1 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5 2018-05-27 Paul Koning <ni1d@arrl.net>
7 * config/pdp11/pdp11.md (truncsihi2): Remove.
9 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
10 Chung-Ju Wu <jasonwucj@gmail.com>
12 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
14 (unaligned_store_dw): Ditto.
15 * config/nds32/nds32-memory-manipulation.c
16 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
17 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
18 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
19 (emit_setmem_word_loop): Rename to ...
20 (emit_setmem_doubleword_loop): ... this.
21 (nds32_gen_dup_4_byte_to_word_value): New function.
22 (nds32_gen_dup_8_byte_to_double_word_value): New function.
23 (nds32_expand_setmem_loop): Refine implementation.
24 (nds32_expand_setmem_loop_v3m): Ditto.
25 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
28 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
30 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
32 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
34 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
35 (nds32_init_machine_status): Initialize machine->attr_naked_p and
36 machine->attr_no_prologue_p.
37 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
38 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
39 (nds32_expand_epilogue): Consider attr_naked_p.
40 (nds32_expand_epilogue_v3pop): Likewise.
41 (nds32_can_use_return_insn): Likewise.
42 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
43 attr_no_prologue_p fields.
44 * config/nds32/nds32.opt (mret-in-naked-func): New option.
46 2018-05-27 Jakub Jelinek <jakub@redhat.com>
49 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
52 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
54 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
56 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
58 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
60 (*<floatsuffix>floatv2div2sf2): Rename to ...
61 (*float<floatunssuffix>v2div2sf2): ... this.
62 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
63 (float<floatunssuffix>v2div2sf2_mask): ... this.
64 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
65 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
66 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
68 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
70 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
72 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
74 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
76 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
78 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
79 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
80 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
81 gen_ufix_truncv8dfv8si2.
82 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
83 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
84 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
85 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
86 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
87 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
88 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
89 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
91 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
95 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
97 2018-05-25 Jim Wilson <jimw@sifive.com>
99 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
100 * config/riscv/riscv.c (struct machine_function): Add
101 interrupt_handler_p and attribute_checked_p fields.
102 (riscv_attribute_table): Add interrupt.
103 (riscv_interrupt_type_p): New.
104 (riscv_save_reg_p): Save extra regs for interrupt handler.
105 (riscv_use_save_libcall): Return false for interrupt handler.
106 (riscv_first_stack_step): Add forward declaration.
107 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
108 for interrupt handler with large frame. Use it for saved reg list.
109 (riscv_expand_prologue): Move flag_stack_usage_info support to
110 eliminate duplication.
111 (riscv_expand_epilogue): Generate mret for interrupt handler.
112 (riscv_epilogue_uses): New.
113 (riscv_can_use_return_insn): Return false for interrupt handler.
114 (riscv_function_ok_for_sibcall): Likewise.
115 (riscv_set_current_function): Add interrupt handler support.
116 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
117 * config/riscv/riscv.md (UNSPECV_MRET): New.
119 (riscv_frflags, riscv_fsflags): Use tab after opcode.
121 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
123 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
125 PR tree-optimization/85712
126 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
127 this candidate has already been replaced in-situ by a copy.
129 2018-05-25 Jason Merrill <jason@redhat.com>
131 PR c++/80485 - inline function non-zero address.
132 * symtab.c (nonzero_address): Check DECL_COMDAT.
134 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
137 * config/alpha/alpha.md (ashlsi3): New insn pattern.
138 (*ashlsi_se): Rename from *ashldi_se. Define as sign
139 extension of SImode operation. Use const123_operand predicate.
141 (*saddl_se_1): Ditto.
143 (*ssubl_se_1): Ditto.
144 * config/alpha/predicates.md (const123_operand): New predicate.
145 * config/alpha/constraints.md (P): Use IN_RANGE.
147 2018-05-25 Richard Biener <rguenther@suse.de>
149 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
151 (ref_maybe_used_by_stmt_p): Likewise.
152 (stmt_may_clobber_ref_p): Likewise.
153 (stmt_may_clobber_ref_p_1): Likewise.
154 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
156 (ref_maybe_used_by_stmt_p): Likewise.
157 (stmt_may_clobber_ref_p): Likewise.
158 (stmt_may_clobber_ref_p_1): Likewise.
