (print_path): Support [a-zA-Z].
8225 2018-02-07 Richard Biener <rguenther@suse.de>
8227 PR tree-optimization/84037
8228 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8229 (cse_and_gimplify_to_preheader): Declare.
8230 (vect_get_place_in_interleaving_chain): Likewise.
8231 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8233 (_loop_vec_info::~_loop_vec_info): Delete it.
8234 (cse_and_gimplify_to_preheader): New function.
8235 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8236 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8237 (vectorizable_load): Likewise. For grouped stores always base
8238 the IV on the first element.
8239 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8240 condition before gimplifying.
8242 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8244 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8245 *DIV_EXPR and *MOD_EXPR.
8247 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8250 * config/i386/i386.c (ix86_option_override_internal): Mask out
8251 the CF_SET bit when checking -fcf-protection.
8253 2018-02-07 Tom de Vries <tom@codesourcery.com>
8256 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8259 2018-02-07 Richard Biener <rguenther@suse.de>
8261 PR tree-optimization/84204
8262 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8265 PR tree-optimization/84205
8266 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8267 special-case isl_ast_op_zdiv_r.
8269 PR tree-optimization/84223
8270 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8271 Only add conditions from within the region.
8272 (gather_bbs::after_dom_children): Adjust.
8274 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8277 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8278 * config/avr/avr.md: Only post-reload split REG-REG moves if
8279 either register is GENERAL_REG_P.
8281 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8283 PR tree-optimization/84235
8284 * tree-ssa-scopedtables.c
8285 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8286 if the subtraction is performed in floating point type where NaNs are
8287 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8288 build 1. Formatting fix.
8290 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8293 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8294 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8295 and skip it regardless of bb boundaries. Use CALL_P macro,
8296 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8297 unnecessarily, formatting fix.
8299 2018-02-06 Michael Collison <michael.collison@arm.com>
8301 * config/arm/thumb2.md:
8302 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8303 (*thumb_mov_notscc): Ditto.
8305 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8308 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8309 unsigned_fix, not "s".
8311 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8313 * configure.ac (gcc_fn_eh_frame_ro): New function.
8314 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8315 correct .eh_frame permissions.
8316 * configure: Regenerate.
8318 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8320 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8323 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8325 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8326 Display warning message for -mno-speculate-indirect-jumps.
8328 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8330 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8332 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8334 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8336 PR tree-optimization/84225
8337 * tree-eh.c (find_trapping_overflow): Only call
8338 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8340 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8343 * config/i386/i386.c: Reimplement the check of possible options
8344 -mibt/-mshstk conbination. Change error messages.
8345 * doc/invoke.texi: Fix a typo: remove extra '='.
8347 2018-02-06 Marek Polacek <polacek@redhat.com>
8349 PR tree-optimization/84228
8350 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8352 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8355 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8356 emitted arch directives.
8357 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8358 __ARM_FEATURE_COPROC before changing architectures.
8360 2018-02-06 Richard Biener <rguenther@suse.de>
8362 * config/i386/i386.c (print_reg): Fix typo.
8363 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8365 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8367 * configure: Regenerate.
8369 2018-02-05 Martin Sebor <msebor@redhat.com>
8371 PR tree-optimization/83369
8372 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8375 2018-02-05 Martin Liska <mliska@suse.cz>
8377 * doc/invoke.texi: Cherry-pick upstream r323995.
8379 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8381 * ira.c (ira_init_register_move_cost): Adjust comment.
8383 2018-02-05 Martin Liska <mliska@suse.cz>
8385 PR gcov-profile/84137
8386 * doc/gcov.texi: Fix typo in documentation.
8388 2018-02-05 Martin Liska <mliska@suse.cz>
8390 PR gcov-profile/83879
8391 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8392 using dlopen functionality.
8394 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8396 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8397 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8398 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8399 _mm_maskz_range_round_ss): New intrinsics.
8400 (__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/i386-builtin.def (__builtin_ia32_rangesd128_round,
8405 __builtin_ia32_rangess128_round): Remove.
8406 (__builtin_ia32_rangesd128_mask_round,
8407 __builtin_ia32_rangess128_mask_round): New builtins.
8408 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8409 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8410 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8411 "<round_saeonly_constraint>")): Changed to ...
8412 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8413 "<round_saeonly_scalar_constraint>")): ... this.
8414 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8415 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8416 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8417 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8418 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8420 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8422 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8424 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8425 Remove all values except native, 8540 and 8548.
8427 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8429 * config/i386/i386.c (ix86_output_function_return): Pass
8430 INVALID_REGNUM, instead of -1, as invalid register number to
8431 indirect_thunk_name and output_indirect_thunk.
8433 2018-02-02 Julia Koval <julia.koval@intel.com>
8435 * config.gcc: Add -march=icelake.
8436 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8437 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8438 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8439 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8440 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8441 (processor_target_table): Add icelake.
8442 (ix86_option_override_internal): Handle new PTAs.
8443 (get_builtin_code_for_version): Handle icelake.
8444 (M_INTEL_COREI7_ICELAKE): New.
8445 (fold_builtin_cpu): Handle icelake.
8446 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8447 * doc/invoke.texi: Add -march=icelake.
8449 2018-02-02 Julia Koval <julia.koval@intel.com>
8451 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8452 to wide_int_bitmask.
8453 * wide-int-bitmask.h: New.
8455 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8458 * config/i386/i386.md: Replace Pmode with word_mode in
8459 builtin_setjmp_setup and builtin_longjmp to support x32.
8461 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8465 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8467 (rs6000_supported_cpu_names): New static variable.
8468 (linux_cpu_translation_table): Likewise.
8469 (elf_platform) <cpu>: Define new static variable and use it.
8470 Translate kernel AT_PLATFORM name to canonical name if needed.
8471 Error if platform name is unknown.
8473 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8476 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8478 2018-02-01 Jeff Law <law@redhat.com>
8481 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8482 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8483 the scratch if RELEASE_VIA_POP is false.
8484 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8485 If we have to save a temporary register, decrement SIZE appropriately.
8486 Pass new arguments to release_scratch_register_on_entry.
8487 (ix86_adjust_stack_and_probe): Likewise.
8488 (ix86_emit_probe_stack_range): Pass new arguments to
8489 release_scratch_register_on_entry.
8491 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8493 PR rtl-optimization/84157
8494 * combine.c (change_zero_ext): Use REG_P predicate in
8495 front of HARD_REGISTER_P predicate.
8497 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8499 * config/avr/avr.c (avr_option_override): Move disabling of
8500 -fdelete-null-pointer-checks to...
8501 * common/config/avr/avr-common.c (avr_option_optimization_table):
8504 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8506 PR tree-optimization/81635
8507 * tree-data-ref.c (split_constant_offset_1): For types that
8508 wrap on overflow, try to use range info to prove that wrapping
8511 2018-02-01 Renlin Li <renlin.li@arm.com>
8514 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8516 (aarch64_register_move_cost): Likewise.
8517 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8519 (REG_CLASS_NAMES): Likewise.
8520 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8521 TAILCALL_ADDR_REGS. Remove IP registers.
8522 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8524 2018-02-01 Richard Biener <rguenther@suse.de>
8526 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8527 for specifying RPO order and allow NULL for that.
8528 * domwalk.c (dom_walker::dom_walker): Likewise.
8529 (dom_walker::walk): Handle NULL RPO order.
8530 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8532 (rewrite_update_dom_walker): Likewise.
8533 (mark_def_dom_walker): Likewise.
8535 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8537 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8538 (aarch64_maybe_expand_sve_subreg_move): Declare.
8539 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8540 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8542 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8543 that are semantically a reverse operation.
8544 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8545 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8546 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8548 (aarch64_can_change_mode_class): For big-endian, forbid changes
8549 between two SVE modes if they have different element sizes.
8551 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8553 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8554 the TImode handling for big-endian targets.
8556 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8558 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8559 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8561 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8562 Remove BSWAP handing for big-endian targets and use the form of
8563 LD1RQ appropariate for the mode.
8565 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8567 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8568 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8571 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8574 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8575 check for operands that need to go through aarch64_sve_reload_be.
8577 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8579 PR tree-optimization/81661
8580 PR tree-optimization/84117
8581 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8582 * tree-eh.c: Include gimplify.h.
8583 (find_trapping_overflow, replace_trapping_overflow,
8584 rewrite_to_non_trapping_overflow): New functions.
8585 * tree-vect-loop.c: Include tree-eh.h.
8586 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8587 * tree-data-ref.c: Include tree-eh.h.
8588 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8590 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8592 PR rtl-optimization/84123
8593 * combine.c (change_zero_ext): Check if hard register satisfies
8594 can_change_dest_mode before calling gen_lowpart_SUBREG.
8596 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8599 * ira.c (ira_init_register_move_cost): Remove assert.
8601 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8603 PR rtl-optimization/84071
8604 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8605 * doc/tm.texi: Regenerate.
8607 2018-01-31 Richard Biener <rguenther@suse.de>
8609 PR tree-optimization/84132
8610 * tree-data-ref.c (analyze_miv_subscript): Properly
8611 check whether evolution_function_is_affine_multivariate_p
8612 before calling gcd_of_steps_may_divide_p.
8614 2018-01-31 Julia Koval <julia.koval@intel.com>
8617 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8618 * config/i386/i386.md (rdpid_rex64) New.
8619 (rdpid): Make 32bit only.
8621 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8624 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8625 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8627 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8630 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8632 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8634 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8636 PR rtl-optimization/84071
8637 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8638 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8640 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8642 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8643 (arc_attribute_table): Add 'aux' attribute.
8644 (arc_in_small_data_p): Consider aux like variables.
8645 (arc_is_aux_reg_p): New function.
8646 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8647 (arc_get_aux_arg): New function.
8648 (prepare_move_operands): Handle aux-register access.
8649 (arc_handle_aux_attribute): New function.
8650 * doc/extend.texi (ARC Variable attributes): Add subsection.
8652 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8654 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8655 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8656 (arc_attribute_table): Add 'uncached' attribute.
8657 (arc_print_operand): Print '.di' flag for uncached memory
8659 (arc_in_small_data_p): Do not consider for small data the uncached
8661 (arc_is_uncached_mem_p): New function.
8662 * config/arc/predicates.md (compact_store_memory_operand): Check
8663 for uncached memory accesses.
8664 (nonvol_nonimm_operand): Likewise.
8665 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8667 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8670 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8671 falign-loops=): Add Optimization flag.
8673 2018-01-30 Jeff Law <law@redhat.com>
8676 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8677 INT_REGISTERS_SAVED. Check it prior to calling
8678 get_scratch_register_on_entry.
8679 (ix86_adjust_stack_and_probe): Similarly.
8680 (ix86_emit_probe_stack_range): Similarly.
8681 (ix86_expand_prologue): Corresponding changes.
8683 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8686 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8687 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8689 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8692 * lra-constraints.c (curr_insn_transform): Process AND in the
8695 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8697 PR rtl-optimization/83986
8698 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8699 dependence against last_pending_memory_flush in addition to
8702 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8704 PR tree-optimization/81611
8705 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8708 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8711 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8714 2018-01-30 Richard Biener <rguenther@suse.de>
8715 Jakub Jelinek <jakub@redhat.com>
8717 PR tree-optimization/84111
8718 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8719 inner loops added during recursion, as they don't have up-to-date
8722 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8725 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8726 (can_inline_edge_by_limits_p): ... here.
8727 (can_early_inline_edge_p, check_callers,
8728 update_caller_keys, update_callee_keys, recursive_inlining,
8729 add_new_edges_to_heap, speculation_useful_p,
8730 inline_small_functions,
8731 inline_small_functions, flatten_function,
8732 inline_to_all_callers_1): Update.
8734 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8736 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8739 2018-01-30 Richard Biener <rguenther@suse.de>
8741 PR tree-optimization/83008
8742 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8743 invariant and constant vector uses in stmts when they need
8746 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8749 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8750 * configure: Regenerate.
8752 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8754 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8756 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8757 Use gen_rtx_REG rather than gen_lowpart.
8759 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8761 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8762 rather than 0 when creating partial subregs.
8764 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8766 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8769 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8772 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8773 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8774 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8775 flags. This restores the settings used before the 2017-07-24.
8776 Turning off pre increment/decrement/modify allows IVOPTS to
8777 optimize DF/SF loops where the index is an int.
8779 2018-01-29 Richard Biener <rguenther@suse.de>
8780 Kelvin Nilsen <kelvin@gcc.gnu.org>
8783 * tree-vect-stmts.c (vectorizable_call): Don't call
8784 targetm.vectorize_builtin_md_vectorized_function if callee is
8787 2018-01-22 Carl Love <cel@us.ibm.com>
8789 * doc/extend.tex: Fix typo in second arg in
8790 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8792 2018-01-29 Richard Biener <rguenther@suse.de>
8794 PR tree-optimization/84086
8795 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8796 (flush_ssaname_freelist): When SSA names were released reset
8797 the SCEV hash table.
8799 2018-01-29 Richard Biener <rguenther@suse.de>
8801 PR tree-optimization/84057
8802 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8803 removed paths when removing edges.
8805 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8807 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8808 -mfunction-return=@var{choice}.
8810 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8813 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8815 (layout::print_source_line): Likewise.
8816 (test_get_line_width_without_trailing_whitespace): Add test cases.
8818 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8821 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8824 2018-01-26 Jim Wilson <jimw@sifive.com>
8826 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8828 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8831 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8833 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8834 and CMP + SUB-immediate -> SUBS.
8836 2018-01-26 Martin Sebor <msebor@redhat.com>
8838 PR tree-optimization/83896
8839 * tree-ssa-strlen.c (get_string_len): Rename...
8840 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8841 Avoid assuming length is constant.
8842 (handle_char_store): Use HOST_WIDE_INT for string length.
8844 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8847 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8848 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8850 2018-01-26 Richard Biener <rguenther@suse.de>
8852 PR rtl-optimization/84003
8853 * dse.c (record_store): Only record redundant stores when
8854 the earlier store aliases at least all accesses the later one does.
8856 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8858 PR rtl-optimization/83985
8859 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8860 REG_CFA_RESTORE insns.
8861 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8862 REG_CFA_RESTORE insns here.
8865 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8866 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8868 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8870 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8871 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8872 (arc_init): Likewise.
8873 (arc_override_options): Likewise.
8874 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8876 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8878 * config/arc/arc.h (TARGET_DBNZ): Define.
8879 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8880 properly set the tune attribute.
8881 (dbnz): Use TARGET_DBNZ guard.
8882 * config/arc/arc.opt (mtune): Add core3 option.
8884 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8886 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8887 recognize new pic like addresses.
8888 (arc_delegitimize_address): Clean up.
8890 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8892 * config/arc/arc-arches.def: Option mrf16 valid for all
8894 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8895 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8896 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8897 * config/arc/arc-tables.opt: Regenerate.
8898 * config/arc/arc.c (arc_conditional_register_usage): Handle
8899 reduced register file case.
8900 (arc_file_start): Set must have build attributes.
8901 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8903 * config/arc/arc.opt (mrf16): Add new option.
8904 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8905 * config/arc/genmultilib.awk: Handle new mrf16 option.
8906 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8907 * config/arc/t-multilib: Regenerate.
8908 * doc/invoke.texi (ARC Options): Document mrf16 option.
8910 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8912 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8913 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8914 (arc_attribute_table): Add 'secure_call' attribute.
8915 (arc_print_operand): Print secure call operand.
8916 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8918 (arc_is_secure_call_p): New function. * config/arc/arc.md
8919 (call_i): Add support for sjli instruction.
8920 (call_value_i): Likewise.
8921 * config/arc/constraints.md (Csc): New constraint.
8923 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8924 John Eric Martin <John.Martin@emmicro-us.com>
8926 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8927 * config/arc/arc.c (_arc_jli_section): New struct.
8928 (arc_jli_section): New type.
8929 (rc_jli_sections): New static variable.
8930 (arc_handle_jli_attribute): New function.
8931 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8932 (arc_file_end): New function.
8933 (TARGET_ASM_FILE_END): Define.
8934 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8935 (arc_add_jli_section): New function.
8936 (jli_call_scan): Likewise.
8937 (arc_reorg): Call jli_call_scan.
8938 (arc_output_addsi): Remove 'S' from printing asm operand.
8939 (arc_is_jli_call_p): New function.
8940 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8942 (movhi_insn): Likewise.
8943 (movsi_insn): Likewise.
8944 (movsi_set_cc_insn): Likewise.
8945 (loadqi_update): Likewise.
8946 (load_zeroextendqisi_update): Likewise.
8947 (load_signextendqisi_update): Likewise.
8948 (loadhi_update): Likewise.
8949 (load_zeroextendhisi_update): Likewise.
8950 (load_signextendhisi_update): Likewise.
8951 (loadsi_update): Likewise.
8952 (loadsf_update): Likewise.
8953 (movsicc_insn): Likewise.
8954 (bset_insn): Likewise.
8955 (bxor_insn): Likewise.
8956 (bclr_insn): Likewise.
8957 (bmsk_insn): Likewise.
8958 (bicsi3_insn): Likewise.
8959 (cmpsi_cc_c_insn): Likewise.
8960 (movsi_ne): Likewise.
8961 (movsi_cond_exec): Likewise.
8962 (clrsbsi2): Likewise.
8973 (call_i): Remove 'S' asm letter, add jli instruction.
8974 (call_value_i): Likewise.
8975 * config/arc/arc.op (mjli-always): New option.
8976 * config/arc/constraints.md (Cji): New constraint.
8977 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8979 (subsf3_fpx): Likewise.
8980 (mulsf3_fpx): Likewise.
8981 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8983 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8985 * doc/invoke.texi (ARC): Document mjli-always option.
8987 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8989 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8990 avoid addition with 0 and use incw and decw where possible.
8992 2018-01-26 Richard Biener <rguenther@suse.de>
8994 PR tree-optimization/81082
8995 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8996 association if it requires casting to unsigned.
8997 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8998 from fold_plusminus_mult_expr to catch important cases late when
8999 range info is available.
9001 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9003 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9004 * configure.ac (hidden_linkonce): New test.
9005 * configure: Regenerate.
9006 * config.in: Regenerate.
9008 2018-01-26 Julia Koval <julia.koval@intel.com>
9010 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9011 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9012 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9013 _mm_mask_bitshuffle_epi64_mask): Fix type.
9014 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9015 USI_FTYPE_V4DI_V4DI_USI): Remove.
9016 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9017 __builtin_ia32_vpshufbitqmb256_mask,
9018 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9019 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9020 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9022 2018-01-26 Alan Modra <amodra@gmail.com>
9025 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9026 UNSPEC_VBPERMQ. Sort other unspecs.
9028 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9030 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9032 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9035 * predict.c (drop_profile): Do not push/pop cfun; update also
9037 (handle_missing_profiles): Fix logic looking for zero profiles.
9039 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9042 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9043 on functions with #pragma omp declare simd or functions with simd
9045 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9046 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9047 Remove trailing \n from warning_at calls.
9049 2018-01-25 Tom de Vries <tom@codesourcery.com>
9052 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9053 for neutered workers.
9055 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9058 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9059 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9061 2018-01-24 Jeff Law <law@redhat.com>
9064 * i386.c (get_probe_interval): Move to earlier point.
9065 (ix86_compute_frame_layout): If -fstack-clash-protection and
9066 the frame is larger than the probe interval, then use pushes
9067 to save registers rather than reg->mem moves.
9068 (ix86_expand_prologue): Remove conditional for int_registers_saved
9071 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9074 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9075 min/max for never referenced object.
9077 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9080 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9082 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9083 attributes from DECL_ATTRIBUTES (decl) without affecting
9084 DECL_ATTRIBUTES (current_function_decl).
9085 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9086 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9088 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9090 PR tree-optimization/83979
9091 * fold-const.c (fold_comparison): Use constant_boolean_node
9092 instead of boolean_{true,false}_node.
9094 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9096 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9099 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9101 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9102 Simplify the clause that sets the length attribute.
9103 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9104 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9105 clause that sets the length attribute.
9106 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9108 2018-01-24 Tom de Vries <tom@codesourcery.com>
9111 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9112 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9113 Add strict parameter.
9114 (prevent_branch_around_nothing): Insert dummy insn between branch to
9115 label and label with no ptx insn inbetween.
9116 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9118 2018-01-24 Tom de Vries <tom@codesourcery.com>
9121 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9122 for neutered threads in warp.
9123 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9125 2018-01-24 Richard Biener <rguenther@suse.de>
9127 PR tree-optimization/83176
9128 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9131 2018-01-24 Richard Biener <rguenther@suse.de>
9133 PR tree-optimization/82819
9134 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9135 code generating pluses that are no-ops in the target precision.
9137 2018-01-24 Richard Biener <rguenther@suse.de>
9140 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9142 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9144 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9145 to merge probabilities.
9146 * predict.c (probably_never_executed): Also mark as cold functions
9147 with global 0 profile and guessed local profile.
9148 * profile-count.c (profile_probability::combine_with_count): New
9150 * profile-count.h (profile_probability::operator*,
9151 profile_probability::operator*=, profile_probability::operator/,
9152 profile_probability::operator/=): Reduce precision to adjusted
9153 and set value to guessed on contradictory divisions.
9154 (profile_probability::combine_with_freq): Remove.
9155 (profile_probability::combine_wiht_count): Declare.
9156 (profile_count::force_nonzero):: Set to adjusted.
9157 (profile_count::probability_in):: Set quality to adjusted.
9158 * tree-ssa-tail-merge.c (replace_block_by): Use
9161 2018-01-23 Andrew Waterman <andrew@sifive.com>
9162 Jim Wilson <jimw@sifive.com>
9164 * config/riscv/riscv.c (riscv_stack_boundary): New.
9165 (riscv_option_override): Set riscv_stack_boundary. Handle
9166 riscv_preferred_stack_boundary_arg.
9167 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9168 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9169 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9170 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9171 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9172 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9174 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9177 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9178 of struct ix86_frame.
9179 (ix86_expand_epilogue): Likewise. Add a local variable for
9180 the reg_save_offset field in struct ix86_frame.