159 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
160 the alias oracle to disambiguate DRs with stmts DR analysis
162 (vect_analyze_data_refs): Do not give up on not analyzable
163 DRs for BB vectorization. Remove code truncating the dataref
166 2018-05-25 Jakub Jelinek <jakub@redhat.com>
169 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
170 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
171 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
173 2018-05-25 Richard Biener <rguenther@suse.de>
175 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
176 function, combining stmt data ref gathering and fatal analysis
178 (vect_analyze_data_refs): Remove now redudnant code and simplify.
179 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
180 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
181 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
182 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
184 2018-05-25 Bin Cheng <bin.cheng@arm.com>
186 PR tree-optimization/85720
187 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
188 SCC if all partitions are builtins.
189 (version_loop_by_alias_check): New parameter. Generate cancelable
190 runtime alias check if all partitions are builtins.
191 (distribute_loop): Update call to above function.
193 2018-05-25 Bin Cheng <bin.cheng@arm.com>
195 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
196 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
197 (parm_default_def_partition_arg): Ditto.
198 (set_parm_default_def_partition): Ditto.
199 (get_parm_default_def_partitions): Ditto and make it static.
200 (get_undefined_value_partitions): Ditto and make it static.
201 (remove_ssa_form): Refactor call to init_var_map here.
202 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
203 computation for loop region.
204 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
205 (register_default_def): Delete.
206 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
207 (parm_default_def_partition_arg): Ditto.
208 (set_parm_default_def_partition): Ditto.
209 (get_parm_default_def_partitions): Ditto and make it static.
210 (get_undefined_value_partitions): Ditto and make it static.
211 (coalesce_with_default, coalesce_with_default): Update comment.
212 (create_coalesce_list_for_region): New func factored out from
213 create_outofssa_var_map.
214 (populate_coalesce_list_for_outofssa): New func factored out from
215 create_outofssa_var_map and coalesce_ssa_name.
216 (create_outofssa_var_map): Delete.
217 (coalesce_ssa_name): Refactor to support live range computation.
218 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
219 (get_parm_default_def_partitions): Delete.
220 (get_undefined_value_partitions): Ditto.
221 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
222 computation for loop region.
223 (new_tree_live_info, loe_visit_block): Ditto.
224 (live_worklist, set_var_live_on_entry): Ditto.
225 (calculate_live_on_exit, verify_live_on_entry): Ditto.
226 * tree-ssa-live.h (struct _var_map): New fields.
227 (init_var_map): Change decl.
228 (region_contains_p): New.
230 2018-05-25 Bin Cheng <bin.cheng@arm.com>
232 * tree-ssa-live.h (live_merge_and_clear): Delete.
234 2018-05-25 Richard Biener <rguenther@suse.de>
237 * tree-dump.c (dequeue_and_dump): Remove access to removed
238 operand 2 of a SWITCH_EXPR.
240 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
242 * doc/sourcebuild.texi (vect_double_cond_arith): Include
243 multiplication and division.
244 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
245 (cond_udiv@var{m}, cond_umod@var{m}): Document.
246 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
247 (cond_udiv_optab, cond_umod_optab): New optabs.
248 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
249 (IFN_COND_RDIV): New internal functions.
250 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
251 TRUNC_MOD_EXPR and RDIV_EXPR.
252 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
253 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
255 (SVE_INT_BINARY): Include mult.
256 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
257 (optab, sve_int_op): Handle mult.
258 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
260 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
261 for SVE_INT_BINARY_SD.
263 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
265 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
266 (optab, sve_int_op): Handle div and udiv.
267 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
268 for SVE_INT_BINARY_SD.
269 (*<optab><mode>3): New insn for the same.
271 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
273 * tree-vect-patterns.c: Include predict.h.
274 (vect_recog_divmod_pattern): Restrict check for division support
275 to when optimizing for size.
277 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
279 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
280 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
281 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
282 (gimple_match_op::set_op): Likewise.
283 (gimple_resimplify4): Declare.
284 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
285 (expr::gen_transform): Likewise.
286 (decision_tree::gen): Generate a simplification routine for 4 operands.
287 * gimple-match-head.c (gimple_simplify): Add an overload for
288 4 operands. In the top-level function, handle up to 4 call
289 arguments and call gimple_resimplify4.
290 (gimple_resimplify4): New function.
291 (build_call_internal): Pass a fourth operand.
292 (maybe_push_to_seq): Likewise.