9182 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9184 PR tree-optimization/82604
9185 * tree-loop-distribution.c (enum partition_kind): New enum item
9186 PKIND_PARTIAL_MEMSET.
9187 (partition_builtin_p): Support above new enum item.
9188 (generate_code_for_partition): Ditto.
9189 (compute_access_range): Differentiate cases that equality can be
9190 proven at all loops, the innermost loops or no loops.
9191 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9192 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9193 (finalize_partitions, distribute_loop): Don't fuse partition of
9194 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9195 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9198 2018-01-23 Martin Liska <mliska@suse.cz>
9200 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9201 order to ignore the predictor.
9202 (PRED_POLYMORPHIC_CALL): Likewise.
9203 (PRED_RECURSIVE_CALL): Likewise.
9205 2018-01-23 Martin Liska <mliska@suse.cz>
9207 * tree-profile.c (tree_profiling): Print function header to
9208 aware reader which function we are working on.
9209 * value-prof.c (gimple_find_values_to_profile): Do not print
9210 not interesting value histograms.
9212 2018-01-23 Martin Liska <mliska@suse.cz>
9214 * profile-count.h (enum profile_quality): Add
9215 profile_uninitialized as the first value. Do not number values
9216 as they are zero based.
9217 (profile_count::verify): Update sanity check.
9218 (profile_probability::verify): Likewise.
9220 2018-01-23 Nathan Sidwell <nathan@acm.org>
9222 * doc/invoke.texi (ffor-scope): Deprecate.
9224 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9226 PR tree-optimization/83510
9227 * domwalk.c (set_all_edges_as_executable): New function.
9228 (dom_walker::dom_walker): Convert bool param
9229 "skip_unreachable_blocks" to enum reachability. Move setup of
9230 edge flags to set_all_edges_as_executable and only do it when
9231 reachability is REACHABLE_BLOCKS.
9232 * domwalk.h (enum dom_walker::reachability): New enum.
9233 (dom_walker::dom_walker): Convert bool param
9234 "skip_unreachable_blocks" to enum reachability.
9235 (set_all_edges_as_executable): New decl.
9236 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9237 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9239 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9240 but converting true to REACHABLE_BLOCKS.
9241 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9243 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9244 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9245 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9247 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9248 if check_all_array_refs will be called.
9250 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9252 * tree.c (selftest::test_location_wrappers): Add more test
9255 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9257 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9258 (selftest::test_bit_in_range): Likewise.
9260 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9263 * doc/sourcebuild.texi (vect_float): Say that the selector
9264 only describes the situation when -funsafe-math-optimizations is on.
9265 (vect_float_strict): Document.
9267 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9269 PR tree-optimization/83965
9270 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9271 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9272 instead of checking only for a reduction.
9273 (vect_recog_widen_sum_pattern): Likewise.
9275 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9277 * predict.c (probably_never_executed): Only use precise profile info.
9278 (compute_function_frequency): Skip after inlining hack since we now
9279 have quality checking.
9281 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9283 * profile-count.h (profile_probability::very_unlikely,
9284 profile_probability::unlikely, profile_probability::even): Set
9285 precision to guessed.
9287 2018-01-23 Richard Biener <rguenther@suse.de>
9289 PR tree-optimization/83963
9290 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9291 Properly terminate dominator walk when crossing the exit edge not
9292 when visiting its source block.
9294 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9297 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9298 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9300 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9302 PR tree-optimization/83957
9303 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9304 semicolon after for body surrounded by braces.
9306 PR tree-optimization/83081
9307 * profile-count.h (profile_probability::split): New method.
9308 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9309 Use profile_probability::split.
9310 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9311 when splitting a single conditional jump into 2.
9313 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9315 PR tree-optimization/69452
9316 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9319 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9321 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9322 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9323 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9325 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9327 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9328 * config/rl78/rl78.md (movdi): New define_expand.
9329 * config/rl78/rl78.c (rl78_split_movdi): New function.
9331 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9334 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9336 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9337 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9338 128-bit to produce an UNSPEC move to get the double word with the
9339 signbit and then a shift directly to do signbit.
9340 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9341 implementation with a new version that just does either a direct
9342 move or a regular move. Move memory interface to separate insns.
9343 Move insns so they are next to the expander.
9344 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9345 with signbit move. Split big and little endian case.
9346 (signbit<mode>2_dm_mem_le): Likewise.
9347 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9348 (signbit<mode>2_dm2): Likewise.
9350 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9352 * config/rl78/rl78.md (anddi3): New define_expand.
9354 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9356 * config/rl78/rl78.md (umindi3): New define_expand.
9358 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9360 * config/rl78/rl78.md (smindi3): New define_expand.
9362 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9364 * config/rl78/rl78.md (smaxdi3): New define_expand.
9366 2018-01-22 Carl Love <cel@us.ibm.com>
9368 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9369 LVX_V1TI): Add macro expansion.
9370 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9371 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9372 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9373 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9374 Change check to determine if the instruction is a byte reversing
9375 entry. Fix typo in comment.
9376 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9377 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9378 Add def_builtin calls for new builtins.
9379 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9380 Add define_insn expansion.
9382 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9384 * config/rl78/rl78.md (umaxdi3): New define_expand.
9386 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9388 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9389 for non-QImode registers.
9391 2018-01-22 Richard Biener <rguenther@suse.de>
9393 PR tree-optimization/83963
9394 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9395 including the loop exit block.
9396 (scop_detection::merge_sese): Likewise.
9397 (scop_detection::add_scop): Do it here instead.
9399 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9401 * doc/sourcebuild.texi (arm_softfloat): Document.
9403 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9406 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9407 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9408 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9410 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9411 David Edelsohn <dje.gcc@gmail.com>
9414 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9415 Change "crset eq" to "crset 2".
9416 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9417 (*call_indirect_aix<mode>_nospec): Likewise.
9418 (*call_value_indirect_aix<mode>_nospec): Likewise.
9419 (*call_indirect_elfv2<mode>_nospec): Likewise.
9420 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9421 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9422 change assembly output from . to $.
9423 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9424 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9425 (*tablejump<mode>_internal1_nospec): Likewise.
9427 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9430 * config/sh/sh_optimize_sett_clrt.cc:
9431 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9433 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9435 PR tree-optimization/83940
9436 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9437 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9438 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9439 use it to pass back the definition type.
9440 (vect_check_store_rhs): Likewise rhs_dt_out.
9441 (vect_build_gather_load_calls): Add a mask_dt argument and use
9442 it instead of a call to vect_is_simple_use.
9443 (vectorizable_store): Update calls to vect_check_load_store_mask
9444 and vect_check_store_rhs. Use the dt returned by the latter instead
9445 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9446 instead of calls to vect_is_simple_use. Pass the scalar rather
9447 than the vector operand to vect_is_simple_use when handling
9448 second and subsequent copies of an rhs value.
9449 (vectorizable_load): Update calls to vect_check_load_store_mask
9450 and vect_build_gather_load_calls. Use the cached mask_dt and
9451 gs_info.offset_dt instead of calls to vect_is_simple_use.
9453 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9456 * tree-emutls.c: Include gimplify.h.
9457 (lower_emutls_2): New function.
9458 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9459 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9460 it before further processing.
9463 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9464 UINTVAL (trueop1) instead of INTVAL (op1).
9466 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9470 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9471 INCOMING_FRAME_SP_OFFSET if not defined.
9472 (scan_trace): Add ENTRY argument. If true and
9473 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9474 emit a note to adjust the CFA offset.
9475 (create_cfi_notes): Adjust scan_trace callers.
9476 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9477 INCOMING_FRAME_SP_OFFSET in the CIE.
9478 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9479 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9481 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9482 * doc/tm.texi: Regenerated.
9484 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9486 PR rtl-optimization/83147
9487 * lra-constraints.c (remove_inheritance_pseudos): Use
9488 lra_substitute_pseudo_within_insn.
9490 2018-01-19 Tom de Vries <tom@codesourcery.com>
9491 Cesar Philippidis <cesar@codesourcery.com>
9494 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9496 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9499 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9500 spaces for function labels.
9502 2018-01-19 Martin Liska <mliska@suse.cz>
9504 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9505 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9506 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9507 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9508 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9509 (PRED_CONST_RETURN): Change from 69 to 65.
9510 (PRED_NULL_RETURN): Change from 91 to 71.
9511 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9512 (PRED_LOOP_GUARD): Change from 66 to 73.
9514 2018-01-19 Martin Liska <mliska@suse.cz>
9516 * predict.c (predict_insn_def): Add new assert.
9517 (struct branch_predictor): Change type to signed integer.
9518 (test_prediction_value_range): Amend test to cover
9520 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9521 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9522 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9523 (PRED_LOOP_IV_COMPARE): Likewise.
9524 * predict.h (PROB_UNINITIALIZED): Define new constant.
9526 2018-01-19 Martin Liska <mliska@suse.cz>
9528 * predict.c (dump_prediction): Add new format for
9529 analyze_brprob.py script which is enabled with -details
9531 * profile-count.h (precise_p): New function.
9533 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9535 PR tree-optimization/83922
9536 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9537 there are no statements that need masking.
9538 (vect_active_double_reduction_p): New function.
9539 (vect_analyze_loop_operations): Use it when handling phis that
9540 are not in the loop header.
9542 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9544 PR tree-optimization/83914
9545 * tree-vect-loop.c (vectorizable_induction): Don't convert
9546 init_expr or apply the peeling adjustment for inductions
9547 that are nested within the vectorized loop.
9549 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9551 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9554 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9558 * function.h (gimplify_parameters): Add gimple_seq * argument.
9559 * function.c: Include gimple.h and options.h.
9560 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9561 for the added local temporaries if needed.
9562 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9563 if there are any parameter cleanups, wrap whole body into a
9564 try/finally with the cleanups.
9566 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9569 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9570 Use GET_MODE_CLASS for scalar floating point.
9572 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9576 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9577 Fix call of call_cgraph_insertion_hooks.
9579 2018-01-18 Martin Sebor <msebor@redhat.com>
9581 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9583 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9586 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9589 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9592 * common.opt: (-ffile-prefix-map): New option.
9593 * opts.c (common_handle_option): Defer it.
9594 * opts-global.c (handle_common_deferred_options): Handle it.
9595 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9596 * file-prefix-map.h: New file.
9597 (remap_debug_filename, add_debug_prefix_map): ...here.
9598 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9599 * final.c (debug_prefix_map, add_debug_prefix_map
9600 remap_debug_filename): Move to...
9601 * file-prefix-map.c: New file.
9602 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9603 generalize, get rid of alloca(), use strrchr() instead of strchr().
9604 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9605 Implement in terms of add_prefix_map().
9606 (remap_macro_filename, remap_debug_filename): Implement in term of
9608 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9609 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9610 * dbxout.c: Include file-prefix-map.h.
9611 * varasm.c: Likewise.
9612 * vmsdbgout.c: Likewise.
9613 * xcoffout.c: Likewise.
9614 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9615 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9616 * doc/invoke.texi (-ffile-prefix-map): Document.
9617 (-fdebug-prefix-map): Update description.
9619 2018-01-18 Martin Liska <mliska@suse.cz>
9621 * config/i386/i386.c (indirect_thunk_name): Document that also
9623 (output_indirect_thunk): Document why both instructions
9624 (pause and lfence) are generated.
9626 2018-01-18 Richard Biener <rguenther@suse.de>
9628 PR tree-optimization/83887
9629 * graphite-scop-detection.c
9630 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9631 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9632 (scop_detection::merge_sese): Re-implement with a flood-fill
9633 algorithm that properly finds a SESE region if it exists.
9635 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9638 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9639 pointer_diff optimizations use view_convert instead of convert.
9641 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9643 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9644 Generate different code for -mno-speculate-indirect-jumps.
9645 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9646 (*call_indirect_aix<mode>): Disable for
9647 -mno-speculate-indirect-jumps.
9648 (*call_indirect_aix<mode>_nospec): New define_insn.
9649 (*call_value_indirect_aix<mode>): Disable for
9650 -mno-speculate-indirect-jumps.
9651 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9652 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9653 -mno-speculate-indirect-jumps.
9654 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9656 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9658 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9659 long double type, set the flags for noting the default long double
9660 type, even if we don't pass or return a long double type.
9662 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9665 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9668 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9670 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9671 support for merge[hl].
9672 (fold_mergehl_helper): New helper function.
9673 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9674 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9675 (altivec_vmrglw_direct): Add xxmrglw insn.
9677 2018-01-17 Andrew Waterman <andrew@sifive.com>
9679 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9680 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9682 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9685 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9686 call the lto_location_cache before reading the
9687 DECL_SOURCE_LOCATION of the types.
9689 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9690 Richard Sandiford <richard.sandiford@linaro.org>
9692 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9693 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9694 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9695 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9696 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9698 * config/aarch64/constraints.md (aarch64_movti_operand):
9700 * config/aarch64/predicates.md (Uti): Add new constraint.
9702 2018-01-17 Carl Love <cel@us.ibm.com>
9704 * config/rs6000/vsx.md (define_expand xl_len_r,
9705 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9706 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9708 (define_expand, define_insn): Move the shift left from the
9709 define_insn to the define_expand for lxvl and stxvl instructions.
9710 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9711 and XL_LEN_R definitions to PURE.
9713 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9715 * config/i386/i386.c (indirect_thunk_name): Declare regno
9716 as unsigned int. Compare regno with INVALID_REGNUM.
9717 (output_indirect_thunk): Ditto.
9718 (output_indirect_thunk_function): Ditto.
9719 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9720 in the call to output_indirect_thunk_function.
9722 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9725 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9726 rather than the size of inner_type to determine the stack slot size
9727 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9729 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9732 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9735 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9737 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9738 endian Linux systems to optionally enable multilibs for selecting
9739 the long double type if the user configured an explicit type.
9740 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9741 have no long double multilibs if not defined.
9742 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9743 warn if the user used -mabi={ieee,ibm}longdouble and we built
9744 multilibs for long double.
9745 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9746 appropriate multilib option.
9747 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9749 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9750 for building long double multilibs.
9751 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9753 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9755 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9758 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9760 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9763 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9766 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9769 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9771 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9772 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9774 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9776 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9777 different rtl trees depending on TARGET_64BIT.
9778 (rs6000_gen_lvx): Likewise.
9780 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9782 * config/visium/visium.md (nop): Tweak comment.
9783 (hazard_nop): Likewise.
9785 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9787 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9788 -mspeculate-indirect-jumps.
9789 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9790 for -mno-speculate-indirect-jumps.
9791 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9792 (*call_value_indirect_elfv2<mode>): Disable for
9793 -mno-speculate-indirect-jumps.
9794 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9795 (indirect_jump): Emit different RTL for
9796 -mno-speculate-indirect-jumps.
9797 (*indirect_jump<mode>): Disable for
9798 -mno-speculate-indirect-jumps.
9799 (*indirect_jump<mode>_nospec): New define_insn.
9800 (tablejump): Emit different RTL for
9801 -mno-speculate-indirect-jumps.
9802 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9803 (tablejumpsi_nospec): New define_expand.
9804 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9805 (tablejumpdi_nospec): New define_expand.
9806 (*tablejump<mode>_internal1): Disable for
9807 -mno-speculate-indirect-jumps.
9808 (*tablejump<mode>_internal1_nospec): New define_insn.
9809 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9812 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9814 * caller-save.c (insert_save): Drop unnecessary parameter. All
9817 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9818 Richard Biener <rguenth@suse.de>
9821 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9822 return early, inline manually is_gimple_sizepos. Make sure if we
9823 call gimplify_expr we don't end up with a gimple constant.
9824 * tree.c (variably_modified_type_p): Don't return true for
9825 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9826 * gimplify.h (is_gimple_sizepos): Remove.
9828 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9830 PR tree-optimization/83857
9831 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9832 vectorizable_live_operation for pure SLP statements.
9833 (vectorizable_live_operation): Handle PHIs.
9835 2018-01-16 Richard Biener <rguenther@suse.de>
9837 PR tree-optimization/83867
9838 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9839 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9841 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9844 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9845 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9846 If off is not INTEGER_CST, issue a may not be aligned warning
9847 rather than isn't aligned. Use isn%'t rather than isn't.
9848 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9850 <case MULT_EXPR>: Improve the case when bottom and one of the
9851 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9852 operand, in that case check if the other operand is multiple of
9853 bottom divided by the INTEGER_CST operand.
9855 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9858 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9859 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9860 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9861 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9862 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9863 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9864 (pa_function_arg_size): New function.
9866 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9868 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9869 in a separate statement.
9871 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9873 PR tree-optimization/83847
9874 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9875 group gathers and scatters.
9877 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9879 PR rtl-optimization/86620
9880 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9882 PR rtl-optimization/83213
9883 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9884 to last if both are JUMP_INSNs.
9886 PR tree-optimization/83843
9887 * gimple-ssa-store-merging.c
9888 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9889 store_immediate_info for bswap/nop orig_stores.
9891 2018-01-15 Andrew Waterman <andrew@sifive.com>
9893 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9895 <UDIV>: Increase cost if !TARGET_DIV.
9897 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9899 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9900 (define_attr "cr_logical_3op"): New.
9901 (cceq_ior_compare): Adjust.
9902 (cceq_ior_compare_complement): Adjust.
9903 (*cceq_rev_compare): Adjust.
9904 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9905 (is_cracked_insn): Adjust.
9906 (insn_must_be_first_in_group): Adjust.
9907 * config/rs6000/40x.md: Adjust.
9908 * config/rs6000/440.md: Adjust.
9909 * config/rs6000/476.md: Adjust.
9910 * config/rs6000/601.md: Adjust.
9911 * config/rs6000/603.md: Adjust.
9912 * config/rs6000/6xx.md: Adjust.
9913 * config/rs6000/7450.md: Adjust.
9914 * config/rs6000/7xx.md: Adjust.
9915 * config/rs6000/8540.md: Adjust.
9916 * config/rs6000/cell.md: Adjust.
9917 * config/rs6000/e300c2c3.md: Adjust.
9918 * config/rs6000/e500mc.md: Adjust.
9919 * config/rs6000/e500mc64.md: Adjust.
9920 * config/rs6000/e5500.md: Adjust.
9921 * config/rs6000/e6500.md: Adjust.
9922 * config/rs6000/mpc.md: Adjust.
9923 * config/rs6000/power4.md: Adjust.
9924 * config/rs6000/power5.md: Adjust.
9925 * config/rs6000/power6.md: Adjust.
9926 * config/rs6000/power7.md: Adjust.
9927 * config/rs6000/power8.md: Adjust.
9928 * config/rs6000/power9.md: Adjust.
9929 * config/rs6000/rs64.md: Adjust.
9930 * config/rs6000/titan.md: Adjust.
9932 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9934 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9935 ix86_indirect_branch_register logic.
9937 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9939 * config/i386/constraints.md (Bs): Update
9940 ix86_indirect_branch_register check. Don't check
9941 ix86_indirect_branch_register with GOT_memory_operand.
9943 * config/i386/predicates.md (GOT_memory_operand): Don't check
9944 ix86_indirect_branch_register here.
9945 (GOT32_symbol_operand): Likewise.
9947 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9949 * config/i386/predicates.md (constant_call_address_operand):
9950 Rewrite ix86_indirect_branch_register logic.
9951 (sibcall_insn_operand): Likewise.
9953 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9955 * config/i386/constraints.md (Bs): Replace
9956 ix86_indirect_branch_thunk_register with
9957 ix86_indirect_branch_register.
9959 * config/i386/i386.md (indirect_jump): Likewise.
9960 (tablejump): Likewise.
9961 (*sibcall_memory): Likewise.
9962 (*sibcall_value_memory): Likewise.
9963 Peepholes of indirect call and jump via memory: Likewise.
9964 * config/i386/i386.opt: Likewise.
9965 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9966 (GOT_memory_operand): Likewise.
9967 (call_insn_operand): Likewise.
9968 (sibcall_insn_operand): Likewise.
9969 (GOT32_symbol_operand): Likewise.
9971 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9974 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9975 type rather than type addr's type points to.
9976 (expand_omp_atomic_mutex): Likewise.
9977 (expand_omp_atomic): Likewise.
9979 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9982 * config/i386/i386.c (output_indirect_thunk_function): Use
9983 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9984 for __x86_return_thunk.
9986 2018-01-15 Richard Biener <rguenther@suse.de>
9989 * expmed.c (extract_bit_field_1): Fix typo.
9991 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9994 * config/arm/iterators.md (VF): New mode iterator.
9995 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9996 Remove integer-related logic from pattern.
9997 (neon_vabd<mode>_3): Likewise.
9999 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10001 PR middle-end/82694
10002 * common.opt (fstrict-overflow): No longer an alias.
10003 (fwrapv-pointer): New option.
10004 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10005 also for pointer types based on flag_wrapv_pointer.
10006 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10007 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10008 opts->x_flag_wrapv got set.
10009 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10010 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10011 POINTER_TYPE_OVERFLOW_UNDEFINED.
10012 * match.pd: Likewise in address comparison pattern.
10013 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10015 2018-01-15 Richard Biener <rguenther@suse.de>
10018 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10019 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10020 Reset type names to their identifier if their TYPE_DECL doesn't
10021 have linkage (and thus is used for ODR and devirt).
10022 (save_debug_info_for_decl): Remove.
10023 (save_debug_info_for_type): Likewise.
10024 (add_tree_to_fld_list): Adjust.
10025 * tree-pretty-print.c (dump_generic_node): Make dumping of
10026 type names more robust.
10028 2018-01-15 Richard Biener <rguenther@suse.de>
10030 * BASE-VER: Bump to 8.0.1.
10032 2018-01-14 Martin Sebor <msebor@redhat.com>
10035 * builtins.c (check_access): Avoid warning when the no-warning bit
10038 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10040 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10041 * ira-color (allocno_hard_regs_compare): Likewise.
10043 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10046 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10047 Use .pushsection/.popsection.
10049 2018-01-14 Martin Sebor <msebor@redhat.com>
10052 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10054 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10056 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10057 entry from extra_headers.
10058 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10059 extra_headers, make the list bitwise identical to the i?86-*-* one.
10061 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10063 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10064 -mcmodel=large with -mindirect-branch=thunk,
10065 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10066 -mfunction-return=thunk-extern.
10067 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10068 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10069 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10071 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10073 * config/i386/i386.c (print_reg): Print the name of the full
10074 integer register without '%'.
10075 (ix86_print_operand): Handle 'V'.
10076 * doc/extend.texi: Document 'V' modifier.
10078 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10080 * config/i386/constraints.md (Bs): Disallow memory operand for
10081 -mindirect-branch-register.
10083 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10084 (GOT_memory_operand): Likewise.
10085 (call_insn_operand): Likewise.
10086 (sibcall_insn_operand): Likewise.
10087 (GOT32_symbol_operand): Likewise.
10088 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10089 for -mindirect-branch-register.
10090 (tablejump): Likewise.
10091 (*sibcall_memory): Likewise.
10092 (*sibcall_value_memory): Likewise.
10093 Disallow peepholes of indirect call and jump via memory for
10094 -mindirect-branch-register.
10095 (*call_pop): Replace m with Bw.
10096 (*call_value_pop): Likewise.
10097 (*sibcall_pop_memory): Replace m with Bs.
10098 * config/i386/i386.opt (mindirect-branch-register): New option.
10099 * doc/invoke.texi: Document -mindirect-branch-register option.
10101 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10103 * config/i386/i386-protos.h (ix86_output_function_return): New.
10104 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10105 set function_return_type.
10106 (indirect_thunk_name): Add ret_p to indicate thunk for function
10108 (output_indirect_thunk_function): Pass false to
10109 indirect_thunk_name.
10110 (ix86_output_indirect_branch_via_reg): Likewise.
10111 (ix86_output_indirect_branch_via_push): Likewise.
10112 (output_indirect_thunk_function): Create alias for function
10113 return thunk if regno < 0.
10114 (ix86_output_function_return): New function.
10115 (ix86_handle_fndecl_attribute): Handle function_return.
10116 (ix86_attribute_table): Add function_return.
10117 * config/i386/i386.h (machine_function): Add
10118 function_return_type.
10119 * config/i386/i386.md (simple_return_internal): Use
10120 ix86_output_function_return.
10121 (simple_return_internal_long): Likewise.
10122 * config/i386/i386.opt (mfunction-return=): New option.
10123 (indirect_branch): Mention -mfunction-return=.
10124 * doc/extend.texi: Document function_return function attribute.
10125 * doc/invoke.texi: Document -mfunction-return= option.
10127 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10129 * config/i386/i386-opts.h (indirect_branch): New.
10130 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10131 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10132 with local indirect jump when converting indirect call and jump.
10133 (ix86_set_indirect_branch_type): New.
10134 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10135 (indirectlabelno): New.
10136 (indirect_thunk_needed): Likewise.
10137 (indirect_thunk_bnd_needed): Likewise.
10138 (indirect_thunks_used): Likewise.
10139 (indirect_thunks_bnd_used): Likewise.
10140 (INDIRECT_LABEL): Likewise.
10141 (indirect_thunk_name): Likewise.
10142 (output_indirect_thunk): Likewise.
10143 (output_indirect_thunk_function): Likewise.
10144 (ix86_output_indirect_branch_via_reg): Likewise.
10145 (ix86_output_indirect_branch_via_push): Likewise.
10146 (ix86_output_indirect_branch): Likewise.
10147 (ix86_output_indirect_jmp): Likewise.
10148 (ix86_code_end): Call output_indirect_thunk_function if needed.
10149 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10151 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10152 (ix86_attribute_table): Add indirect_branch.
10153 * config/i386/i386.h (machine_function): Add indirect_branch_type
10154 and has_local_indirect_jump.
10155 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10157 (tablejump): Likewise.
10158 (*indirect_jump): Use ix86_output_indirect_jmp.
10159 (*tablejump_1): Likewise.
10160 (simple_return_indirect_internal): Likewise.
10161 * config/i386/i386.opt (mindirect-branch=): New option.
10162 (indirect_branch): New.
10165 (thunk-inline): Likewise.
10166 (thunk-extern): Likewise.
10167 * doc/extend.texi: Document indirect_branch function attribute.
10168 * doc/invoke.texi: Document -mindirect-branch= option.
10170 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10173 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10175 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10177 * ipa-inline.c (want_inline_small_function_p): Return false if
10178 inlining has already failed with CIF_FINAL_ERROR.
10179 (update_caller_keys): Call want_inline_small_function_p before
10181 (update_callee_keys): Likewise.
10183 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10185 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10187 (rs6000_quadword_masked_address_p): Likewise.
10188 (quad_aligned_load_p): Likewise.
10189 (quad_aligned_store_p): Likewise.
10190 (const_load_sequence_p): Add comment to describe the outer-most loop.
10191 (mimic_memory_attributes_and_flags): New function.
10192 (rs6000_gen_stvx): Likewise.
10193 (replace_swapped_aligned_store): Likewise.
10194 (rs6000_gen_lvx): Likewise.
10195 (replace_swapped_aligned_load): Likewise.
10196 (replace_swapped_load_constant): Capitalize argument name in
10197 comment describing this function.
10198 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10199 and stores that access quad-word aligned addresses and replace
10200 with stvx or lvx instructions when appropriate.
10201 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10202 New function prototype.
10203 (rs6000_quadword_masked_address_p): Likewise.
10204 (rs6000_gen_lvx): Likewise.
10205 (rs6000_gen_stvx): Likewise.
10206 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10207 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10208 when memory address is aligned.
10209 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10210 this split to select lvx instruction when memory address is aligned.
10211 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10212 instruction when memory address is aligned.
10213 (*vsx_le_perm_load_v16qi): Likewise.
10214 (four unnamed splitters): Modify to select the stvx instruction
10215 when memory is aligned.
10217 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10219 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10220 which appears in the queue multiple times.
10222 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10223 Alan Hayward <alan.hayward@arm.com>
10224 David Sherwood <david.sherwood@arm.com>
10226 * tree-vectorizer.h (vec_lower_bound): New structure.
10227 (_loop_vec_info): Add check_nonzero and lower_bounds.
10228 (LOOP_VINFO_CHECK_NONZERO): New macro.
10229 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10230 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10231 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10232 fields. Make seg_len the distance travelled, not including the
10234 (dr_direction_indicator): Declare.
10235 (dr_zero_step_indicator): Likewise.
10236 (dr_known_forward_stride_p): Likewise.
10237 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10239 (runtime_alias_check_p): Allow runtime alias checks with
10241 (operator ==): Compare access_size and align.
10242 (prune_runtime_alias_test_list): Rework for new distinction between
10243 the access_size and seg_len.
10244 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10246 (get_segment_min_max): New function.
10247 (create_intersect_range_checks): Use it.
10248 (dr_step_indicator): New function.
10249 (dr_direction_indicator): Likewise.
10250 (dr_zero_step_indicator): Likewise.
10251 (dr_known_forward_stride_p): Likewise.
10252 * tree-loop-distribution.c (data_ref_segment_size): Return
10253 DR_STEP * (niters - 1).
10254 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10256 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10257 (vect_preserves_scalar_order_p): New function, split out from...
10258 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10259 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10260 (vect_vfa_access_size): New function.
10261 (vect_vfa_align): Likewise.
10262 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10263 (dump_lower_bound): New function.
10264 (vect_check_lower_bound): Likewise.
10265 (vect_small_gap_p): Likewise.
10266 (vectorizable_with_step_bound_p): Likewise.
10267 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10268 depencies if the vectorization factor is 1. Convert the checks
10269 for nonzero steps into checks on the bounds of DR_STEP. Try using
10270 a bunds check for variable steps if the minimum required step is
10271 relatively small. Update calls to the dr_with_seg_len
10272 constructor and to vect_compile_time_alias.
10273 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10275 (vect_loop_versioning): Call it.
10276 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10278 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10280 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10281 Alan Hayward <alan.hayward@arm.com>
10282 David Sherwood <david.sherwood@arm.com>
10284 * doc/sourcebuild.texi (vect_scatter_store): Document.
10285 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10287 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10289 * genopinit.c (main): Add supports_vec_scatter_store and
10290 supports_vec_scatter_store_cached to target_optabs.
10291 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10292 IFN_MASK_SCATTER_STORE.
10293 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10295 * internal-fn.h (internal_store_fn_p): Declare.
10296 (internal_fn_stored_value_index): Likewise.
10297 * internal-fn.c (scatter_store_direct): New macro.
10298 (expand_scatter_store_optab_fn): New function.
10299 (direct_scatter_store_optab_supported_p): New macro.
10300 (internal_store_fn_p): New function.
10301 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10302 IFN_MASK_SCATTER_STORE.
10303 (internal_fn_mask_index): Likewise.
10304 (internal_fn_stored_value_index): New function.
10305 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10306 for scatter stores.
10307 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10308 * optabs-query.c (supports_vec_scatter_store_p): New function.
10309 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10310 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10311 true for scatter stores.
10312 (vect_gather_scatter_fn_p): Handle scatter stores too.
10313 (vect_check_gather_scatter): Consider using scatter stores if
10314 supports_vec_scatter_store_p.
10315 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10316 scatter stores too.
10317 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10318 internal_fn_stored_value_index.
10319 (check_load_store_masking): Handle scatter stores too.
10320 (vect_get_store_rhs): Make public.
10321 (vectorizable_call): Use internal_store_fn_p.
10322 (vectorizable_store): Handle scatter store internal functions.
10323 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10324 when deciding whether the end of the group has been reached.
10325 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10326 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10327 (mask_scatter_store<mode>): New insns.
10329 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10330 Alan Hayward <alan.hayward@arm.com>
10331 David Sherwood <david.sherwood@arm.com>
10333 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10334 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10335 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10337 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10338 Use vect_truncate_gather_scatter_offset if we can't treat the
10339 operation as a normal gather load or scatter store.
10340 (get_group_load_store_type): Take the gather_scatter_info
10341 as argument. Try using a gather load or scatter store for
10342 single-element groups.
10343 (get_load_store_type): Update calls to get_group_load_store_type
10344 and vect_use_strided_gather_scatters_p.
10346 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10347 Alan Hayward <alan.hayward@arm.com>
10348 David Sherwood <david.sherwood@arm.com>
10350 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10351 optional tree argument.
10352 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10354 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10355 but continue to use the current value as a fallback.
10356 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10357 to compare the updates.
10358 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10359 (get_load_store_type): Use it when handling a strided access.
10360 (vect_get_strided_load_store_ops): New function.
10361 (vect_get_data_ptr_increment): Likewise.
10362 (vectorizable_load): Handle strided gather loads. Always pass
10363 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10365 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10366 Alan Hayward <alan.hayward@arm.com>
10367 David Sherwood <david.sherwood@arm.com>
10369 * doc/md.texi (gather_load@var{m}): Document.
10370 (mask_gather_load@var{m}): Likewise.
10371 * genopinit.c (main): Add supports_vec_gather_load and
10372 supports_vec_gather_load_cached to target_optabs.
10373 * optabs-tree.c (init_tree_optimization_optabs): Use
10374 ggc_cleared_alloc to allocate target_optabs.
10375 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10376 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10378 * internal-fn.h (internal_load_fn_p): Declare.
10379 (internal_gather_scatter_fn_p): Likewise.
10380 (internal_fn_mask_index): Likewise.
10381 (internal_gather_scatter_fn_supported_p): Likewise.
10382 * internal-fn.c (gather_load_direct): New macro.
10383 (expand_gather_load_optab_fn): New function.
10384 (direct_gather_load_optab_supported_p): New macro.
10385 (direct_internal_fn_optab): New function.
10386 (internal_load_fn_p): Likewise.
10387 (internal_gather_scatter_fn_p): Likewise.
10388 (internal_fn_mask_index): Likewise.
10389 (internal_gather_scatter_fn_supported_p): Likewise.
10390 * optabs-query.c (supports_at_least_one_mode_p): New function.
10391 (supports_vec_gather_load_p): Likewise.
10392 * optabs-query.h (supports_vec_gather_load_p): Declare.
10393 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10394 and memory_type field.
10395 (NUM_PATTERNS): Bump to 15.
10396 * tree-vect-data-refs.c: Include internal-fn.h.
10397 (vect_gather_scatter_fn_p): New function.
10398 (vect_describe_gather_scatter_call): Likewise.
10399 (vect_check_gather_scatter): Try using internal functions for
10400 gather loads. Recognize existing calls to a gather load function.
10401 (vect_analyze_data_refs): Consider using gather loads if
10402 supports_vec_gather_load_p.
10403 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10404 (vect_get_gather_scatter_offset_type): Likewise.
10405 (vect_convert_mask_for_vectype): Likewise.
10406 (vect_add_conversion_to_patterm): Likewise.
10407 (vect_try_gather_scatter_pattern): Likewise.
10408 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10409 (vect_vect_recog_func_ptrs): Add it.
10410 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10411 internal_fn_mask_index and internal_gather_scatter_fn_p.
10412 (check_load_store_masking): Take the gather_scatter_info as an
10413 argument and handle gather loads.
10414 (vect_get_gather_scatter_ops): New function.
10415 (vectorizable_call): Check internal_load_fn_p.
10416 (vectorizable_load): Likewise. Handle gather load internal
10418 (vectorizable_store): Update call to check_load_store_masking.
10419 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10420 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10421 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10422 (aarch64_gather_scale_operand_d): New predicates.
10423 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10424 (mask_gather_load<mode>): New insns.
10426 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10427 Alan Hayward <alan.hayward@arm.com>
10428 David Sherwood <david.sherwood@arm.com>
10430 * optabs.def (fold_left_plus_optab): New optab.
10431 * doc/md.texi (fold_left_plus_@var{m}): Document.
10432 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10433 * internal-fn.c (fold_left_direct): Define.
10434 (expand_fold_left_optab_fn): Likewise.
10435 (direct_fold_left_optab_supported_p): Likewise.
10436 * fold-const-call.c (fold_const_fold_left): New function.
10437 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10438 * tree-parloops.c (valid_reduction_p): New function.
10439 (gather_scalar_reductions): Use it.
10440 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10441 (vect_finish_replace_stmt): Declare.
10442 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10443 (needs_fold_left_reduction_p): New function, split out from...
10444 (vect_is_simple_reduction): ...here. Accept reductions that
10445 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10446 (vect_force_simple_reduction): Also store the reduction type in
10447 the assignment's STMT_VINFO_REDUC_TYPE.
10448 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10449 (merge_with_identity): New function.
10450 (vect_expand_fold_left): Likewise.
10451 (vectorize_fold_left_reduction): Likewise.
10452 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10453 scalar phi in place for it. Check for target support and reject
10454 cases that would reassociate the operation. Defer the transform
10455 phase to vectorize_fold_left_reduction.
10456 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10457 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10458 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10460 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10462 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10463 call to ifc_temp_var.
10465 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10466 Alan Hayward <alan.hayward@arm.com>
10467 David Sherwood <david.sherwood@arm.com>
10469 * target.def (legitimize_address_displacement): Take the original
10470 offset as a poly_int.
10471 * targhooks.h (default_legitimize_address_displacement): Update
10473 * targhooks.c (default_legitimize_address_displacement): Likewise.
10474 * doc/tm.texi: Regenerate.
10475 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10476 as an argument, moving assert of ad->disp == ad->disp_term to...
10477 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10478 Try calling targetm.legitimize_address_displacement before expanding
10479 the address rather than afterwards, and adjust for the new interface.
10480 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10481 Match the new hook interface. Handle SVE addresses.
10482 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10483 new hook interface.
10485 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10487 * Makefile.in (OBJS): Add early-remat.o.
10488 * target.def (select_early_remat_modes): New hook.
10489 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10490 * doc/tm.texi: Regenerate.
10491 * targhooks.h (default_select_early_remat_modes): Declare.
10492 * targhooks.c (default_select_early_remat_modes): New function.
10493 * timevar.def (TV_EARLY_REMAT): New timevar.
10494 * passes.def (pass_early_remat): New pass.
10495 * tree-pass.h (make_pass_early_remat): Declare.
10496 * early-remat.c: New file.
10497 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10499 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10501 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10502 Alan Hayward <alan.hayward@arm.com>
10503 David Sherwood <david.sherwood@arm.com>
10505 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10506 vfm1 with a bound_epilog parameter.
10507 (vect_do_peeling): Update calls accordingly, and move the prologue
10508 call earlier in the function. Treat the base bound_epilog as 0 for
10509 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10510 this base when peeling for gaps.
10511 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10512 with fully-masked loops.
10513 (vect_estimate_min_profitable_iters): Handle the single peeled
10514 iteration in that case.
10516 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10517 Alan Hayward <alan.hayward@arm.com>
10518 David Sherwood <david.sherwood@arm.com>
10520 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10521 single-element interleaving even if the size is not a power of 2.
10522 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10523 accesses for single-element interleaving if the group size is
10526 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10527 Alan Hayward <alan.hayward@arm.com>
10528 David Sherwood <david.sherwood@arm.com>
10530 * doc/md.texi (fold_extract_last_@var{m}): Document.
10531 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10532 * optabs.def (fold_extract_last_optab): New optab.
10533 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10534 * internal-fn.c (fold_extract_direct): New macro.
10535 (expand_fold_extract_optab_fn): Likewise.
10536 (direct_fold_extract_optab_supported_p): Likewise.
10537 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10538 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10539 EXTRACT_LAST_REDUCTION.
10540 (get_initial_def_for_reduction): Do not create an initial vector
10541 for EXTRACT_LAST_REDUCTION reductions.
10542 (vectorizable_reduction): Leave the scalar phi in place for
10543 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10544 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10545 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10546 transform phase to vectorizable_condition.
10547 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10549 (vect_finish_stmt_generation): ...here.
10550 (vect_finish_replace_stmt): New function.
10551 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10552 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10554 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10556 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10557 Alan Hayward <alan.hayward@arm.com>
10558 David Sherwood <david.sherwood@arm.com>
10560 * doc/md.texi (extract_last_@var{m}): Document.
10561 * optabs.def (extract_last_optab): New optab.
10562 * internal-fn.def (EXTRACT_LAST): New internal function.
10563 * internal-fn.c (cond_unary_direct): New macro.
10564 (expand_cond_unary_optab_fn): Likewise.
10565 (direct_cond_unary_optab_supported_p): Likewise.
10566 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10567 loops using EXTRACT_LAST.
10568 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10569 (extract_last_<mode>): ...this optab.
10570 (vec_extract<mode><Vel>): Update accordingly.
10572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10573 Alan Hayward <alan.hayward@arm.com>
10574 David Sherwood <david.sherwood@arm.com>
10576 * target.def (empty_mask_is_expensive): New hook.
10577 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10578 * doc/tm.texi: Regenerate.
10579 * targhooks.h (default_empty_mask_is_expensive): Declare.
10580 * targhooks.c (default_empty_mask_is_expensive): New function.
10581 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10582 if the target says that empty masks are expensive.
10583 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10585 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10587 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10588 Alan Hayward <alan.hayward@arm.com>
10589 David Sherwood <david.sherwood@arm.com>
10591 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10592 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10593 (vect_use_loop_mask_for_alignment_p): New function.
10594 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10595 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10596 niters_skip argument. Make sure that the first niters_skip elements
10597 of the first iteration are inactive.
10598 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10599 Update call to vect_set_loop_masks_directly.
10600 (get_misalign_in_elems): New function, split out from...
10601 (vect_gen_prolog_loop_niters): ...here.
10602 (vect_update_init_of_dr): Take a code argument that specifies whether
10603 the adjustment should be added or subtracted.
10604 (vect_update_init_of_drs): Likewise.
10605 (vect_prepare_for_masked_peels): New function.
10606 (vect_do_peeling): Skip prologue peeling if we're using a mask
10607 instead. Update call to vect_update_inits_of_drs.
10608 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10610 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10611 alignment. Do not include the number of peeled iterations in
10612 the minimum threshold in that case.
10613 (vectorizable_induction): Adjust the start value down by
10614 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10615 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10616 Take the number of skipped iterations into account when calculating
10618 * tree-vect-stmts.c (vect_gen_while_not): New function.
10620 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10621 Alan Hayward <alan.hayward@arm.com>
10622 David Sherwood <david.sherwood@arm.com>
10624 * doc/sourcebuild.texi (vect_fully_masked): Document.
10625 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10626 default value to 0.
10627 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10629 (vect_analyze_loop_2): ...here. Don't check the vectorization
10630 factor against the number of loop iterations if the loop is
10633 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10634 Alan Hayward <alan.hayward@arm.com>
10635 David Sherwood <david.sherwood@arm.com>
10637 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10638 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10639 (dump_groups): Update accordingly.
10640 (iv_use::mem_type): New member variable.
10641 (address_p): New function.
10642 (record_use): Add a mem_type argument and initialize the new
10644 (record_group_use): Add a mem_type argument. Use address_p.
10645 Remove obsolete null checks of base_object. Update call to record_use.
10646 (find_interesting_uses_op): Update call to record_group_use.
10647 (find_interesting_uses_cond): Likewise.
10648 (find_interesting_uses_address): Likewise.
10649 (get_mem_type_for_internal_fn): New function.
10650 (find_address_like_use): Likewise.
10651 (find_interesting_uses_stmt): Try find_address_like_use before
10652 calling find_interesting_uses_op.
10653 (addr_offset_valid_p): Use the iv mem_type field as the type
10654 of the addressed memory.
10655 (add_autoinc_candidates): Likewise.
10656 (get_address_cost): Likewise.
10657 (split_small_address_groups_p): Use address_p.
10658 (split_address_groups): Likewise.
10659 (add_iv_candidate_for_use): Likewise.
10660 (autoinc_possible_for_pair): Likewise.
10661 (rewrite_groups): Likewise.
10662 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10663 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10664 (get_alias_ptr_type_for_ptr_address): New function.