293 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
294 Fold VEC_COND_EXPRs of an operation and a default value into
295 an IFN_COND_* function if possible.
296 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
298 (SVE_COND_FP_BINARY): Include them.
299 (optab, sve_fp_op): Handle them.
300 (SVE_INT_BINARY_REV): New code iterator.
301 (SVE_COND_FP_BINARY_REV): New int iterator.
302 (commutative): New int attribute.
303 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
305 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
307 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
308 (*cond_<optab><mode>): New patterns for reversed operands.
310 2018-05-25 Richard Biener <rguenther@suse.de>
312 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
313 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
314 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
315 (STMT_VINFO_GROUPED_ACCESS): Adjust.
316 * tree-vect-data-refs.c (everywhere): Adjust users.
317 * tree-vect-loop.c (everywhere): Likewise.
318 * tree-vect-slp.c (everywhere): Likewise.
319 * tree-vect-stmts.c (everywhere): Likewise.
320 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
322 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
324 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
326 (gcc_cv_as_section_exclude): ... this.
327 Try Solaris as #exclude syntax.
328 * configure: Regenerate.
329 * config.in: Regenerate.
330 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
332 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
333 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
335 * varasm.c (default_elf_asm_named_section): Don't check if
336 HAVE_GAS_SECTION_EXCLUDE is defined.
338 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
340 * doc/md.texi: Update the documentation of the cond_* optabs
341 to mention the new final operand. Fix GET_MODE_NUNITS call.
342 Describe the scalar case too.
343 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
344 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
346 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
347 (get_conditional_internal_fn): Update comment.
348 * tree-vect-loop.c (vectorizable_reduction): Pass the original
349 accumulator value as a final argument to conditional functions.
350 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
351 a define_expand and add an "else" operand. Assert for now that
352 the else operand is equal to operand 2. Use SVE_INT_BINARY and
353 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
354 (*cond_<optab><mode>): New patterns.
355 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
356 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
357 (UNSPEC_COND_EOR): Delete.
358 (optab): Remove associated mappings.
359 (SVE_INT_BINARY): New code iterator.
360 (sve_int_op): Remove int attribute and add "minus" to the code
362 (SVE_COND_INT_OP): Delete.
363 (SVE_COND_FP_OP): Rename to...
364 (SVE_COND_FP_BINARY): ...this.
366 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
368 * optabs.c (can_reuse_operands_p): New function.
369 (maybe_legitimize_operands): Try to reuse the results for
372 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
374 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
375 Add {q} suffix to insn mnemonic.
377 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
379 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
380 (msp430_warn_func_return): New.
382 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
384 * fold-const.c (tree_nonzero_bits): New function.
385 * fold-const.h (tree_nonzero_bits): Likewise.
386 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
387 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
389 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
393 * varasm.c (assemble_alias): Check ifunc_resolver only on
396 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
399 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
400 when memory input operand is handled.
402 2018-05-24 Luis Machado <luis.machado@linaro.org>
404 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
406 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
408 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
410 * match.pd: Delay FMA folds until after vectorization.
412 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
415 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
416 address check not strict.
418 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
420 * gimple-match.h (gimple_match_op): New class.
421 (mprts_hook): Replace parameters with a gimple_match_op *.
422 (maybe_build_generic_op): Likewise.
423 (gimple_simplified_result_is_gimple_val): Replace parameters with
424 a const gimple_match_op *.
425 (gimple_simplify): Replace code_helper * and tree * parameters with
426 a gimple_match_op * parameter.
427 (gimple_resimplify1): Replace code_helper *, tree and tree *
428 parameters with a gimple_match_op * parameter.
429 (gimple_resimplify2): Likewise.
430 (gimple_resimplify3): Likewise.
431 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
432 parameters with a gimple_match_op * parameter.
433 * gimple-match-head.c (gimple_simplify): Change prototypes of
434 auto-generated functions to take a gimple_match_op * instead of
435 separate code_helper * and tree * parameters. Make the same
436 change in the top-level overload and update calls to the
437 gimple_resimplify routines. Update calls to the auto-generated
438 functions and to maybe_push_res_to_seq in the publicly-facing
439 operation-specific gimple_simplify overloads.
440 (gimple_match_op::MAX_NUM_OPS): Define.
441 (gimple_resimplify1): Replace rcode and ops with a single res_op
442 parameter. Update call to gimple_simplify.
443 (gimple_resimplify2): Likewise.
444 (gimple_resimplify3): Likewise.
445 (mprts_hook): Replace parameters with a gimple_match_op *.
446 (maybe_build_generic_op): Likewise.
447 (build_call_internal): Replace type, nargs and ops with
449 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
450 with a single gimple_match_op *. Update calls to mprts_hook,
451 build_call_internal and gimple_simplified_result_is_gimple_val.
452 Factor out code that is common to the tree_code and combined_fn cases.
453 * genmatch.c (expr::gen_transform): Replace tem_code and
454 tem_ops with a gimple_match_op called tem_op. Update calls
455 to the gimple_resimplify functions and maybe_push_res_to_seq.
456 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
457 res_ops. Update call to the gimple_resimplify functions.
458 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
459 (decision_tree::gen): Make the functions take a gimple_match_op *
460 called res_op instead of separate res_code and res_ops parameters.
461 Update call accordingly.
462 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
463 and ops with a single res_op parameter. Update calls to
464 maybe_build_generic_op and maybe_push_res_to_seq.
465 (fold_stmt_1): Update calls to gimple_simplify and
466 replace_stmt_with_simplification.
467 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
468 and gimple_simplified_result_is_gimple_val.
469 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
471 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
472 with a gimple_match_op *.
473 (vn_nary_build_or_lookup): Likewise. Update call to
474 vn_nary_build_or_lookup_1.
475 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
476 gimple_match_op *. Update calls to the gimple_resimplify routines
477 and to gimple_simplified_result_is_gimple_val.
478 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
479 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
480 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
481 (visit_nary_op): Likewise.
482 (visit_reference_op_load): Likewise.
484 2018-05-23 Luis Machado <luis.machado@linaro.org>
486 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
487 modifier for printing the step amount.
489 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
492 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
495 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
497 * doc/sourcebuild.texi (Endianness): New subsubsection.
499 2018-05-23 Luis Machado <luis.machado@linaro.org>
501 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
502 <prefetch_dynamic_strides>: New const bool field.
503 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
504 prefetch_dynamic_strides.
505 (exynosm1_prefetch_tune): Likewise.
506 (thunderxt88_prefetch_tune): Likewise.
507 (thunderx_prefetch_tune): Likewise.
508 (thunderx2t99_prefetch_tune): Likewise.
509 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
511 (aarch64_override_options_internal): Update to set
512 PARAM_PREFETCH_DYNAMIC_STRIDES.
513 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
514 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
515 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
516 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
517 prefetch-dynamic-strides setting.
519 2018-05-23 Luis Machado <luis.machado@linaro.org>
521 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
522 <minimum_stride>: New const int field.
523 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
524 minimum_stride field defaulting to -1.
525 (exynosm1_prefetch_tune): Likewise.
526 (thunderxt88_prefetch_tune): Likewise.
527 (thunderx_prefetch_tune): Likewise.
528 (thunderx2t99_prefetch_tune): Likewise.
529 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
530 <default_opt_level>: Set to 3.
531 (aarch64_override_options_internal): Update to set
532 PARAM_PREFETCH_MINIMUM_STRIDE.
533 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
534 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
535 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
536 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
537 stride is constant and is below the minimum stride threshold.
539 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
541 * config/arm/arm-cpus.in (mode26): Delete.
542 (armv4): Delete mode26 reference.
543 * config/arm/arm.c (arm_configure_build_target): Delete use of
546 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
548 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
550 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
551 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
553 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
554 (floatunsdidf2): Ditto.
556 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
558 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
559 (fixuns_trunc<mode>si2_avx512f): Ditto.
560 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
561 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
562 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
564 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
566 PR rtl-optimization/79985
567 * df-scan.c (df_insn_refs_collect): Remove special case for
568 global registers and asm statements.
570 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
572 * extend.texi (Global Register Variables): Rewrite the bullet list.
573 Note that the register is available for allocation. Note that access
574 via inline asm must use constraints. Add note about async-signal
575 handlers. Remove paragraph about automagic register selection.
577 2018-05-23 Richard Biener <rguenther@suse.de>
579 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
580 of fixed offset from memset VN.
582 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
584 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
586 (alloc_cand_and_find_basis): Initialize first_interp field.
587 (slsr_process_mul): Modify first_interp field.
588 (slsr_process_add): Likewise.