10665 (rewrite_use_address): Rewrite address uses in calls that were
10666 identified by find_address_like_use.
10668 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10669 Alan Hayward <alan.hayward@arm.com>
10670 David Sherwood <david.sherwood@arm.com>
10672 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10674 * gimple-expr.h (is_gimple_addressable: Likewise.
10675 * gimple-expr.c (is_gimple_address): Likewise.
10676 * internal-fn.c (expand_call_mem_ref): New function.
10677 (expand_mask_load_optab_fn): Use it.
10678 (expand_mask_store_optab_fn): Likewise.
10680 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10681 Alan Hayward <alan.hayward@arm.com>
10682 David Sherwood <david.sherwood@arm.com>
10684 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10685 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10686 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10687 (cond_umax@var{mode}): Document.
10688 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10689 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10690 (cond_umin_optab, cond_umax_optab): New optabs.
10691 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10692 (COND_IOR, COND_XOR): New internal functions.
10693 * internal-fn.h (get_conditional_internal_fn): Declare.
10694 * internal-fn.c (cond_binary_direct): New macro.
10695 (expand_cond_binary_optab_fn): Likewise.
10696 (direct_cond_binary_optab_supported_p): Likewise.
10697 (get_conditional_internal_fn): New function.
10698 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10699 Cope with reduction statements that are vectorized as calls rather
10701 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10702 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10703 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10704 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10705 (UNSPEC_COND_EOR): New unspecs.
10706 (optab): Add mappings for them.
10707 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10708 (sve_int_op, sve_fp_op): New int attributes.
10710 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10711 Alan Hayward <alan.hayward@arm.com>
10712 David Sherwood <david.sherwood@arm.com>
10714 * optabs.def (while_ult_optab): New optab.
10715 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10716 * internal-fn.def (WHILE_ULT): New internal function.
10717 * internal-fn.h (direct_internal_fn_supported_p): New override
10718 that takes two types as argument.
10719 * internal-fn.c (while_direct): New macro.
10720 (expand_while_optab_fn): New function.
10721 (convert_optab_supported_p): Likewise.
10722 (direct_while_optab_supported_p): New macro.
10723 * wide-int.h (wi::udiv_ceil): New function.
10724 * tree-vectorizer.h (rgroup_masks): New structure.
10725 (vec_loop_masks): New typedef.
10726 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10727 and fully_masked_p.
10728 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10729 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10730 (vect_max_vf): New function.
10731 (slpeel_make_loop_iterate_ntimes): Delete.
10732 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10733 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10734 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10735 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10736 internal-fn.h, stor-layout.h and optabs-query.h.
10737 (vect_set_loop_mask): New function.
10738 (add_preheader_seq): Likewise.
10739 (add_header_seq): Likewise.
10740 (interleave_supported_p): Likewise.
10741 (vect_maybe_permute_loop_masks): Likewise.
10742 (vect_set_loop_masks_directly): Likewise.
10743 (vect_set_loop_condition_masked): Likewise.
10744 (vect_set_loop_condition_unmasked): New function, split out from
10745 slpeel_make_loop_iterate_ntimes.
10746 (slpeel_make_loop_iterate_ntimes): Rename to..
10747 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10748 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10749 (vect_do_peeling): Update call accordingly.
10750 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10752 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10753 mask_compare_type, can_fully_mask_p and fully_masked_p.
10754 (release_vec_loop_masks): New function.
10755 (_loop_vec_info): Use it to free the loop masks.
10756 (can_produce_all_loop_masks_p): New function.
10757 (vect_get_max_nscalars_per_iter): Likewise.
10758 (vect_verify_full_masking): Likewise.
10759 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10760 retries, and free the mask rgroups before retrying. Check loop-wide
10761 reasons for disallowing fully-masked loops. Make the final decision
10762 about whether use a fully-masked loop or not.
10763 (vect_estimate_min_profitable_iters): Do not assume that peeling
10764 for the number of iterations will be needed for fully-masked loops.
10765 (vectorizable_reduction): Disable fully-masked loops.
10766 (vectorizable_live_operation): Likewise.
10767 (vect_halve_mask_nunits): New function.
10768 (vect_double_mask_nunits): Likewise.
10769 (vect_record_loop_mask): Likewise.
10770 (vect_get_loop_mask): Likewise.
10771 (vect_transform_loop): Handle the case in which the final loop
10772 iteration might handle a partial vector. Call vect_set_loop_condition
10773 instead of slpeel_make_loop_iterate_ntimes.
10774 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10775 (check_load_store_masking): New function.
10776 (prepare_load_store_mask): Likewise.
10777 (vectorizable_store): Handle fully-masked loops.
10778 (vectorizable_load): Likewise.
10779 (supportable_widening_operation): Use vect_halve_mask_nunits for
10781 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10782 (vect_gen_while): New function.
10783 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10784 (aarch64_uqdec<mode>): New insn.
10786 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10787 Alan Hayward <alan.hayward@arm.com>
10788 David Sherwood <david.sherwood@arm.com>
10790 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10791 (reduc_xor_scal_optab): New optabs.
10792 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10793 (reduc_xor_scal_@var{m}): Document.
10794 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10795 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10796 internal functions.
10797 * fold-const-call.c (fold_const_call): Handle them.
10798 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10799 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10800 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10801 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10802 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10803 (UNSPEC_XORV): New unspecs.
10804 (optab): Add entries for them.
10805 (BITWISEV): New int iterator.
10806 (bit_reduc_op): New int attributes.
10808 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10809 Alan Hayward <alan.hayward@arm.com>
10810 David Sherwood <david.sherwood@arm.com>
10812 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10813 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10814 * optabs.def (vec_shl_insert_optab): New optab.
10815 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10816 (duplicate_and_interleave): Likewise.
10817 * tree-vect-loop.c: Include internal-fn.h.
10818 (neutral_op_for_slp_reduction): New function, split out from
10819 get_initial_defs_for_reduction.
10820 (get_initial_def_for_reduction): Handle option 2 for variable-length
10821 vectors by loading the neutral value into a vector and then shifting
10822 the initial value into element 0.
10823 (get_initial_defs_for_reduction): Replace the code argument with
10824 the neutral value calculated by neutral_op_for_slp_reduction.
10825 Use gimple_build_vector for constant-length vectors.
10826 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10827 but the first group_size elements have a neutral value.
10828 Use duplicate_and_interleave otherwise.
10829 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10830 Update call to get_initial_defs_for_reduction. Handle SLP
10831 reductions for variable-length vectors by creating one vector
10832 result for each scalar result, with the elements associated
10833 with other scalar results stubbed out with the neutral value.
10834 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10835 Require IFN_VEC_SHL_INSERT for double reductions on
10836 variable-length vectors, or SLP reductions that have
10837 a neutral value. Require can_duplicate_and_interleave_p
10838 support for variable-length unchained SLP reductions if there
10839 is no neutral value, such as for MIN/MAX reductions. Also require
10840 the number of vector elements to be a multiple of the number of
10841 SLP statements when doing variable-length unchained SLP reductions.
10842 Update call to vect_create_epilog_for_reduction.
10843 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10844 and remove initial values.
10845 (duplicate_and_interleave): Make public.
10846 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10847 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10849 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10850 Alan Hayward <alan.hayward@arm.com>
10851 David Sherwood <david.sherwood@arm.com>
10853 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10854 (can_duplicate_and_interleave_p): New function.
10855 (vect_get_and_check_slp_defs): Take the vector of statements
10856 rather than just the current one. Remove excess parentheses.
10857 Restriction rejectinon of vect_constant_def and vect_external_def
10858 for variable-length vectors to boolean types, or types for which
10859 can_duplicate_and_interleave_p is false.
10860 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10861 (duplicate_and_interleave): New function.
10862 (vect_get_constant_vectors): Use gimple_build_vector for
10863 constant-length vectors and suitable variable-length constant
10864 vectors. Use duplicate_and_interleave for other variable-length
10865 vectors. Don't defer the update when inserting new statements.
10867 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10868 Alan Hayward <alan.hayward@arm.com>
10869 David Sherwood <david.sherwood@arm.com>
10871 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10872 min_profitable_iters doesn't go negative.
10874 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10875 Alan Hayward <alan.hayward@arm.com>
10876 David Sherwood <david.sherwood@arm.com>
10878 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10879 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10880 * optabs.def (vec_mask_load_lanes_optab): New optab.
10881 (vec_mask_store_lanes_optab): Likewise.
10882 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10883 (MASK_STORE_LANES): Likewise.
10884 * internal-fn.c (mask_load_lanes_direct): New macro.
10885 (mask_store_lanes_direct): Likewise.
10886 (expand_mask_load_optab_fn): Handle masked operations.
10887 (expand_mask_load_lanes_optab_fn): New macro.
10888 (expand_mask_store_optab_fn): Handle masked operations.
10889 (expand_mask_store_lanes_optab_fn): New macro.
10890 (direct_mask_load_lanes_optab_supported_p): Likewise.
10891 (direct_mask_store_lanes_optab_supported_p): Likewise.
10892 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10894 (vect_load_lanes_supported): Likewise.
10895 * tree-vect-data-refs.c (strip_conversion): New function.
10896 (can_group_stmts_p): Likewise.
10897 (vect_analyze_data_ref_accesses): Use it instead of checking
10898 for a pair of assignments.
10899 (vect_store_lanes_supported): Take a masked_p parameter.
10900 (vect_load_lanes_supported): Likewise.
10901 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10902 vect_store_lanes_supported and vect_load_lanes_supported.
10903 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10904 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10905 parameter. Don't allow gaps for masked accesses.
10906 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10907 and vect_load_lanes_supported.
10908 (get_load_store_type): Take a masked_p parameter and update
10909 call to get_group_load_store_type.
10910 (vectorizable_store): Update call to get_load_store_type.
10911 Handle IFN_MASK_STORE_LANES.
10912 (vectorizable_load): Update call to get_load_store_type.
10913 Handle IFN_MASK_LOAD_LANES.
10915 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10916 Alan Hayward <alan.hayward@arm.com>
10917 David Sherwood <david.sherwood@arm.com>
10919 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10921 * config/aarch64/aarch64-protos.h
10922 (aarch64_sve_struct_memory_operand_p): Declare.
10923 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10924 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10925 (VPRED, vpred): Handle SVE structure modes.
10926 * config/aarch64/constraints.md (Utx): New constraint.
10927 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10928 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10929 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10930 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10931 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10932 structure modes. Split into pieces after RA.
10933 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10934 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10936 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10937 SVE structure modes.
10938 (aarch64_classify_address): Likewise.
10939 (sizetochar): Move earlier in file.
10940 (aarch64_print_operand): Handle SVE register lists.
10941 (aarch64_array_mode): New function.
10942 (aarch64_sve_struct_memory_operand_p): Likewise.
10943 (TARGET_ARRAY_MODE): Redefine.
10945 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10946 Alan Hayward <alan.hayward@arm.com>
10947 David Sherwood <david.sherwood@arm.com>
10949 * target.def (array_mode): New target hook.
10950 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10951 * doc/tm.texi: Regenerate.
10952 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10953 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10954 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10955 targetm.array_mode.
10956 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10959 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10960 Alan Hayward <alan.hayward@arm.com>
10961 David Sherwood <david.sherwood@arm.com>
10963 * fold-const.c (fold_binary_loc): Check the argument types
10964 rather than the result type when testing for a vector operation.
10966 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10968 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10969 * doc/tm.texi: Regenerate.
10971 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10972 Alan Hayward <alan.hayward@arm.com>
10973 David Sherwood <david.sherwood@arm.com>
10975 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10976 (sve): Document new AArch64 extension.
10977 * doc/md.texi (w): Extend the description of the AArch64
10978 constraint to include SVE vectors.
10979 (Upl, Upa): Document new AArch64 predicate constraints.
10980 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10982 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10983 (msve-vector-bits=): New option.
10984 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10985 SVE when these are disabled.
10986 (sve): New extension.
10987 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10988 modes. Adjust their number of units based on aarch64_sve_vg.
10989 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10990 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10991 aarch64_addr_query_type.
10992 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10993 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10994 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10995 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10996 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10997 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10998 (aarch64_simd_imm_zero_p): Delete.
10999 (aarch64_check_zero_based_sve_index_immediate): Declare.
11000 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11001 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11002 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11003 (aarch64_sve_float_mul_immediate_p): Likewise.
11004 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11005 rather than an rtx.
11006 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11007 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11008 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11009 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11010 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11011 (aarch64_regmode_natural_size): Likewise.
11012 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11013 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11015 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11016 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11017 for VG and the SVE predicate registers.
11018 (V_ALIASES): Add a "z"-prefixed alias.
11019 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11020 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11021 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11022 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11023 (REG_CLASS_NAMES): Add entries for them.
11024 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11025 and the predicate registers.
11026 (aarch64_sve_vg): Declare.
11027 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11028 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11029 (REGMODE_NATURAL_SIZE): Define.
11030 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11032 * config/aarch64/aarch64.c: Include cfgrtl.h.
11033 (simd_immediate_info): Add a constructor for series vectors,
11034 and an associated step field.
11035 (aarch64_sve_vg): New variable.
11036 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11037 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11038 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11039 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11040 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11041 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11042 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11043 (aarch64_get_mask_mode): New functions.
11044 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11045 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11046 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11047 predicate modes and predicate registers. Explicitly restrict
11048 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11049 to store a vector mode if it is recognized by
11050 aarch64_classify_vector_mode.
11051 (aarch64_regmode_natural_size): New function.
11052 (aarch64_hard_regno_caller_save_mode): Return the original mode
11054 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11055 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11056 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11057 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11059 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11060 does not overlap dest if the function is frame-related. Handle
11062 (aarch64_split_add_offset): New function.
11063 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11064 them aarch64_add_offset.
11065 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11066 and update call to aarch64_sub_sp.
11067 (aarch64_add_cfa_expression): New function.
11068 (aarch64_expand_prologue): Pass extra temporary registers to the
11069 functions above. Handle the case in which we need to emit new
11070 DW_CFA_expressions for registers that were originally saved
11071 relative to the stack pointer, but now have to be expressed
11072 relative to the frame pointer.
11073 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11075 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11076 IP0 and IP1 values for SVE frames.
11077 (aarch64_expand_vec_series): New function.
11078 (aarch64_expand_sve_widened_duplicate): Likewise.
11079 (aarch64_expand_sve_const_vector): Likewise.
11080 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11081 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11082 into the register, rather than emitting a SET directly.
11083 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11084 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11085 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11086 (offset_9bit_signed_scaled_p): New functions.
11087 (aarch64_replicate_bitmask_imm): New function.
11088 (aarch64_bitmask_imm): Use it.
11089 (aarch64_cannot_force_const_mem): Reject expressions involving
11090 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11091 (aarch64_classify_index): Handle SVE indices, by requiring
11092 a plain register index with a scale that matches the element size.
11093 (aarch64_classify_address): Handle SVE addresses. Assert that
11094 the mode of the address is VOIDmode or an integer mode.
11095 Update call to aarch64_classify_symbol.
11096 (aarch64_classify_symbolic_expression): Update call to
11097 aarch64_classify_symbol.
11098 (aarch64_const_vec_all_in_range_p): New function.
11099 (aarch64_print_vector_float_operand): Likewise.
11100 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11101 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11102 and the FP immediates 1.0 and 0.5.
11103 (aarch64_print_address_internal): Handle SVE addresses.
11104 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11105 (aarch64_regno_regclass): Handle predicate registers.
11106 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11108 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11109 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11110 (aarch64_convert_sve_vector_bits): New function.
11111 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11112 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11113 rather than an rtx.
11114 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11115 Handle SVE vector and predicate modes. Accept VL-based constants
11116 that need only one temporary register, and VL offsets that require
11117 no temporary registers.
11118 (aarch64_conditional_register_usage): Mark the predicate registers
11119 as fixed if SVE isn't available.
11120 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11121 Return true for SVE vector and predicate modes.
11122 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11123 rather than an unsigned int. Handle SVE modes.
11124 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11126 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11128 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11129 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11130 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11131 (aarch64_sve_float_mul_immediate_p): New functions.
11132 (aarch64_sve_valid_immediate): New function.
11133 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11134 Explicitly reject structure modes. Check for INDEX constants.
11135 Handle PTRUE and PFALSE constants.
11136 (aarch64_check_zero_based_sve_index_immediate): New function.
11137 (aarch64_simd_imm_zero_p): Delete.
11138 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11139 vector modes. Accept constants in the range of CNT[BHWD].
11140 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11141 ask for an Advanced SIMD mode.
11142 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11143 (aarch64_simd_vector_alignment): Handle SVE predicates.
11144 (aarch64_vectorize_preferred_vector_alignment): New function.
11145 (aarch64_simd_vector_alignment_reachable): Use it instead of
11147 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11148 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11150 (MAX_VECT_LEN): Delete.
11151 (expand_vec_perm_d): Add a vec_flags field.
11152 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11153 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11154 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11156 (aarch64_evpc_rev): Rename to...
11157 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11158 (aarch64_evpc_rev_global): New function.
11159 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11160 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11162 (aarch64_evpc_sve_tbl): New function.
11163 (aarch64_expand_vec_perm_const_1): Update after rename of
11164 aarch64_evpc_rev. Handle SVE permutes too, trying
11165 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11166 than aarch64_evpc_tbl.
11167 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11168 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11169 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11170 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11171 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11172 (aarch64_expand_sve_vcond): New functions.
11173 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11174 of aarch64_vector_mode_p.
11175 (aarch64_dwarf_poly_indeterminate_value): New function.
11176 (aarch64_compute_pressure_classes): Likewise.
11177 (aarch64_can_change_mode_class): Likewise.
11178 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11179 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11180 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11181 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11182 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11183 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11184 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11185 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11187 (Dn, Dl, Dr): Accept const as well as const_vector.
11188 (Dz): Likewise. Compare against CONST0_RTX.
11189 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11190 of "vector" where appropriate.
11191 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11192 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11193 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11194 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11195 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11196 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11197 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11198 (v_int_equiv): Extend to SVE modes.
11199 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11201 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11202 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11203 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11204 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11205 (SVE_COND_FP_CMP): New int iterators.
11206 (perm_hilo): Handle the new unpack unspecs.
11207 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11209 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11210 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11211 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11212 (aarch64_equality_operator, aarch64_constant_vector_operand)
11213 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11214 (aarch64_sve_nonimmediate_operand): Likewise.
11215 (aarch64_sve_general_operand): Likewise.
11216 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11217 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11218 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11219 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11220 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11221 (aarch64_sve_float_arith_immediate): Likewise.
11222 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11223 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11224 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11225 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11226 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11227 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11228 (aarch64_sve_float_arith_operand): Likewise.
11229 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11230 (aarch64_sve_float_mul_operand): Likewise.
11231 (aarch64_sve_vec_perm_operand): Likewise.
11232 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11233 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11234 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11235 as well as const_vector.
11236 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11237 in file. Use CONST0_RTX and CONSTM1_RTX.
11238 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11239 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11240 Use aarch64_simd_imm_zero.
11241 * config/aarch64/aarch64-sve.md: New file.
11242 * config/aarch64/aarch64.md: Include it.
11243 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11244 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11245 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11246 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11247 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11248 (sve): New attribute.
11249 (enabled): Disable instructions with the sve attribute unless
11251 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11252 aarch64_expand_mov_immediate.
11253 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11254 CNT[BHSD] immediates.
11255 (movti): Split CONST_POLY_INT moves into two halves.
11256 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11257 Split additions that need a temporary here if the destination
11258 is the stack pointer.
11259 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11260 (*add<mode>3_poly_1): New instruction.
11261 (set_clobber_cc): New expander.
11263 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11265 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11266 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11267 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11268 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11269 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11270 Change innermode from fixed_mode_size to machine_mode.
11271 (simplify_subreg): Update call accordingly. Handle a constant-sized
11272 subreg of a variable-length CONST_VECTOR.
11274 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11275 Alan Hayward <alan.hayward@arm.com>
11276 David Sherwood <david.sherwood@arm.com>
11278 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11279 (add_offset_to_base): New function, split out from...
11280 (create_mem_ref): ...here. When handling a scale other than 1,
11281 check first whether the address is valid without the offset.
11282 Add it into the base if so, leaving the index and scale as-is.
11284 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11287 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11288 fold_for_warn before checking if arg2 is INTEGER_CST.
11290 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11292 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11293 (store_multiple_operation): Delete.
11294 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11295 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11296 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11297 guarded by TARGET_STRING.
11298 (rs6000_output_load_multiple): Delete.
11299 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11300 OPTION_MASK_STRING / TARGET_STRING handling.
11301 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11302 (const rs6000_opt_masks) <"string">: Change mask to 0.
11303 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11304 (MASK_STRING): Delete.
11305 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11307 (load_multiple): Delete.
11314 (store_multiple): Delete.
11321 (movmemsi_8reg): Delete.
11322 (corresponding unnamed define_insn): Delete.
11323 (movmemsi_6reg): Delete.
11324 (corresponding unnamed define_insn): Delete.
11325 (movmemsi_4reg): Delete.
11326 (corresponding unnamed define_insn): Delete.
11327 (movmemsi_2reg): Delete.
11328 (corresponding unnamed define_insn): Delete.
11329 (movmemsi_1reg): Delete.
11330 (corresponding unnamed define_insn): Delete.
11331 * config/rs6000/rs6000.opt (mno-string): New.
11332 (mstring): Replace by deprecation warning stub.
11333 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11335 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11337 * regrename.c (regrename_do_replace): If replacing the same
11338 reg multiple times, try to reuse last created gen_raw_REG.
11341 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11342 main to workaround a bug in GDB.
11344 2018-01-12 Tom de Vries <tom@codesourcery.com>
11347 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11349 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11351 PR rtl-optimization/80481
11352 * ira-color.c (get_cap_member): New function.
11353 (allocnos_conflict_by_live_ranges_p): Use it.
11354 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11355 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11357 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11360 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11361 (*saddl_se_1): Ditto.
11362 (*ssubsi_1): Ditto.
11363 (*ssubl_se_1): Ditto.
11365 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11367 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11368 rather than wi::to_widest for DR_INITs.
11369 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11370 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11371 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11373 (vect_analyze_group_access_1): Note that here.
11375 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11377 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11378 polynomial type sizes.
11380 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11382 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11383 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11384 (gimple_add_tmp_var): Likewise.
11386 2018-01-12 Martin Liska <mliska@suse.cz>
11388 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11389 (gimple_alloc_sizes): Likewise.
11390 (dump_gimple_statistics): Use PRIu64 in printf format.
11391 * gimple.h: Change uint64_t to int.
11393 2018-01-12 Martin Liska <mliska@suse.cz>
11395 * tree-core.h: Use uint64_t instead of int.
11396 * tree.c (tree_node_counts): Likewise.
11397 (tree_node_sizes): Likewise.
11398 (dump_tree_statistics): Use PRIu64 in printf format.
11400 2018-01-12 Martin Liska <mliska@suse.cz>
11402 * Makefile.in: As qsort_chk is implemented in vec.c, add
11403 vec.o to linkage of gencfn-macros.
11404 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11405 passing the info to record_node_allocation_statistics.
11406 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11408 * ggc-common.c (struct ggc_usage): Add operator== and use
11409 it in operator< and compare function.
11410 * mem-stats.h (struct mem_usage): Likewise.
11411 * vec.c (struct vec_usage): Remove operator< and compare
11412 function. Can be simply inherited.
11414 2018-01-12 Martin Jambor <mjambor@suse.cz>
11417 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11418 * tree-ssa-math-opts.c: Include domwalk.h.
11419 (convert_mult_to_fma_1): New function.
11420 (fma_transformation_info): New type.
11421 (fma_deferring_state): Likewise.
11422 (cancel_fma_deferring): New function.
11423 (result_of_phi): Likewise.
11424 (last_fma_candidate_feeds_initial_phi): Likewise.
11425 (convert_mult_to_fma): Added deferring logic, split actual
11426 transformation to convert_mult_to_fma_1.
11427 (math_opts_dom_walker): New type.
11428 (math_opts_dom_walker::after_dom_children): New method, body moved
11429 here from pass_optimize_widening_mul::execute, added deferring logic
11431 (pass_optimize_widening_mul::execute): Moved most of code to
11432 math_opts_dom_walker::after_dom_children.
11433 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11434 * config/i386/i386.c (ix86_option_override_internal): Added
11435 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11437 2018-01-12 Richard Biener <rguenther@suse.de>
11440 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11441 inline instance vars.
11443 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11446 * config/rx/rx.c (rx_is_restricted_memory_address):
11447 Handle SUBREG case.
11449 2018-01-12 Richard Biener <rguenther@suse.de>
11451 PR tree-optimization/80846
11452 * target.def (split_reduction): New target hook.
11453 * targhooks.c (default_split_reduction): New function.
11454 * targhooks.h (default_split_reduction): Declare.
11455 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11456 target requests first reduce vectors by combining low and high
11458 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11459 (get_vectype_for_scalar_type_and_size): Export.
11460 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11461 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11462 * doc/tm.texi: Regenerate.
11463 * config/i386/i386.c (ix86_split_reduction): Implement
11464 TARGET_VECTORIZE_SPLIT_REDUCTION.
11466 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11469 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11470 in PIC mode except for TARGET_VXWORKS_RTP.
11471 * config/sparc/sparc.c: Include cfgrtl.h.
11472 (TARGET_INIT_PIC_REG): Define.
11473 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11474 (sparc_pic_register_p): New predicate.
11475 (sparc_legitimate_address_p): Use it.
11476 (sparc_legitimize_pic_address): Likewise.
11477 (sparc_delegitimize_address): Likewise.
11478 (sparc_mode_dependent_address_p): Likewise.
11479 (gen_load_pcrel_sym): Remove 4th parameter.
11480 (load_got_register): Adjust call to above. Remove obsolete stuff.
11481 (sparc_expand_prologue): Do not call load_got_register here.
11482 (sparc_flat_expand_prologue): Likewise.
11483 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11484 (sparc_use_pseudo_pic_reg): New function.
11485 (sparc_init_pic_reg): Likewise.
11486 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11487 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11489 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11491 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11492 Add item for branch_cost.
11494 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11496 PR rtl-optimization/83565
11497 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11498 not extend the result to a larger mode for rotate operations.
11499 (num_sign_bit_copies1): Likewise.
11501 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11504 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11506 Use values-Xc.o for -pedantic.
11507 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11509 2018-01-12 Martin Liska <mliska@suse.cz>
11512 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11514 (possible_polymorphic_call_targets): Use it.
11515 (ipa_devirt): Likewise.
11517 2018-01-12 Martin Liska <mliska@suse.cz>
11519 * profile-count.h (enum profile_quality): Use 0 as invalid
11520 enum value of profile_quality.
11522 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11524 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11525 -mext-string options.
11527 2018-01-12 Richard Biener <rguenther@suse.de>
11529 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11530 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11531 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11533 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11535 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11537 * configure.ac (--with-long-double-format): Add support for the
11538 configuration option to change the default long double format on
11540 * config.gcc (powerpc*-linux*-*): Likewise.
11541 * configure: Regenerate.
11542 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11543 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11544 used without modification.
11546 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11548 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11549 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11550 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11551 MISC_BUILTIN_SPEC_BARRIER.
11552 (rs6000_init_builtins): Likewise.
11553 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11555 (speculation_barrier): New define_insn.
11556 * doc/extend.texi: Document __builtin_speculation_barrier.
11558 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11561 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11562 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11563 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11565 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11566 integral modes instead of "ss" and "sd".
11567 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11568 vectors with 32-bit and 64-bit elements.
11569 (vecdupssescalarmodesuffix): New mode attribute.
11570 (vec_dup<mode>): Use it.
11572 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11575 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11576 frame if argument is passed on stack.
11578 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11581 * ree.c (combine_reaching_defs): Optimize also
11582 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11583 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11585 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11587 PR middle-end/83189
11588 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11590 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11592 PR middle-end/83718
11593 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11594 after they are computed.
11596 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11598 PR tree-optimization/83695
11599 * gimple-loop-linterchange.cc
11600 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11601 reset cached scev information after interchange.
11602 (pass_linterchange::execute): Remove call to scev_reset_htab.
11604 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11606 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11607 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11608 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11609 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11610 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11611 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11612 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11613 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11614 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11615 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11616 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11617 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11618 (V_lane_reg): Likewise.
11619 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11621 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11622 (vfmal_lane_low<mode>_intrinsic,
11623 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11624 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11625 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11626 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11627 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11628 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11630 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11632 * config/arm/arm-cpus.in (fp16fml): New feature.
11633 (ALL_SIMD): Add fp16fml.
11634 (armv8.2-a): Add fp16fml as an option.
11635 (armv8.3-a): Likewise.
11636 (armv8.4-a): Add fp16fml as part of fp16.
11637 * config/arm/arm.h (TARGET_FP16FML): Define.
11638 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11640 * config/arm/arm-modes.def (V2HF): Define.
11641 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11642 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11643 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11644 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11645 vfmsl_low, vfmsl_high): New set of builtins.
11646 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11647 (vfml_op): New code attribute.
11648 (VFMLHALVES): New int iterator.
11649 (VFML, VFMLSEL): New mode attributes.
11650 (V_reg): Define mapping for V2HF.
11651 (V_hi, V_lo): New mode attributes.
11652 (VF_constraint): Likewise.
11653 (vfml_half, vfml_half_selector): New int attributes.
11654 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11656 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11657 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11659 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11660 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11661 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11662 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11664 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11665 Document new effective target and option set.
11667 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11669 * config/arm/arm-cpus.in (armv8_4): New feature.
11670 (ARMv8_4a): New fgroup.
11671 (armv8.4-a): New arch.
11672 * config/arm/arm-tables.opt: Regenerate.
11673 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11674 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11675 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11676 Add matching rules for -march=armv8.4-a and extensions.
11677 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11679 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11682 * config/rx/rx.md (BW): New mode attribute.
11683 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11685 2018-01-11 Richard Biener <rguenther@suse.de>
11687 PR tree-optimization/83435
11688 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11689 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11690 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11692 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11693 Alan Hayward <alan.hayward@arm.com>
11694 David Sherwood <david.sherwood@arm.com>
11696 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11698 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11699 (aarch64_print_address_internal): Use it to check for a zero offset.
11701 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11702 Alan Hayward <alan.hayward@arm.com>
11703 David Sherwood <david.sherwood@arm.com>
11705 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11706 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11707 Return a poly_int64 rather than a HOST_WIDE_INT.
11708 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11709 rather than a HOST_WIDE_INT.
11710 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11711 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11712 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11713 final_offset from HOST_WIDE_INT to poly_int64.
11714 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11715 to_constant when getting the number of units in an Advanced SIMD
11717 (aarch64_builtin_vectorized_function): Check for a constant number
11719 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11721 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11722 attribute instead of GET_MODE_NUNITS.
11723 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11724 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11725 GET_MODE_SIZE for fixed-size registers.
11726 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11727 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11728 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11729 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11730 (aarch64_print_operand, aarch64_print_address_internal)
11731 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11732 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11733 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11734 Handle polynomial GET_MODE_SIZE.
11735 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11736 wider than SImode without modification.
11737 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11738 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11739 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11740 passing and returning SVE modes.
11741 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11742 rather than GEN_INT.
11743 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11744 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11745 (aarch64_allocate_and_probe_stack_space): Likewise.
11746 (aarch64_layout_frame): Cope with polynomial offsets.
11747 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11748 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11749 polynomial offsets.
11750 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11751 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11752 poly_int64 rather than a HOST_WIDE_INT.
11753 (aarch64_get_separate_components, aarch64_process_components)
11754 (aarch64_expand_prologue, aarch64_expand_epilogue)
11755 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11756 (aarch64_anchor_offset): New function, split out from...
11757 (aarch64_legitimize_address): ...here.
11758 (aarch64_builtin_vectorization_cost): Handle polynomial
11759 TYPE_VECTOR_SUBPARTS.
11760 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11762 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11763 number of elements from the PARALLEL rather than the mode.
11764 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11765 rather than GET_MODE_BITSIZE.
11766 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11767 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11768 (aarch64_expand_vec_perm_const_1): Handle polynomial
11769 d->perm.length () and d->perm elements.
11770 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11771 Apply to_constant to d->perm elements.
11772 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11773 polynomial CONST_VECTOR_NUNITS.
11774 (aarch64_move_pointer): Take amount as a poly_int64 rather
11776 (aarch64_progress_pointer): Avoid temporary variable.
11777 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11778 the mode attribute instead of GET_MODE.
11780 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11781 Alan Hayward <alan.hayward@arm.com>
11782 David Sherwood <david.sherwood@arm.com>
11784 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11785 x exists before using it.
11786 (aarch64_add_constant_internal): Rename to...
11787 (aarch64_add_offset_1): ...this. Replace regnum with separate
11788 src and dest rtxes. Handle the case in which they're different,
11789 including when the offset is zero. Replace scratchreg with an rtx.
11790 Use 2 additions if there is no spare register into which we can
11791 move a 16-bit constant.
11792 (aarch64_add_constant): Delete.
11793 (aarch64_add_offset): Replace reg with separate src and dest
11794 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11795 Use aarch64_add_offset_1.
11796 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11797 an rtx rather than an int. Take the delta as a poly_int64
11798 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11799 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11800 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11801 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11802 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11803 and aarch64_add_sp.
11804 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11805 aarch64_add_constant.
11807 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11809 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11810 Use scalar_float_mode.
11812 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11814 * config/aarch64/aarch64-simd.md
11815 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11816 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11817 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11818 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11819 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11820 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11821 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11822 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11823 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11824 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11826 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11829 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11830 targ_options->x_arm_arch_string is non NULL.
11832 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11834 * config/aarch64/aarch64.h
11835 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11837 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11840 * expmed.c (emit_store_flag_force): Swap if const op0
11841 and change VOIDmode to mode of op0.
11843 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11845 PR rtl-optimization/83761
11846 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11847 than bytes to mode_for_size.
11849 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11851 PR middle-end/83189
11852 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11853 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11856 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11858 PR middle-end/83575
11859 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11860 when in layout mode.
11861 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11862 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11865 2018-01-10 Michael Collison <michael.collison@arm.com>
11867 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11868 * config/aarch64/aarch64-option-extension.def: Add
11869 AARCH64_OPT_EXTENSION of 'fp16fml'.
11870 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11871 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11872 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11873 * config/aarch64/constraints.md (Ui7): New constraint.
11874 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11875 (VFMLA_SEL_W): Ditto.
11878 (VFMLA16_LOW): New int iterator.
11879 (VFMLA16_HIGH): Ditto.
11880 (UNSPEC_FMLAL): New unspec.
11881 (UNSPEC_FMLSL): Ditto.
11882 (UNSPEC_FMLAL2): Ditto.
11883 (UNSPEC_FMLSL2): Ditto.
11884 (f16mac): New code attribute.
11885 * config/aarch64/aarch64-simd-builtins.def
11886 (aarch64_fmlal_lowv2sf): Ditto.
11887 (aarch64_fmlsl_lowv2sf): Ditto.
11888 (aarch64_fmlalq_lowv4sf): Ditto.
11889 (aarch64_fmlslq_lowv4sf): Ditto.
11890 (aarch64_fmlal_highv2sf): Ditto.
11891 (aarch64_fmlsl_highv2sf): Ditto.
11892 (aarch64_fmlalq_highv4sf): Ditto.
11893 (aarch64_fmlslq_highv4sf): Ditto.
11894 (aarch64_fmlal_lane_lowv2sf): Ditto.
11895 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11896 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11897 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11898 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11899 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11900 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11901 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11902 (aarch64_fmlal_lane_highv2sf): Ditto.
11903 (aarch64_fmlsl_lane_highv2sf): Ditto.
11904 (aarch64_fmlal_laneq_highv2sf): Ditto.
11905 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11906 (aarch64_fmlalq_lane_highv4sf): Ditto.
11907 (aarch64_fmlsl_lane_highv4sf): Ditto.
11908 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11909 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11910 * config/aarch64/aarch64-simd.md:
11911 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11912 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11913 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11914 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11915 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11916 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11917 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11918 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11919 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11920 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11921 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11922 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11923 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11924 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11925 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11926 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11927 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11928 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11929 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11930 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11931 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11932 (vfmlsl_low_u32): Ditto.
11933 (vfmlalq_low_u32): Ditto.
11934 (vfmlslq_low_u32): Ditto.
11935 (vfmlal_high_u32): Ditto.
11936 (vfmlsl_high_u32): Ditto.
11937 (vfmlalq_high_u32): Ditto.
11938 (vfmlslq_high_u32): Ditto.
11939 (vfmlal_lane_low_u32): Ditto.
11940 (vfmlsl_lane_low_u32): Ditto.
11941 (vfmlal_laneq_low_u32): Ditto.
11942 (vfmlsl_laneq_low_u32): Ditto.
11943 (vfmlalq_lane_low_u32): Ditto.
11944 (vfmlslq_lane_low_u32): Ditto.
11945 (vfmlalq_laneq_low_u32): Ditto.
11946 (vfmlslq_laneq_low_u32): Ditto.
11947 (vfmlal_lane_high_u32): Ditto.
11948 (vfmlsl_lane_high_u32): Ditto.
11949 (vfmlal_laneq_high_u32): Ditto.
11950 (vfmlsl_laneq_high_u32): Ditto.
11951 (vfmlalq_lane_high_u32): Ditto.
11952 (vfmlslq_lane_high_u32): Ditto.
11953 (vfmlalq_laneq_high_u32): Ditto.
11954 (vfmlslq_laneq_high_u32): Ditto.
11955 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11956 (AARCH64_FL_FOR_ARCH8_4): New.
11957 (AARCH64_ISA_F16FML): New ISA flag.
11958 (TARGET_F16FML): New feature flag for fp16fml.
11959 (doc/invoke.texi): Document new fp16fml option.
11961 2018-01-10 Michael Collison <michael.collison@arm.com>
11963 * config/aarch64/aarch64-builtins.c:
11964 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11965 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11966 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11967 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11968 (AARCH64_ISA_SHA3): New ISA flag.
11969 (TARGET_SHA3): New feature flag for sha3.
11970 * config/aarch64/iterators.md (sha512_op): New int attribute.
11971 (CRYPTO_SHA512): New int iterator.
11972 (UNSPEC_SHA512H): New unspec.
11973 (UNSPEC_SHA512H2): Ditto.
11974 (UNSPEC_SHA512SU0): Ditto.
11975 (UNSPEC_SHA512SU1): Ditto.
11976 * config/aarch64/aarch64-simd-builtins.def
11977 (aarch64_crypto_sha512hqv2di): New builtin.
11978 (aarch64_crypto_sha512h2qv2di): Ditto.
11979 (aarch64_crypto_sha512su0qv2di): Ditto.
11980 (aarch64_crypto_sha512su1qv2di): Ditto.
11981 (aarch64_eor3qv8hi): Ditto.
11982 (aarch64_rax1qv2di): Ditto.
11983 (aarch64_xarqv2di): Ditto.
11984 (aarch64_bcaxqv8hi): Ditto.
11985 * config/aarch64/aarch64-simd.md:
11986 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11987 (aarch64_crypto_sha512su0qv2di): Ditto.
11988 (aarch64_crypto_sha512su1qv2di): Ditto.
11989 (aarch64_eor3qv8hi): Ditto.
11990 (aarch64_rax1qv2di): Ditto.
11991 (aarch64_xarqv2di): Ditto.
11992 (aarch64_bcaxqv8hi): Ditto.
11993 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11994 (vsha512h2q_u64): Ditto.
11995 (vsha512su0q_u64): Ditto.
11996 (vsha512su1q_u64): Ditto.
11997 (veor3q_u16): Ditto.
11998 (vrax1q_u64): Ditto.
11999 (vxarq_u64): Ditto.
12000 (vbcaxq_u16): Ditto.
12001 * config/arm/types.md (crypto_sha512): New type attribute.
12002 (crypto_sha3): Ditto.
12003 (doc/invoke.texi): Document new sha3 option.
12005 2018-01-10 Michael Collison <michael.collison@arm.com>
12007 * config/aarch64/aarch64-builtins.c:
12008 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12009 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12010 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12011 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12012 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12013 (AARCH64_ISA_SM4): New ISA flag.
12014 (TARGET_SM4): New feature flag for sm4.
12015 * config/aarch64/aarch64-simd-builtins.def
12016 (aarch64_sm3ss1qv4si): Ditto.
12017 (aarch64_sm3tt1aq4si): Ditto.
12018 (aarch64_sm3tt1bq4si): Ditto.
12019 (aarch64_sm3tt2aq4si): Ditto.
12020 (aarch64_sm3tt2bq4si): Ditto.
12021 (aarch64_sm3partw1qv4si): Ditto.
12022 (aarch64_sm3partw2qv4si): Ditto.
12023 (aarch64_sm4eqv4si): Ditto.
12024 (aarch64_sm4ekeyqv4si): Ditto.
12025 * config/aarch64/aarch64-simd.md:
12026 (aarch64_sm3ss1qv4si): Ditto.
12027 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12028 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12029 (aarch64_sm4eqv4si): Ditto.
12030 (aarch64_sm4ekeyqv4si): Ditto.
12031 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12032 (sm3part_op): Ditto.
12033 (CRYPTO_SM3TT): Ditto.
12034 (CRYPTO_SM3PART): Ditto.
12035 (UNSPEC_SM3SS1): New unspec.
12036 (UNSPEC_SM3TT1A): Ditto.
12037 (UNSPEC_SM3TT1B): Ditto.
12038 (UNSPEC_SM3TT2A): Ditto.
12039 (UNSPEC_SM3TT2B): Ditto.
12040 (UNSPEC_SM3PARTW1): Ditto.
12041 (UNSPEC_SM3PARTW2): Ditto.
12042 (UNSPEC_SM4E): Ditto.
12043 (UNSPEC_SM4EKEY): Ditto.
12044 * config/aarch64/constraints.md (Ui2): New constraint.
12045 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12046 * config/arm/types.md (crypto_sm3): New type attribute.
12047 (crypto_sm4): Ditto.
12048 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12049 (vsm3tt1aq_u32): Ditto.
12050 (vsm3tt1bq_u32): Ditto.
12051 (vsm3tt2aq_u32): Ditto.
12052 (vsm3tt2bq_u32): Ditto.
12053 (vsm3partw1q_u32): Ditto.
12054 (vsm3partw2q_u32): Ditto.
12055 (vsm4eq_u32): Ditto.
12056 (vsm4ekeyq_u32): Ditto.
12057 (doc/invoke.texi): Document new sm4 option.
12059 2018-01-10 Michael Collison <michael.collison@arm.com>
12061 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12062 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12063 (AARCH64_FL_FOR_ARCH8_4): New.
12064 (AARCH64_FL_V8_4): New flag.
12065 (doc/invoke.texi): Document new armv8.4-a option.
12067 2018-01-10 Michael Collison <michael.collison@arm.com>
12069 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12070 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12071 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12072 * config/aarch64/aarch64-option-extension.def: Add
12073 AARCH64_OPT_EXTENSION of 'sha2'.
12074 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12075 (crypto): Disable sha2 and aes if crypto disabled.
12076 (crypto): Enable aes and sha2 if enabled.
12077 (simd): Disable sha2 and aes if simd disabled.
12078 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12080 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12081 (TARGET_SHA2): New feature flag for sha2.
12082 (TARGET_AES): New feature flag for aes.
12083 * config/aarch64/aarch64-simd.md:
12084 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12085 conditional on TARGET_AES.
12086 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12087 (aarch64_crypto_sha1hsi): Make pattern conditional
12089 (aarch64_crypto_sha1hv4si): Ditto.
12090 (aarch64_be_crypto_sha1hv4si): Ditto.
12091 (aarch64_crypto_sha1su1v4si): Ditto.
12092 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12093 (aarch64_crypto_sha1su0v4si): Ditto.
12094 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12095 (aarch64_crypto_sha256su0v4si): Ditto.
12096 (aarch64_crypto_sha256su1v4si): Ditto.
12097 (doc/invoke.texi): Document new aes and sha2 options.
12099 2018-01-10 Martin Sebor <msebor@redhat.com>
12101 PR tree-optimization/83781
12102 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12105 2018-01-11 Martin Sebor <msebor@gmail.com>
12106 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12108 PR tree-optimization/83501
12109 PR tree-optimization/81703
12111 * tree-ssa-strlen.c (get_string_cst): Rename...
12112 (get_string_len): ...to this. Handle global constants.
12113 (handle_char_store): Adjust.
12115 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12116 Jim Wilson <jimw@sifive.com>
12118 * config/riscv/riscv-protos.h (riscv_output_return): New.
12119 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12120 (riscv_attribute_table, riscv_output_return),
12121 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12122 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12123 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12125 (riscv_expand_prologue): Add early return for naked function.
12126 (riscv_expand_epilogue): Likewise.
12127 (riscv_function_ok_for_sibcall): Return false for naked function.
12128 (riscv_set_current_function): New.
12129 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12130 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12131 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12132 * doc/extend.texi (RISC-V Function Attributes): New.
12134 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12136 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12137 check for 128-bit long double before checking TCmode.
12138 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12139 128-bit long doubles before checking TFmode or TCmode.
12140 (FLOAT128_IBM_P): Likewise.
12142 2018-01-10 Martin Sebor <msebor@redhat.com>
12144 PR tree-optimization/83671
12145 * builtins.c (c_strlen): Unconditionally return zero for the empty
12147 Use -Warray-bounds for warnings.
12148 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12149 for non-constant array indices with COMPONENT_REF, arrays of
12150 arrays, and pointers to arrays.
12151 (gimple_fold_builtin_strlen): Determine and set length range for
12152 non-constant character arrays.
12154 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12156 PR middle-end/81897
12157 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12160 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12162 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12164 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12167 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12168 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12169 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12170 indexed_or_indirect_operand predicate.
12171 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12172 (*vsx_le_perm_load_v8hi): Likewise.
12173 (*vsx_le_perm_load_v16qi): Likewise.
12174 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12175 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12176 (*vsx_le_perm_store_v8hi): Likewise.
12177 (*vsx_le_perm_store_v16qi): Likewise.
12178 (eight unnamed splitters): Likewise.
12180 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12182 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12183 * config/rs6000/emmintrin.h: Likewise.
12184 * config/rs6000/mmintrin.h: Likewise.
12185 * config/rs6000/xmmintrin.h: Likewise.
12187 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12190 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12192 * tree.c (tree_nop_conversion): Return true for location wrapper
12194 (maybe_wrap_with_location): New function.
12195 (selftest::check_strip_nops): New function.
12196 (selftest::test_location_wrappers): New function.
12197 (selftest::tree_c_tests): Call it.
12198 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12199 (maybe_wrap_with_location): New decl.
12200 (EXPR_LOCATION_WRAPPER_P): New macro.
12201 (location_wrapper_p): New inline function.
12202 (tree_strip_any_location_wrapper): New inline function.
12204 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12207 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12208 stack_realign_offset for the largest alignment of stack slot
12210 (ix86_find_max_used_stack_alignment): New function.
12211 (ix86_finalize_stack_frame_flags): Use it. Set
12212 max_used_stack_alignment if we don't realign stack.
12213 * config/i386/i386.h (machine_function): Add
12214 max_used_stack_alignment.
12216 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12218 * config/arm/arm.opt (-mbranch-cost): New option.
12219 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12222 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12225 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12226 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12228 2018-01-10 Richard Biener <rguenther@suse.de>
12231 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12232 early out so it also covers the case where we have a non-NULL
12235 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12237 PR tree-optimization/83753
12238 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12239 for non-strided grouped accesses if the number of elements is 1.
12241 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12244 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12245 * i386.h (TARGET_USE_GATHER): Define.
12246 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12248 2018-01-10 Martin Liska <mliska@suse.cz>
12251 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12252 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12254 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12255 CLEANUP_NO_PARTITIONING is not set.
12257 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12259 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12260 for vectors, as a partial revert of r254296.
12261 * rtl.h (const_vec_p): Delete.
12262 (const_vec_duplicate_p): Don't test for vector CONSTs.
12263 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12264 * expmed.c (make_tree): Likewise.
12267 * common.md (E, F): Use CONSTANT_P instead of checking for
12269 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12270 checking for CONST_VECTOR.
12272 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12274 PR middle-end/83575
12275 * predict.c (force_edge_cold): Handle in more sane way edges
12276 with no prediction.
12278 2018-01-09 Carl Love <cel@us.ibm.com>
12280 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12282 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12283 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12284 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12285 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12286 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12287 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12288 * config/rs6000/rs6000-protos.h: Add extern defition for
12289 rs6000_generate_float2_double_code.
12290 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12292 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12293 (float2_v2df): Add define_expand.
12295 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12298 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12299 op_mode in the force_to_mode call.
12301 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12303 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12304 instead of checking each element individually.
12305 (aarch64_evpc_uzp): Likewise.
12306 (aarch64_evpc_zip): Likewise.
12307 (aarch64_evpc_ext): Likewise.
12308 (aarch64_evpc_rev): Likewise.
12309 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12310 instead of checking each element individually. Return true without
12312 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12313 whether all selected elements come from the same input, instead of
12314 checking each element individually. Remove calls to gen_rtx_REG,
12315 start_sequence and end_sequence and instead assert that no rtl is
12318 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12320 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12321 order of HIGH and CONST checks.
12323 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12325 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12326 if the destination isn't an SSA_NAME.
12328 2018-01-09 Richard Biener <rguenther@suse.de>
12330 PR tree-optimization/83668
12331 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12333 (canonicalize_loop_form): ... here, renamed from ...
12334 (canonicalize_loop_closed_ssa_form): ... this and amended to
12335 swap successor edges for loop exit blocks to make us use
12336 the RPO order we need for initial schedule generation.
12338 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12340 PR tree-optimization/64811
12341 * match.pd: When optimizing comparisons with Inf, avoid
12342 introducing or losing exceptions from comparisons with NaN.
12344 2018-01-09 Martin Liska <mliska@suse.cz>
12347 * asan.c (shadow_mem_size): Add gcc_assert.
12349 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12351 Don't save registers in main().
12354 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12355 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12356 * config/avr/avr.c (avr_set_current_function): Don't error if
12357 naked, OS_task or OS_main are specified at the same time.
12358 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12360 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12362 * common/config/avr/avr-common.c (avr_option_optimization_table):
12363 Switch on -mmain-is-OS_task for optimizing compilations.
12365 2018-01-09 Richard Biener <rguenther@suse.de>
12367 PR tree-optimization/83572
12368 * graphite.c: Include cfganal.h.
12369 (graphite_transform_loops): Connect infinite loops to exit
12370 and remove fake edges at the end.
12372 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12374 * ipa-inline.c (edge_badness): Revert accidental checkin.
12376 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12379 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12380 symbols; not inline clones.
12382 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12385 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12386 hard registers. Formatting fixes.
12388 PR preprocessor/83722
12389 * gcc.c (try_generate_repro): Pass
12390 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12391 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12394 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12395 Kito Cheng <kito.cheng@gmail.com>
12397 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12398 (riscv_leaf_function_p): Delete.
12399 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12401 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12403 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12405 (do_ifelse): New function.
12406 (do_isel): New function.
12407 (do_sub3): New function.
12408 (do_add3): New function.
12409 (do_load_mask_compare): New function.
12410 (do_overlap_load_compare): New function.
12411 (expand_compare_loop): New function.
12412 (expand_block_compare): Call expand_compare_loop() when appropriate.
12413 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12414 option description.
12415 (-mblock-compare-inline-loop-limit): New option.
12417 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12420 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12421 Reverse order of second and third operands in first alternative.
12422 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12423 of first and second elements in UNSPEC_VPERMR vector.
12424 (altivec_expand_vec_perm_le): Likewise.
12426 2018-01-08 Jeff Law <law@redhat.com>
12428 PR rtl-optimizatin/81308
12429 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12430 (process_switch): If group_case_labels makes a change, then set
12432 (pass_convert_switch::execute): If a switch is converted, then
12433 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12435 PR rtl-optimization/81308
12436 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12439 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12441 PR target/83663 - Revert r255946
12442 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12443 generation for cases where splatting a value is not useful.
12444 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12445 across a vec_duplicate and a paradoxical subreg forming a vector
12446 mode to a vec_concat.
12448 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12450 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12451 -march=armv8.3-a variants.
12452 * config/arm/t-multilib: Likewise.
12453 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12455 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12457 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12459 (cceq_ior_compare_complement): Give it a name so I can use it, and
12460 change boolean_or_operator predicate to boolean_operator so it can
12461 be used to generate a crand.
12462 (eqne): New code iterator.
12463 (bd/bd_neg): New code_attrs.
12464 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12465 a single define_insn.
12466 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12467 decrement (bdnzt/bdnzf/bdzt/bdzf).
12468 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12469 with the new names of the branch decrement patterns, and added the
12470 names of the branch decrement conditional patterns.
12472 2018-01-08 Richard Biener <rguenther@suse.de>
12474 PR tree-optimization/83563
12475 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12478 2018-01-08 Richard Biener <rguenther@suse.de>
12480 PR middle-end/83713
12481 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12483 2018-01-08 Richard Biener <rguenther@suse.de>
12485 PR tree-optimization/83685
12486 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12487 references to abnormals.
12489 2018-01-08 Richard Biener <rguenther@suse.de>
12492 * dwarf2out.c (output_indirect_strings): Handle empty
12493 skeleton_debug_str_hash.
12494 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12496 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12498 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12499 (emit_store_direct): Likewise.
12500 (arc_trampoline_adjust_address): Likewise.
12501 (arc_asm_trampoline_template): New function.
12502 (arc_initialize_trampoline): Use asm_trampoline_template.
12503 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12504 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12505 * config/arc/arc.md (flush_icache): Delete pattern.
12507 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12509 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12510 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12513 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12516 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12517 by not USED_FOR_TARGET.
12518 (make_pass_resolve_sw_modes): Likewise.
12520 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12522 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12525 2018-01-08 Richard Biener <rguenther@suse.de>
12527 PR middle-end/83580
12528 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12530 2018-01-08 Richard Biener <rguenther@suse.de>
12532 PR middle-end/83517
12533 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12535 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12537 PR middle-end/81897
12538 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12539 basic blocks with a small number of successors.
12540 (convert_control_dep_chain_into_preds): Improve handling of
12542 (dump_predicates): Split apart into...
12543 (dump_pred_chain): ...here...
12544 (dump_pred_info): ...and here.
12545 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12546 (can_chain_union_be_invalidated_p): Improve check for invalidation
12548 (uninit_uses_cannot_happen): Avoid unnecessary if
12549 convert_control_dep_chain_into_preds yielded nothing.
12551 2018-01-06 Martin Sebor <msebor@redhat.com>
12553 PR tree-optimization/83640
12554 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12555 subtracting negative offset from size.
12556 (builtin_access::overlap): Adjust offset bounds of the access to fall
12557 within the size of the object if possible.
12559 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12561 PR rtl-optimization/83699
12562 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12563 extract_bit_field_as_subreg to cases in which the extracted
12564 value is also a vector.
12566 * lra-constraints.c (process_alt_operands): Test for the equivalence
12567 substitutions when detecting a possible reload cycle.
12569 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12572 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12573 by default if flag_selective_schedling{,2}. Formatting fixes.
12575 PR rtl-optimization/83682
12576 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12577 if it has non-VECTOR_MODE element mode.
12578 (vec_duplicate_p): Likewise.
12580 PR middle-end/83694
12581 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12582 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12584 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12587 * config/i386/i386-builtin.def
12588 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12589 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12590 Require also OPTION_MASK_ISA_AVX512F in addition to
12591 OPTION_MASK_ISA_GFNI.
12592 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12593 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12594 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12595 to OPTION_MASK_ISA_GFNI.
12596 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12597 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12598 OPTION_MASK_ISA_AVX512BW.
12599 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12600 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12601 addition to OPTION_MASK_ISA_GFNI.
12602 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12603 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12604 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12605 to OPTION_MASK_ISA_GFNI.
12606 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12607 a requirement for all ISAs rather than any of them with a few
12609 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12611 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12612 bitmasks to be enabled with 3 exceptions, instead of requiring any
12613 enabled ISA with lots of exceptions.
12614 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12615 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12616 Change avx512bw in isa attribute to avx512f.
12617 * config/i386/sgxintrin.h: Add license boilerplate.
12618 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12619 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12620 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12621 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12623 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12624 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12625 temporarily sse2 rather than sse if not enabled already.
12628 * config/i386/sse.md (VI248_VLBW): Rename to ...
12629 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12630 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12631 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12632 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12633 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12634 mode iterator instead of VI248_VLBW.
12636 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12638 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12639 (record_modified): Skip clobbers; add debug output.
12640 (param_change_prob): Use sreal frequencies.
12642 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12644 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12645 punt for user-aligned variables.
12647 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12649 * tree-chrec.c (chrec_contains_symbols): Return true for
12652 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12655 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12656 of (x|y) == x for BICS pattern.
12658 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12660 PR tree-optimization/83605
12661 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12662 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12665 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12667 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12668 * config/epiphany/rtems.h: New file.
12670 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12671 Uros Bizjak <ubizjak@gmail.com>
12674 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12675 QIreg_operand instead of register_operand predicate.
12676 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12677 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12678 comments instead of -fmitigate[-_]rop.
12680 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12683 * cgraphunit.c (symbol_table::compile): Switch to text_section
12684 before calling assembly_start debug hook.
12685 * run-rtl-passes.c (run_rtl_passes): Likewise.
12688 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12690 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12691 range_int_cst_p rather than !symbolic_range_p before calling
12692 extract_range_from_multiplicative_op_1.
12694 2018-01-04 Jeff Law <law@redhat.com>
12696 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12697 redundant test in assertion.
12699 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12701 * doc/rtl.texi: Document machine_mode wrapper classes.
12703 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12705 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12706 using tree_to_uhwi.
12708 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12710 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12711 the VEC_PERM_EXPR fold to fail.
12713 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12716 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12717 to switched_sections.
12719 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12722 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12723 test for d.testing.
12725 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12728 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12729 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12731 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12734 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12735 is BLKmode and bitpos not zero or mode change is needed.
12737 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12740 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12743 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12746 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12747 instead of MULT rtx. Update all corresponding splitters.
12748 (*saddl_se): Ditto.
12749 (*ssub<modesuffix>): Ditto.
12750 (*ssubl_se): Ditto.
12751 (*cmp_sadd_di): Update split patterns.
12752 (*cmp_sadd_si): Ditto.
12753 (*cmp_sadd_sidi): Ditto.
12754 (*cmp_ssub_di): Ditto.
12755 (*cmp_ssub_si): Ditto.
12756 (*cmp_ssub_sidi): Ditto.
12757 * config/alpha/predicates.md (const23_operand): New predicate.
12758 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12759 Look for ASHIFT, not MULT inner operand.
12760 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12762 2018-01-04 Martin Liska <mliska@suse.cz>
12764 PR gcov-profile/83669
12765 * gcov.c (output_intermediate_file): Add version to intermediate
12767 * doc/gcov.texi: Document new field 'version' in intermediate
12768 file format. Fix location of '-k' option of gcov command.
12770 2018-01-04 Martin Liska <mliska@suse.cz>
12773 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12775 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12777 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12779 2018-01-03 Martin Sebor <msebor@redhat.com>
12781 PR tree-optimization/83655
12782 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12783 checking calls with invalid arguments.
12785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12787 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12788 (vectorizable_mask_load_store): Delete.
12789 (vectorizable_call): Return false for masked loads and stores.
12790 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12791 instead of gimple_assign_rhs1.
12792 (vectorizable_load): Handle IFN_MASK_LOAD.
12793 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12797 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12799 (vectorizable_mask_load_store): ...here.
12800 (vectorizable_load): ...and here.
12802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12804 * tree-vect-stmts.c (vect_build_all_ones_mask)
12805 (vect_build_zero_merge_argument): New functions, split out from...
12806 (vectorizable_load): ...here.
12808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12810 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12812 (vectorizable_mask_load_store): ...here.
12813 (vectorizable_store): ...and here.
12815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12817 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12819 (vectorizable_mask_load_store): ...here.
12821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12823 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12824 (vect_model_store_cost): Take a vec_load_store_type instead of a
12826 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12827 (vect_model_store_cost): Take a vec_load_store_type instead of a
12829 (vectorizable_mask_load_store): Update accordingly.
12830 (vectorizable_store): Likewise.
12831 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12835 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12836 IFN_MASK_LOAD calls here rather than...
12837 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12840 Alan Hayward <alan.hayward@arm.com>
12841 David Sherwood <david.sherwood@arm.com>
12843 * expmed.c (extract_bit_field_1): For vector extracts,
12844 fall back to extract_bit_field_as_subreg if vec_extract
12847 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12848 Alan Hayward <alan.hayward@arm.com>
12849 David Sherwood <david.sherwood@arm.com>
12851 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12852 they are variable or constant sized.
12853 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12854 slots for constant-sized data.
12856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12857 Alan Hayward <alan.hayward@arm.com>
12858 David Sherwood <david.sherwood@arm.com>
12860 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12861 handling COND_EXPRs with boolean comparisons, try to find a better
12862 basis for the mask type than the boolean itself.
12864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12866 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12867 is calculated and how it can be overridden.
12868 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12869 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12871 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12875 Alan Hayward <alan.hayward@arm.com>
12876 David Sherwood <david.sherwood@arm.com>
12878 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12879 Remove the mode argument.
12880 (aarch64_simd_valid_immediate): Remove the mode and inverse
12882 * config/aarch64/iterators.md (bitsize): New iterator.
12883 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12884 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12885 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12886 aarch64_simd_valid_immediate.
12887 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12888 (aarch64_reg_or_bic_imm): Likewise.
12889 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12890 with an insn_type enum and msl with a modifier_type enum.
12891 Replace element_width with a scalar_mode. Change the shift
12892 to unsigned int. Add constructors for scalar_float_mode and
12893 scalar_int_mode elements.
12894 (aarch64_vect_float_const_representable_p): Delete.
12895 (aarch64_can_const_movi_rtx_p)
12896 (aarch64_simd_scalar_immediate_valid_for_move)
12897 (aarch64_simd_make_constant): Update call to
12898 aarch64_simd_valid_immediate.
12899 (aarch64_advsimd_valid_immediate_hs): New function.
12900 (aarch64_advsimd_valid_immediate): Likewise.
12901 (aarch64_simd_valid_immediate): Remove mode and inverse
12902 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12903 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12904 and aarch64_float_const_representable_p on the result.
12905 (aarch64_output_simd_mov_immediate): Remove mode argument.
12906 Update call to aarch64_simd_valid_immediate and use of
12907 simd_immediate_info.
12908 (aarch64_output_scalar_simd_mov_immediate): Update call
12911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12912 Alan Hayward <alan.hayward@arm.com>
12913 David Sherwood <david.sherwood@arm.com>
12915 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12916 (mode_nunits): Likewise CONST_MODE_NUNITS.
12917 * machmode.def (ADJUST_NUNITS): Document.
12918 * genmodes.c (mode_data::need_nunits_adj): New field.
12919 (blank_mode): Update accordingly.
12920 (adj_nunits): New variable.
12921 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12923 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12924 listed in adj_nunits.
12925 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12926 listed in adj_nunits. Don't emit case statements for such modes.
12927 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12928 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12929 nothing if adj_nunits is nonnull.
12930 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12931 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12932 (emit_mode_fbit): Update use of print_maybe_const_decl.
12933 (emit_move_size): Likewise. Treat the array as non-const
12935 (emit_mode_adjustments): Handle adj_nunits.
12937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12939 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12940 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12941 (VECTOR_MODES): Use it.
12942 (make_vector_modes): Take the prefix as an argument.
12944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12945 Alan Hayward <alan.hayward@arm.com>
12946 David Sherwood <david.sherwood@arm.com>
12948 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12949 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12950 for MODE_VECTOR_BOOL.
12951 * machmode.def (VECTOR_BOOL_MODE): Document.
12952 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12953 (make_vector_bool_mode): New function.
12954 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12956 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12957 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12959 * stor-layout.c (int_mode_for_mode): Likewise.
12960 * tree.c (build_vector_type_for_mode): Likewise.
12961 * varasm.c (output_constant_pool_2): Likewise.
12962 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12963 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12964 for MODE_VECTOR_BOOL.
12965 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12966 of mode class checks.
12967 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12968 instead of a list of mode class checks.
12969 (expand_vector_scalar_condition): Likewise.
12970 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12973 Alan Hayward <alan.hayward@arm.com>
12974 David Sherwood <david.sherwood@arm.com>
12976 * machmode.h (mode_size): Change from unsigned short to
12978 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12979 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12980 or if measurement_type is not polynomial.
12981 (fixed_size_mode::includes_p): Check for constant-sized modes.
12982 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12983 return a poly_uint16 rather than an unsigned short.
12984 (emit_mode_size): Change the type of mode_size from unsigned short
12985 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12986 (emit_mode_adjustments): Cope with polynomial vector sizes.
12987 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12989 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12991 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12992 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12993 * caller-save.c (setup_save_areas): Likewise.
12994 (replace_reg_with_saved_mem): Likewise.
12995 * calls.c (emit_library_call_value_1): Likewise.
12996 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12997 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12998 (gen_lowpart_for_combine): Likewise.
12999 * convert.c (convert_to_integer_1): Likewise.
13000 * cse.c (equiv_constant, cse_insn): Likewise.
13001 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13002 (cselib_subst_to_values): Likewise.
13003 * dce.c (word_dce_process_block): Likewise.
13004 * df-problems.c (df_word_lr_mark_ref): Likewise.
13005 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13006 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13007 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13008 (rtl_for_decl_location): Likewise.
13009 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13010 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13011 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13012 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13013 (expand_expr_real_1): Likewise.
13014 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13015 (pad_below): Likewise.
13016 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13017 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13018 * ira.c (get_subreg_tracking_sizes): Likewise.
13019 * ira-build.c (ira_create_allocno_objects): Likewise.
13020 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13021 (ira_sort_regnos_for_alter_reg): Likewise.
13022 * ira-costs.c (record_operand_costs): Likewise.
13023 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13024 (resolve_simple_move): Likewise.
13025 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13026 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13027 (lra_constraints): Likewise.
13028 (CONST_POOL_OK_P): Reject variable-sized modes.
13029 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13030 (add_pseudo_to_slot, lra_spill): Likewise.
13031 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13032 * optabs-query.c (get_best_extraction_insn): Likewise.
13033 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13034 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13035 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13036 * recog.c (offsettable_address_addr_space_p): Likewise.
13037 * regcprop.c (maybe_mode_change): Likewise.
13038 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13039 * regrename.c (build_def_use): Likewise.
13040 * regstat.c (dump_reg_info): Likewise.
13041 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13042 (find_reloads, find_reloads_subreg_address): Likewise.
13043 * reload1.c (eliminate_regs_1): Likewise.
13044 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13045 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13046 (simplify_binary_operation_1, simplify_subreg): Likewise.
13047 * targhooks.c (default_function_arg_padding): Likewise.
13048 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13049 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13050 (verify_gimple_assign_ternary): Likewise.
13051 * tree-inline.c (estimate_move_cost): Likewise.
13052 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13053 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13054 (get_address_cost_ainc): Likewise.
13055 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13056 (vect_supportable_dr_alignment): Likewise.
13057 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13058 (vectorizable_reduction): Likewise.
13059 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13060 (vectorizable_operation, vectorizable_load): Likewise.
13061 * tree.c (build_same_sized_truth_vector_type): Likewise.
13062 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13063 * var-tracking.c (emit_note_insn_var_location): Likewise.
13064 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13065 (ADDR_VEC_ALIGN): Likewise.
13067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13068 Alan Hayward <alan.hayward@arm.com>
13069 David Sherwood <david.sherwood@arm.com>
13071 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13073 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13074 or if measurement_type is polynomial.
13075 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13076 * combine.c (make_extraction): Likewise.
13077 * dse.c (find_shift_sequence): Likewise.
13078 * dwarf2out.c (mem_loc_descriptor): Likewise.
13079 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13080 (extract_bit_field, extract_low_bits): Likewise.
13081 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13082 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13083 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13084 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13085 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13086 * reload.c (find_reloads): Likewise.
13087 * reload1.c (alter_reg): Likewise.
13088 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13089 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13090 * tree-if-conv.c (predicate_mem_writes): Likewise.
13091 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13092 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13093 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13094 * valtrack.c (dead_debug_insert_temp): Likewise.
13095 * varasm.c (mergeable_constant_section): Likewise.
13096 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13099 Alan Hayward <alan.hayward@arm.com>
13100 David Sherwood <david.sherwood@arm.com>
13102 * expr.c (expand_assignment): Cope with polynomial mode sizes
13103 when assigning to a CONCAT.
13105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13106 Alan Hayward <alan.hayward@arm.com>
13107 David Sherwood <david.sherwood@arm.com>
13109 * machmode.h (mode_precision): Change from unsigned short to
13111 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13113 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13114 or if measurement_type is not polynomial.
13115 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13116 in which the mode is already known to be a scalar_int_mode.
13117 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13118 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13120 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13121 for GET_MODE_PRECISION.
13122 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13123 for GET_MODE_PRECISION.
13124 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13126 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13127 (expand_field_assignment, make_extraction): Likewise.
13128 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13129 (get_last_value): Likewise.
13130 * convert.c (convert_to_integer_1): Likewise.
13131 * cse.c (cse_insn): Likewise.
13132 * expr.c (expand_expr_real_1): Likewise.
13133 * lra-constraints.c (simplify_operand_subreg): Likewise.
13134 * optabs-query.c (can_atomic_load_p): Likewise.
13135 * optabs.c (expand_atomic_load): Likewise.
13136 (expand_atomic_store): Likewise.
13137 * ree.c (combine_reaching_defs): Likewise.
13138 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13139 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13140 * tree.h (type_has_mode_precision_p): Likewise.
13141 * ubsan.c (instrument_si_overflow): Likewise.
13143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13144 Alan Hayward <alan.hayward@arm.com>
13145 David Sherwood <david.sherwood@arm.com>
13147 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13148 polynomial numbers of units.
13149 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13150 (valid_vector_subparts_p): New function.
13151 (build_vector_type): Remove temporary shim and take the number
13152 of units as a poly_uint64 rather than an int.
13153 (build_opaque_vector_type): Take the number of units as a
13154 poly_uint64 rather than an int.
13155 * tree.c (build_vector_from_ctor): Handle polynomial
13156 TYPE_VECTOR_SUBPARTS.
13157 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13158 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13159 (build_vector_from_val): If the number of units is variable,
13160 use build_vec_duplicate_cst for constant operands and
13161 VEC_DUPLICATE_EXPR otherwise.
13162 (make_vector_type): Remove temporary is_constant ().
13163 (build_vector_type, build_opaque_vector_type): Take the number of
13164 units as a poly_uint64 rather than an int.
13165 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13167 * cfgexpand.c (expand_debug_expr): Likewise.
13168 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13169 (store_constructor, expand_expr_real_1): Likewise.
13170 (const_scalar_mask_from_tree): Likewise.
13171 * fold-const-call.c (fold_const_reduction): Likewise.
13172 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13173 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13174 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13175 (fold_relational_const): Likewise.
13176 (native_interpret_vector): Likewise. Change the size from an
13177 int to an unsigned int.
13178 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13179 TYPE_VECTOR_SUBPARTS.
13180 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13181 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13182 duplicating a non-constant operand into a variable-length vector.
13183 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13184 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13185 * ipa-icf.c (sem_variable::equals): Likewise.
13186 * match.pd: Likewise.
13187 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13188 * print-tree.c (print_node): Likewise.
13189 * stor-layout.c (layout_type): Likewise.
13190 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13191 * tree-cfg.c (verify_gimple_comparison): Likewise.
13192 (verify_gimple_assign_binary): Likewise.
13193 (verify_gimple_assign_ternary): Likewise.
13194 (verify_gimple_assign_single): Likewise.
13195 * tree-pretty-print.c (dump_generic_node): Likewise.
13196 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13197 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13198 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13199 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13200 (vect_shift_permute_load_chain): Likewise.
13201 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13202 (expand_vector_condition, optimize_vector_constructor): Likewise.
13203 (lower_vec_perm, get_compute_type): Likewise.
13204 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13205 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13206 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13207 (vect_recog_mask_conversion_pattern): Likewise.
13208 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13209 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13210 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13211 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13212 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13213 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13214 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13215 (supportable_widening_operation): Likewise.
13216 (supportable_narrowing_operation): Likewise.
13217 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13219 * varasm.c (output_constant): Likewise.
13221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13222 Alan Hayward <alan.hayward@arm.com>
13223 David Sherwood <david.sherwood@arm.com>
13225 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13226 so that both the length == 3 and length != 3 cases set up their
13227 own permute vectors. Add comments explaining why we know the
13228 number of elements is constant.
13229 (vect_permute_load_chain): Likewise.
13231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13232 Alan Hayward <alan.hayward@arm.com>
13233 David Sherwood <david.sherwood@arm.com>
13235 * machmode.h (mode_nunits): Change from unsigned char to
13237 (ONLY_FIXED_SIZE_MODES): New macro.
13238 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13239 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13240 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13242 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13243 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13244 or if measurement_type is not polynomial.
13245 * genmodes.c (ZERO_COEFFS): New macro.
13246 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13248 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13249 Use ZERO_COEFFS when emitting initializers.
13250 * data-streamer.h (bp_pack_poly_value): New function.
13251 (bp_unpack_poly_value): Likewise.
13252 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13253 for GET_MODE_NUNITS.
13254 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13255 for GET_MODE_NUNITS.
13256 * tree.c (make_vector_type): Remove temporary shim and make
13257 the real function take the number of units as a poly_uint64
13258 rather than an int.
13259 (build_vector_type_for_mode): Handle polynomial nunits.
13260 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13261 * emit-rtl.c (const_vec_series_p_1): Likewise.
13262 (gen_rtx_CONST_VECTOR): Likewise.
13263 * fold-const.c (test_vec_duplicate_folding): Likewise.
13264 * genrecog.c (validate_pattern): Likewise.
13265 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13266 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13267 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13268 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13269 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13270 * rtlanal.c (subreg_get_info): Likewise.
13271 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13272 (vect_grouped_load_supported): Likewise.
13273 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13274 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13275 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13276 (simplify_const_unary_operation, simplify_binary_operation_1)
13277 (simplify_const_binary_operation, simplify_ternary_operation)
13278 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13279 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13280 instead of CONST_VECTOR_NUNITS.
13281 * varasm.c (output_constant_pool_2): Likewise.
13282 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13283 explicit-encoded elements in the XVEC for variable-length vectors.
13285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13287 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13290 Alan Hayward <alan.hayward@arm.com>
13291 David Sherwood <david.sherwood@arm.com>
13293 * coretypes.h (fixed_size_mode): Declare.
13294 (fixed_size_mode_pod): New typedef.
13295 * builtins.h (target_builtins::x_apply_args_mode)
13296 (target_builtins::x_apply_result_mode): Change type to
13297 fixed_size_mode_pod.
13298 * builtins.c (apply_args_size, apply_result_size, result_vector)
13299 (expand_builtin_apply_args_1, expand_builtin_apply)
13300 (expand_builtin_return): Update accordingly.
13302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13304 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13305 * cselib.c (cselib_hash_rtx): Likewise.
13306 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13307 CONST_VECTOR encoding.
13309 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13310 Jeff Law <law@redhat.com>
13313 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13314 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13315 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13316 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13319 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13320 explicitly probe *sp in a noreturn function if there were any callee
13321 register saves or frame pointer is needed.
13323 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13326 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13327 BLKmode for ternary, binary or unary expressions.
13330 * var-tracking.c (delete_vta_debug_insn): New inline function.
13331 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13332 insns from get_insns () to NULL instead of each bb separately.
13333 Use delete_vta_debug_insn. No longer static.
13334 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13335 delete_vta_debug_insns callers.
13336 * rtl.h (delete_vta_debug_insns): Declare.
13337 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13338 instead of variable_tracking_main.
13340 2018-01-03 Martin Sebor <msebor@redhat.com>
13342 PR tree-optimization/83603
13343 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13344 arguments past the endof the argument list in functions declared
13345 without a prototype.
13346 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13347 Avoid checking when arguments are null.
13349 2018-01-03 Martin Sebor <msebor@redhat.com>
13352 * doc/extend.texi (attribute const): Fix a typo.
13353 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13354 issuing -Wsuggest-attribute for void functions.
13356 2018-01-03 Martin Sebor <msebor@redhat.com>
13358 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13359 offset_int::from instead of wide_int::to_shwi.
13360 (maybe_diag_overlap): Remove assertion.
13361 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13362 * gimple-ssa-sprintf.c (format_directive): Same.
13363 (parse_directive): Same.
13364 (sprintf_dom_walker::compute_format_length): Same.
13365 (try_substitute_return_value): Same.
13367 2018-01-03 Jeff Law <law@redhat.com>
13369 PR middle-end/83654
13370 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13371 non-constant residual for zero at runtime and avoid probing in
13372 that case. Reorganize code for trailing problem to mirror handling
13375 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13377 PR tree-optimization/83501
13378 * tree-ssa-strlen.c (get_string_cst): New.
13379 (handle_char_store): Call get_string_cst.
13381 2018-01-03 Martin Liska <mliska@suse.cz>
13383 PR tree-optimization/83593
13384 * tree-ssa-strlen.c: Include tree-cfg.h.
13385 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13386 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13387 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13389 (strlen_dom_walker::before_dom_children): Call
13390 gimple_purge_dead_eh_edges. Dump tranformation with details
13392 (strlen_dom_walker::before_dom_children): Update call by adding
13393 new argument cleanup_eh.
13394 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13396 2018-01-03 Martin Liska <mliska@suse.cz>
13399 * cif-code.def (VARIADIC_THUNK): New enum value.
13400 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13403 2018-01-03 Jan Beulich <jbeulich@suse.com>
13405 * sse.md (mov<mode>_internal): Tighten condition for when to use
13406 vmovdqu<ssescalarsize> for TI and OI modes.
13408 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13410 Update copyright years.
13412 2018-01-03 Martin Liska <mliska@suse.cz>
13415 * ipa-visibility.c (function_and_variable_visibility): Skip
13416 functions with noipa attribure.
13418 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13420 * gcc.c (process_command): Update copyright notice dates.
13421 * gcov-dump.c (print_version): Ditto.
13422 * gcov.c (print_version): Ditto.
13423 * gcov-tool.c (print_version): Ditto.
13424 * gengtype.c (create_file): Ditto.
13425 * doc/cpp.texi: Bump @copying's copyright year.
13426 * doc/cppinternals.texi: Ditto.
13427 * doc/gcc.texi: Ditto.
13428 * doc/gccint.texi: Ditto.
13429 * doc/gcov.texi: Ditto.
13430 * doc/install.texi: Ditto.
13431 * doc/invoke.texi: Ditto.
13433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13435 * vector-builder.h (vector_builder::m_full_nelts): Change from
13436 unsigned int to poly_uint64.
13437 (vector_builder::full_nelts): Update prototype accordingly.
13438 (vector_builder::new_vector): Likewise.
13439 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13440 (vector_builder::operator ==): Likewise.
13441 (vector_builder::finalize): Likewise.
13442 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13443 Take the number of elements as a poly_uint64 rather than an
13445 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13446 from unsigned int to poly_uint64.
13447 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13448 (vec_perm_indices::new_vector): Likewise.
13449 (vec_perm_indices::length): Likewise.
13450 (vec_perm_indices::nelts_per_input): Likewise.
13451 (vec_perm_indices::input_nelts): Likewise.
13452 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13453 number of elements per input as a poly_uint64 rather than an
13454 unsigned int. Use the original encoding for variable-length
13455 vectors, rather than clamping each individual element.
13456 For the second and subsequent elements in each pattern,
13457 clamp the step and base before clamping their sum.
13458 (vec_perm_indices::series_p): Handle polynomial element counts.
13459 (vec_perm_indices::all_in_range_p): Likewise.
13460 (vec_perm_indices_to_tree): Likewise.
13461 (vec_perm_indices_to_rtx): Likewise.
13462 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13463 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13464 (tree_vector_builder::new_binary_operation): Handle polynomial
13465 element counts. Return false if we need to know the number
13466 of elements at compile time.
13467 * fold-const.c (fold_vec_perm): Punt if the number of elements
13468 isn't known at compile time.
13470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13472 * vec-perm-indices.h (vec_perm_builder): Change element type
13473 from HOST_WIDE_INT to poly_int64.
13474 (vec_perm_indices::element_type): Update accordingly.
13475 (vec_perm_indices::clamp): Handle polynomial element_types.
13476 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13477 (vec_perm_indices::all_in_range_p): Likewise.
13478 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13480 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13481 polynomial vec_perm_indices element types.
13482 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13483 * fold-const.c (fold_vec_perm): Likewise.
13484 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13485 * tree-vect-generic.c (lower_vec_perm): Likewise.
13486 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13487 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13488 element type to HOST_WIDE_INT.
13490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13491 Alan Hayward <alan.hayward@arm.com>
13492 David Sherwood <david.sherwood@arm.com>
13494 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13495 rather than an int. Use plus_constant.
13496 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13497 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13500 Alan Hayward <alan.hayward@arm.com>
13501 David Sherwood <david.sherwood@arm.com>
13503 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13504 a HOST_WIDE_INT to a poly_int64.
13506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13507 Alan Hayward <alan.hayward@arm.com>
13508 David Sherwood <david.sherwood@arm.com>
13510 * calls.c (load_register_parameters): Cope with polynomial
13511 mode sizes. Require a constant size for BLKmode parameters
13512 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13513 forces a parameter to be padded at the lsb end in order to
13514 fill a complete number of words, require the parameter size
13515 to be ordered wrt UNITS_PER_WORD.
13517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13518 Alan Hayward <alan.hayward@arm.com>
13519 David Sherwood <david.sherwood@arm.com>
13521 * reload1.c (spill_stack_slot_width): Change element type
13522 from unsigned int to poly_uint64_pod.
13523 (alter_reg): Treat mode sizes as polynomial.
13525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13526 Alan Hayward <alan.hayward@arm.com>
13527 David Sherwood <david.sherwood@arm.com>
13529 * reload.c (complex_word_subreg_p): New function.
13530 (reload_inner_reg_of_subreg, push_reload): Use it.
13532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13533 Alan Hayward <alan.hayward@arm.com>
13534 David Sherwood <david.sherwood@arm.com>
13536 * lra-constraints.c (process_alt_operands): Reject matched
13537 operands whose sizes aren't ordered.
13538 (match_reload): Refer to this check here.
13540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13541 Alan Hayward <alan.hayward@arm.com>
13542 David Sherwood <david.sherwood@arm.com>
13544 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13545 that the mode size is in the set {1, 2, 4, 8, 16}.
13547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13548 Alan Hayward <alan.hayward@arm.com>
13549 David Sherwood <david.sherwood@arm.com>
13551 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13552 Use plus_constant instead of gen_rtx_PLUS.
13554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13555 Alan Hayward <alan.hayward@arm.com>
13556 David Sherwood <david.sherwood@arm.com>
13558 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13559 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13560 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13561 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13562 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13563 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13564 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13565 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13566 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13567 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13569 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13570 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13571 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13572 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13573 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13574 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13575 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13576 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13577 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13578 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13580 * expr.c (emit_move_resolve_push): Treat the input and result
13581 of PUSH_ROUNDING as a poly_int64.
13582 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13583 (emit_push_insn): Likewise.
13584 * lra-eliminations.c (mark_not_eliminable): Likewise.
13585 * recog.c (push_operand): Likewise.
13586 * reload1.c (elimination_effects): Likewise.
13587 * rtlanal.c (nonzero_bits1): Likewise.
13588 * calls.c (store_one_arg): Likewise. Require the padding to be
13589 known at compile time.
13591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13592 Alan Hayward <alan.hayward@arm.com>
13593 David Sherwood <david.sherwood@arm.com>
13595 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13596 Use plus_constant instead of gen_rtx_PLUS.
13598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13599 Alan Hayward <alan.hayward@arm.com>
13600 David Sherwood <david.sherwood@arm.com>
13602 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13603 rather than an int.
13605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13606 Alan Hayward <alan.hayward@arm.com>
13607 David Sherwood <david.sherwood@arm.com>
13609 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13610 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13611 via stack temporaries. Treat the mode size as polynomial too.
13613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13614 Alan Hayward <alan.hayward@arm.com>
13615 David Sherwood <david.sherwood@arm.com>
13617 * expr.c (expand_expr_real_2): When handling conversions involving
13618 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13619 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13620 as a poly_uint64 too.
13622 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13623 Alan Hayward <alan.hayward@arm.com>
13624 David Sherwood <david.sherwood@arm.com>
13626 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13629 Alan Hayward <alan.hayward@arm.com>
13630 David Sherwood <david.sherwood@arm.com>
13632 * combine.c (can_change_dest_mode): Handle polynomial
13633 REGMODE_NATURAL_SIZE.
13634 * expmed.c (store_bit_field_1): Likewise.
13635 * expr.c (store_constructor): Likewise.
13636 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13637 and polynomial REGMODE_NATURAL_SIZE.
13638 (gen_lowpart_common): Likewise.
13639 * reginfo.c (record_subregs_of_mode): Likewise.
13640 * rtlanal.c (read_modify_subreg_p): Likewise.
13642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13643 Alan Hayward <alan.hayward@arm.com>
13644 David Sherwood <david.sherwood@arm.com>
13646 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13647 numbers of elements.
13649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13650 Alan Hayward <alan.hayward@arm.com>
13651 David Sherwood <david.sherwood@arm.com>
13653 * match.pd: Cope with polynomial numbers of vector elements.
13655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13656 Alan Hayward <alan.hayward@arm.com>
13657 David Sherwood <david.sherwood@arm.com>
13659 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13660 in a POINTER_PLUS_EXPR.
13662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13663 Alan Hayward <alan.hayward@arm.com>
13664 David Sherwood <david.sherwood@arm.com>
13666 * omp-simd-clone.c (simd_clone_subparts): New function.
13667 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13668 (ipa_simd_modify_function_body): Likewise.
13670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13671 Alan Hayward <alan.hayward@arm.com>
13672 David Sherwood <david.sherwood@arm.com>
13674 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13675 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13676 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13677 (expand_vector_condition, vector_element): Likewise.
13678 (subparts_gt): New function.
13679 (get_compute_type): Use subparts_gt.
13680 (count_type_subparts): Delete.
13681 (expand_vector_operations_1): Use subparts_gt instead of
13682 count_type_subparts.
13684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13685 Alan Hayward <alan.hayward@arm.com>
13686 David Sherwood <david.sherwood@arm.com>
13688 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13689 (vect_compile_time_alias): ...this new function. Do the calculation
13690 on poly_ints rather than trees.
13691 (vect_prune_runtime_alias_test_list): Update call accordingly.
13693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13694 Alan Hayward <alan.hayward@arm.com>
13695 David Sherwood <david.sherwood@arm.com>
13697 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13699 (vect_schedule_slp_instance): Likewise.
13701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13702 Alan Hayward <alan.hayward@arm.com>
13703 David Sherwood <david.sherwood@arm.com>
13705 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13706 constant and extern definitions for variable-length vectors.
13707 (vect_get_constant_vectors): Note that the number of units
13708 is known to be constant.
13710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13711 Alan Hayward <alan.hayward@arm.com>
13712 David Sherwood <david.sherwood@arm.com>
13714 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13715 of units as polynomial. Choose between WIDE and NARROW based
13718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13719 Alan Hayward <alan.hayward@arm.com>
13720 David Sherwood <david.sherwood@arm.com>
13722 * tree-vect-stmts.c (simd_clone_subparts): New function.
13723 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13726 Alan Hayward <alan.hayward@arm.com>
13727 David Sherwood <david.sherwood@arm.com>
13729 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13730 vectors as polynomial. Use build_index_vector for
13731 IFN_GOMP_SIMD_LANE.
13733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13734 Alan Hayward <alan.hayward@arm.com>
13735 David Sherwood <david.sherwood@arm.com>
13737 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13738 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13739 for variable-length vectors.
13740 (vectorizable_mask_load_store): Treat the number of units as
13741 polynomial, asserting that it is constant if the condition has
13742 already been enforced.
13743 (vectorizable_store, vectorizable_load): Likewise.
13745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13746 Alan Hayward <alan.hayward@arm.com>
13747 David Sherwood <david.sherwood@arm.com>
13749 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13750 of units as polynomial. Punt if we can't tell at compile time
13751 which vector contains the final result.
13753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13754 Alan Hayward <alan.hayward@arm.com>
13755 David Sherwood <david.sherwood@arm.com>
13757 * tree-vect-loop.c (vectorizable_induction): Treat the number
13758 of units as polynomial. Punt on SLP inductions. Use an integer
13759 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13760 cast of such a series for variable-length floating-point
13763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13764 Alan Hayward <alan.hayward@arm.com>
13765 David Sherwood <david.sherwood@arm.com>
13767 * tree.h (build_index_vector): Declare.
13768 * tree.c (build_index_vector): New function.
13769 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13770 of units as polynomial, forcibly converting it to a constant if
13771 vectorizable_reduction has already enforced the condition.
13772 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13773 to create a {1,2,3,...} vector.
13774 (vectorizable_reduction): Treat the number of units as polynomial.
13775 Choose vectype_in based on the largest scalar element size rather
13776 than the smallest number of units. Enforce the restrictions
13779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13780 Alan Hayward <alan.hayward@arm.com>
13781 David Sherwood <david.sherwood@arm.com>
13783 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13784 number of units as polynomial.
13786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13787 Alan Hayward <alan.hayward@arm.com>
13788 David Sherwood <david.sherwood@arm.com>
13790 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13791 * target.def (autovectorize_vector_sizes): Return the vector sizes
13792 by pointer, using vector_sizes rather than a bitmask.
13793 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13794 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13795 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13797 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13798 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13799 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13800 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13801 * omp-general.c (omp_max_vf): Likewise.
13802 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13803 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13804 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13805 * tree-vect-slp.c (vect_slp_bb): Likewise.
13806 * doc/tm.texi: Regenerate.
13807 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13809 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13810 the vector size as a poly_uint64 rather than an unsigned int.
13811 (current_vector_size): Change from an unsigned int to a poly_uint64.
13812 (get_vectype_for_scalar_type): Update accordingly.
13813 * tree.h (build_truth_vector_type): Take the size and number of
13814 units as a poly_uint64 rather than an unsigned int.
13815 (build_vector_type): Add a temporary overload that takes
13816 the number of units as a poly_uint64 rather than an unsigned int.
13817 * tree.c (make_vector_type): Likewise.
13818 (build_truth_vector_type): Take the number of units as a poly_uint64
13819 rather than an unsigned int.
13821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13822 Alan Hayward <alan.hayward@arm.com>
13823 David Sherwood <david.sherwood@arm.com>
13825 * target.def (get_mask_mode): Take the number of units and length
13826 as poly_uint64s rather than unsigned ints.
13827 * targhooks.h (default_get_mask_mode): Update accordingly.
13828 * targhooks.c (default_get_mask_mode): Likewise.
13829 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13830 * doc/tm.texi: Regenerate.
13832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13833 Alan Hayward <alan.hayward@arm.com>
13834 David Sherwood <david.sherwood@arm.com>
13836 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13837 * omp-general.c (omp_max_vf): Likewise.
13838 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13839 (expand_omp_simd): Handle polynomial safelen.
13840 * omp-low.c (omplow_simd_context): Add a default constructor.
13841 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13842 (lower_rec_simd_input_clauses): Update accordingly.
13843 (lower_rec_input_clauses): Likewise.
13845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13846 Alan Hayward <alan.hayward@arm.com>
13847 David Sherwood <david.sherwood@arm.com>
13849 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13850 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13851 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13852 (vect_analyze_slp_cost): Likewise.
13853 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13854 (vect_model_load_cost): Likewise.
13856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13857 Alan Hayward <alan.hayward@arm.com>
13858 David Sherwood <david.sherwood@arm.com>
13860 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13861 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13862 from an unsigned int * to a poly_uint64_pod *.
13863 (calculate_unrolling_factor): New function.
13864 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13867 Alan Hayward <alan.hayward@arm.com>
13868 David Sherwood <david.sherwood@arm.com>
13870 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13871 from an unsigned int to a poly_uint64.
13872 (_loop_vec_info::slp_unrolling_factor): Likewise.
13873 (_loop_vec_info::vectorization_factor): Change from an int
13875 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13876 (vect_get_num_vectors): New function.
13877 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13878 (vect_get_num_copies): Use vect_get_num_vectors.
13879 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13880 to an unsigned int *.
13881 (vect_analyze_data_refs): Change min_vf from an int * to a
13883 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13884 than an unsigned HOST_WIDE_INT.
13885 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13886 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13887 to an unsigned int *.
13888 (vect_analyze_data_ref_dependences): Likewise.
13889 (vect_compute_data_ref_alignment): Handle polynomial vf.
13890 (vect_enhance_data_refs_alignment): Likewise.
13891 (vect_prune_runtime_alias_test_list): Likewise.
13892 (vect_shift_permute_load_chain): Likewise.
13893 (vect_supportable_dr_alignment): Likewise.
13894 (dependence_distance_ge_vf): Take the vectorization factor as a
13895 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13896 (vect_analyze_data_refs): Change min_vf from an int * to a
13898 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13899 vfm1 as a poly_uint64 rather than an int. Make the same change
13900 for the returned bound_scalar.
13901 (vect_gen_vector_loop_niters): Handle polynomial vf.
13902 (vect_do_peeling): Likewise. Update call to
13903 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13904 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13906 * tree-vect-loop.c (vect_determine_vectorization_factor)
13907 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13908 (vect_get_known_peeling_cost): Likewise.
13909 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13910 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13911 (vect_transform_loop): Likewise. Use the lowest possible VF when
13912 updating the upper bounds of the loop.
13913 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13914 rather than an int.
13915 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13916 polynomial unroll factors.
13917 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13918 (vect_make_slp_decision): Likewise.
13919 (vect_supported_load_permutation_p): Likewise, and polynomial
13921 (vect_analyze_slp_cost): Handle polynomial vf.
13922 (vect_slp_analyze_node_operations): Likewise.
13923 (vect_slp_analyze_bb_1): Likewise.
13924 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13925 than an unsigned HOST_WIDE_INT.
13926 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13927 (vectorizable_load): Handle polynomial vf.
13928 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13930 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13933 Alan Hayward <alan.hayward@arm.com>
13934 David Sherwood <david.sherwood@arm.com>
13936 * match.pd: Handle bit operations involving three constants
13937 and try to fold one pair.
13939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13941 * tree-vect-loop-manip.c: Include gimple-fold.h.
13942 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13943 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13944 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13945 Add a path that uses a step of VF instead of 1, but disable it
13947 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13948 and niters_no_overflow parameters. Update calls to
13949 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13950 Create a new SSA name if the latter choses to use a ste other
13951 than zero, and return it via niters_vector_mult_vf_var.
13952 * tree-vect-loop.c (vect_transform_loop): Update calls to
13953 vect_do_peeling, vect_gen_vector_loop_niters and
13954 slpeel_make_loop_iterate_ntimes.
13955 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13956 (vect_gen_vector_loop_niters): Update declarations after above changes.
13958 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13960 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13961 128-bit round to integer instructions.
13962 (ceil<mode>2): Likewise.
13963 (btrunc<mode>2): Likewise.
13964 (round<mode>2): Likewise.
13966 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13968 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13969 unaligned VSX load/store on P8/P9.
13970 (expand_block_clear): Allow the use of unaligned VSX
13971 load/store on P8/P9.
13973 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13975 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13977 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13978 swap associated with both a load and a store.
13980 2018-01-02 Andrew Waterman <andrew@sifive.com>
13982 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13983 * config/riscv/riscv.md (clear_cache): Use it.
13985 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13987 * web.c: Remove out-of-date comment.
13989 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13991 * expr.c (fixup_args_size_notes): Check that any existing
13992 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13993 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13994 (emit_single_push_insn): ...here.
13996 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13998 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13999 (const_vector_encoded_nelts): New function.
14000 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14001 (const_vector_int_elt, const_vector_elt): Declare.
14002 * emit-rtl.c (const_vector_int_elt_1): New function.
14003 (const_vector_elt): Likewise.
14004 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14005 of CONST_VECTOR_ELT.
14007 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14009 * expr.c: Include rtx-vector-builder.h.
14010 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14011 directly on the tree encoding.
14012 (const_vector_from_tree): Likewise.
14013 * optabs.c: Include rtx-vector-builder.h.
14014 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14015 sequence of "u" values.
14016 * vec-perm-indices.c: Include rtx-vector-builder.h.
14017 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14018 directly on the vec_perm_indices encoding.
14020 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14022 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14023 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14024 * rtx-vector-builder.h: New file.
14025 * rtx-vector-builder.c: Likewise.
14026 * rtl.h (rtx_def::u2): Add a const_vector field.
14027 (CONST_VECTOR_NPATTERNS): New macro.
14028 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14029 (CONST_VECTOR_DUPLICATE_P): Likewise.
14030 (CONST_VECTOR_STEPPED_P): Likewise.
14031 (CONST_VECTOR_ENCODED_ELT): Likewise.
14032 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14033 (unwrap_const_vec_duplicate): Likewise.
14034 (const_vec_series_p): Check for a non-duplicated vector encoding.
14035 Say that the function only returns true for integer vectors.
14036 * emit-rtl.c: Include rtx-vector-builder.h.
14037 (gen_const_vec_duplicate_1): Delete.
14038 (gen_const_vector): Call gen_const_vec_duplicate instead of
14039 gen_const_vec_duplicate_1.
14040 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14041 (gen_const_vec_duplicate): Use rtx_vector_builder.
14042 (gen_const_vec_series): Likewise.
14043 (gen_rtx_CONST_VECTOR): Likewise.
14044 * config/powerpcspe/powerpcspe.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.
14048 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14049 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14050 Build a new vector rather than modifying a CONST_VECTOR in-place.
14051 (handle_special_swappables): Update call accordingly.
14053 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14055 * simplify-rtx.c (simplify_const_binary_operation): Use
14056 CONST_VECTOR_ELT instead of XVECEXP.
14058 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14060 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14061 the selector elements to be different from the data elements
14062 if the selector is a VECTOR_CST.
14063 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14064 ssizetype for the selector.
14066 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14068 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14069 before testing each element individually.
14070 * tree-vect-generic.c (lower_vec_perm): Likewise.
14072 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14074 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14075 * selftest-run-tests.c (selftest::run_tests): Call it.
14076 * vector-builder.h (vector_builder::operator ==): New function.
14077 (vector_builder::operator !=): Likewise.
14078 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14079 (vec_perm_indices::all_from_input_p): New function.
14080 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14081 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14082 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14083 instead of reading the VECTOR_CST directly. Detect whether both
14084 vector inputs are the same before constructing the vec_perm_indices,
14085 and update the number of inputs argument accordingly. Use the
14086 utility functions added above. Only construct sel2 if we need to.
14088 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14090 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14091 the broadcast of the low byte.
14092 (expand_mult_highpart): Use an explicit encoding for the permutes.
14093 * optabs-query.c (can_mult_highpart_p): Likewise.
14094 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14095 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14096 (vectorizable_bswap): Likewise.
14097 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14098 explicit encoding for the power-of-2 permutes.
14099 (vect_permute_store_chain): Likewise.
14100 (vect_grouped_load_supported): Likewise.
14101 (vect_permute_load_chain): Likewise.
14103 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14105 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14106 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14107 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14108 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14109 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14110 (vect_gen_perm_mask_any): Likewise.
14112 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14114 * int-vector-builder.h: New file.
14115 * vec-perm-indices.h: Include int-vector-builder.h.
14116 (vec_perm_indices): Redefine as an int_vector_builder.
14117 (auto_vec_perm_indices): Delete.
14118 (vec_perm_builder): Redefine as a stand-alone class.
14119 (vec_perm_indices::vec_perm_indices): New function.
14120 (vec_perm_indices::clamp): Likewise.
14121 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14122 (vec_perm_indices::new_vector): New function.
14123 (vec_perm_indices::new_expanded_vector): Update for new
14124 vec_perm_indices class.
14125 (vec_perm_indices::rotate_inputs): New function.
14126 (vec_perm_indices::all_in_range_p): Operate directly on the
14127 encoded form, without computing elided elements.
14128 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14129 encoding. Update for new vec_perm_indices class.
14130 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14131 the given vec_perm_builder.
14132 (expand_vec_perm_var): Update vec_perm_builder constructor.
14133 (expand_mult_highpart): Use vec_perm_builder instead of
14134 auto_vec_perm_indices.
14135 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14136 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14137 or double series encoding as appropriate.
14138 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14139 vec_perm_indices instead of auto_vec_perm_indices.
14140 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14141 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14142 (vect_permute_store_chain): Likewise.
14143 (vect_grouped_load_supported): Likewise.
14144 (vect_permute_load_chain): Likewise.
14145 (vect_shift_permute_load_chain): Likewise.
14146 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14147 (vect_transform_slp_perm_load): Likewise.
14148 (vect_schedule_slp_instance): Likewise.
14149 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14150 (vectorizable_mask_load_store): Likewise.
14151 (vectorizable_bswap): Likewise.
14152 (vectorizable_store): Likewise.
14153 (vectorizable_load): Likewise.
14154 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14155 vec_perm_indices instead of auto_vec_perm_indices. Use
14156 tree_to_vec_perm_builder to read the vector from a tree.
14157 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14158 vec_perm_builder instead of a vec_perm_indices.
14159 (have_whole_vector_shift): Use vec_perm_builder and
14160 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14161 truncation to calc_vec_perm_mask_for_shift.
14162 (vect_create_epilog_for_reduction): Likewise.
14163 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14164 from auto_vec_perm_indices to vec_perm_indices.
14165 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14166 instead of changing individual elements.
14167 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14168 the vector in d.perm.
14169 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14170 from auto_vec_perm_indices to vec_perm_indices.
14171 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14172 instead of changing individual elements.
14173 (arm_vectorize_vec_perm_const): Use new_vector to install
14174 the vector in d.perm.
14175 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14176 Update vec_perm_builder constructor.
14177 (rs6000_expand_interleave): Likewise.
14178 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14179 (rs6000_expand_interleave): Likewise.
14181 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14183 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14184 to qimode could truncate the indices.
14185 * optabs.c (expand_vec_perm_var): Likewise.
14187 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14189 * Makefile.in (OBJS): Add vec-perm-indices.o.
14190 * vec-perm-indices.h: New file.
14191 * vec-perm-indices.c: Likewise.
14192 * target.h (vec_perm_indices): Replace with a forward class
14194 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14195 * optabs.h: Include vec-perm-indices.h.
14196 (expand_vec_perm): Delete.
14197 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14198 (expand_vec_perm_const): Declare.
14199 * target.def (vec_perm_const_ok): Replace with...
14200 (vec_perm_const): ...this new hook.
14201 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14202 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14203 * doc/tm.texi: Regenerate.
14204 * optabs.def (vec_perm_const): Delete.
14205 * doc/md.texi (vec_perm_const): Likewise.
14206 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14207 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14208 expand_vec_perm for constant permutation vectors. Assert that
14209 the mode of variable permutation vectors is the integer equivalent
14210 of the mode that is being permuted.
14211 * optabs-query.h (selector_fits_mode_p): Declare.
14212 * optabs-query.c: Include vec-perm-indices.h.
14213 (selector_fits_mode_p): New function.
14214 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14215 is defined, instead of checking whether the vec_perm_const_optab
14216 exists. Use targetm.vectorize.vec_perm_const instead of
14217 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14218 fit in the vector mode before using a variable permute.
14219 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14220 vec_perm_indices instead of an rtx.
14221 (expand_vec_perm): Replace with...
14222 (expand_vec_perm_const): ...this new function. Take the selector
14223 as a vec_perm_indices rather than an rtx. Also take the mode of
14224 the selector. Update call to shift_amt_for_vec_perm_mask.
14225 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14226 Use vec_perm_indices::new_expanded_vector to expand the original
14227 selector into bytes. Check whether the indices fit in the vector
14228 mode before using a variable permute.
14229 (expand_vec_perm_var): Make global.
14230 (expand_mult_highpart): Use expand_vec_perm_const.
14231 * fold-const.c: Includes vec-perm-indices.h.
14232 * tree-ssa-forwprop.c: Likewise.
14233 * tree-vect-data-refs.c: Likewise.
14234 * tree-vect-generic.c: Likewise.
14235 * tree-vect-loop.c: Likewise.
14236 * tree-vect-slp.c: Likewise.
14237 * tree-vect-stmts.c: Likewise.
14238 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14240 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14241 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14242 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14243 (aarch64_vectorize_vec_perm_const): ...this new function.
14244 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14245 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14246 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14247 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14248 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14249 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14250 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14252 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14253 check for NEON modes.
14254 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14255 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14256 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14257 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14259 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14260 the old VEC_PERM_CONST conditions.
14261 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14262 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14263 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14264 (ia64_vectorize_vec_perm_const_ok): Merge into...
14265 (ia64_vectorize_vec_perm_const): ...this new function.
14266 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14267 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14268 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14269 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14270 * config/mips/mips.c (mips_expand_vec_perm_const)
14271 (mips_vectorize_vec_perm_const_ok): Merge into...
14272 (mips_vectorize_vec_perm_const): ...this new function.
14273 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14274 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14275 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14276 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14277 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14278 (rs6000_expand_vec_perm_const): Delete.
14279 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14281 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14282 (altivec_expand_vec_perm_const_le): Take each operand individually.
14283 Operate on constant selectors rather than rtxes.
14284 (altivec_expand_vec_perm_const): Likewise. Update call to
14285 altivec_expand_vec_perm_const_le.
14286 (rs6000_expand_vec_perm_const): Delete.
14287 (rs6000_vectorize_vec_perm_const_ok): Delete.
14288 (rs6000_vectorize_vec_perm_const): New function.
14289 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14290 an element count and rtx array.
14291 (rs6000_expand_extract_even): Update call accordingly.
14292 (rs6000_expand_interleave): Likewise.
14293 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14294 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14295 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14296 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14297 (rs6000_expand_vec_perm_const): Delete.
14298 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14299 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14300 (altivec_expand_vec_perm_const_le): Take each operand individually.
14301 Operate on constant selectors rather than rtxes.
14302 (altivec_expand_vec_perm_const): Likewise. Update call to
14303 altivec_expand_vec_perm_const_le.
14304 (rs6000_expand_vec_perm_const): Delete.
14305 (rs6000_vectorize_vec_perm_const_ok): Delete.
14306 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14307 reference to the SPE evmerge intructions.
14308 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14309 an element count and rtx array.
14310 (rs6000_expand_extract_even): Update call accordingly.
14311 (rs6000_expand_interleave): Likewise.
14312 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14313 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14315 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14317 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14319 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14320 vector mode and that that mode matches the mode of the data
14322 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14323 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14324 directly using expand_vec_perm_1 when forcing selectors into
14326 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14328 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14330 * optabs-query.h (can_vec_perm_p): Delete.
14331 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14332 * optabs-query.c (can_vec_perm_p): Split into...
14333 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14334 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14335 particular selector is valid.
14336 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14337 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14338 (vect_grouped_load_supported): Likewise.
14339 (vect_shift_permute_load_chain): Likewise.
14340 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14341 (vect_transform_slp_perm_load): Likewise.
14342 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14343 (vectorizable_bswap): Likewise.
14344 (vect_gen_perm_mask_checked): Likewise.
14345 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14346 implementations of variable permutation vectors into account
14347 when deciding which selector to use.
14348 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14349 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14350 with a false third argument.
14351 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14352 to test whether the constant selector is valid and can_vec_perm_var_p
14353 to test whether a variable selector is valid.
14355 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14357 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14358 * optabs-query.c (can_vec_perm_p): Likewise.
14359 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14360 instead of vec_perm_indices.
14361 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14362 (vect_gen_perm_mask_checked): Likewise,
14363 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14364 (vect_gen_perm_mask_checked): Likewise,
14366 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14368 * optabs-query.h (qimode_for_vec_perm): Declare.
14369 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14370 (qimode_for_vec_perm): ...this new function.
14371 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14373 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14375 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14376 does not have a conditional at the top.
14378 2018-01-02 Richard Biener <rguenther@suse.de>
14380 * ipa-inline.c (big_speedup_p): Fix expression.
14382 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14385 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14388 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14392 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14393 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14394 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14395 cond_taken_branch_cost 3->4.
14397 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14399 PR tree-optimization/83581
14400 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14401 TODO_cleanup_cfg if any changes have been made.
14403 PR middle-end/83608
14404 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14405 convert_modes if target mode has the right side, but different mode
14408 PR middle-end/83609
14409 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14410 last argument when extracting from CONCAT. If either from_real or
14411 from_imag is NULL, use expansion through memory. If result is not
14412 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14413 the parts directly to inner mode, if even that fails, use expansion
14416 PR middle-end/83623
14417 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14418 check for bswap in mode rather than HImode and use that in expand_unop
14421 Copyright (C) 2018 Free Software Foundation, Inc.
14423 Copying and distribution of this file, with or without modification,
14424 are permitted in any medium without royalty provided the copyright
14425 notice and this notice are preserved.