589 (slsr_process_cast): Modify first_interp field for each new
591 (slsr_process_copy): Likewise.
592 (dump_candidate): Dump first_interp field.
593 (replace_mult_candidate): Process all interpretations, not just
595 (replace_rhs_if_not_dup): Likewise.
596 (replace_one_candidate): Likewise.
598 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
600 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
602 (aarch64_needs_frame_chain): New function.
603 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
605 2018-05-23 Sudakshina Das <sudi.das@arm.com>
608 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
609 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
610 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
611 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
612 as true for strict-align.
613 (aarch64_can_inline_p): Perform checks even when callee has no
614 attributes to check for strict alignment.
615 * doc/extend.texi (AArch64 Function Attributes): Document
617 * doc/invoke.texi: (AArch64 Options): Likewise.
619 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
621 PR tree-optimization/85853
622 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
623 the handling of the root of the node to...
624 (vect_slp_analyze_node_operations_1): ...this new function,
625 and run the whole thing with the child nodes' def types
626 set according to their SLP node's def type.
628 2018-05-23 Richard Biener <rguenther@suse.de>
631 * tree-data-ref.c (create_runtime_alias_checks): Defer
632 and ignore overflow warnings.
634 2018-05-23 Yury Gribov <tetra2005@gmail.com>
636 PR tree-optimization/85822
637 * tree-vrp.c (is_masked_range_test): Fix handling of negative
640 2018-05-23 Richard Biener <rguenther@suse.de>
642 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
643 memset constants via native_interpret_expr.
645 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
648 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
650 (cgraph_node::create_alias): Likewise.
651 (cgraph_node::get_availability): Check ifunc_resolver instead
652 of looking up ifunc attribute.
653 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
654 * varasm.c (do_assemble_alias): Likewise.
655 (assemble_alias): Likewise.
656 (default_binds_local_p_3): Likewise.
657 * cgraph.h (cgraph_node): Add ifunc_resolver.
658 (cgraph_node::only_called_directly_or_aliased_p): Return false
660 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
662 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
663 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
664 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
665 instead of looking up ifunc attribute.
667 2018-05-22 Luis Machado <luis.machado@linaro.org>
669 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
671 2018-05-22 Martin Sebor <msebor@redhat.com>
674 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
675 only when expasion succeeds.
676 (expand_builtin_strcmp): Same.
677 (expand_builtin_strncmp): Same.
679 2018-05-22 Martin Sebor <msebor@redhat.com>
681 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
683 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
684 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
686 * config/aarch64/aarch64-ldpstp.md: Replace uses of
687 aarch64_mem_pair_operand with memory_operand and delete operand swapping
689 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
690 Add check for legitimate_address.
691 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
692 (aarch64_swap_ldrstr_operands): New.
693 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
696 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
697 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
699 * config/aarch64/aarch64.md: New patterns to generate stp
701 (store_pair_sw, store_pair_dw): New patterns to generate stp for
702 single words and double words.
703 (load_pair_sw, load_pair_dw): Likewise.
704 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
706 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
708 * config/aarch64/aarch64-ldpstp.md: Modify peephole
709 for different mode ldpstp and add peephole for merged zero stores.
711 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
713 (aarch64_gen_store_pair): Rename calls to match new patterns.
714 (aarch64_gen_load_pair): Rename calls to match new patterns.
715 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
716 (load_pair<DREG:mode><DREG2:mode>): ... This.
717 (store_pair<mode>): Rename to...
718 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
719 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
721 (V_INT_EQUIV): Handle SImode.
722 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
725 2018-05-22 Martin Sebor <msebor@redhat.com>
728 * calls.c (maybe_warn_nonstring_arg): Use string length to set
729 or ajust the presumed bound on an operation to avoid unnecessary
732 2018-05-22 Martin Sebor <msebor@redhat.com>
734 PR tree-optimization/85826
735 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
736 assuming that a DECL necesarily has a constant size.
738 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
741 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
743 2018-05-22 Richard Biener <rguenther@suse.de>
745 PR tree-optimization/85834
746 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
747 non-constant and non-zero memset arguments.
749 2018-05-22 Martin Liska <mliska@suse.cz>
752 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
754 2018-05-22 Richard Biener <rguenther@suse.de>
756 PR tree-optimization/85863
757 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
758 comparisons when vectype is specified.
759 (vectorizable_condition): Do not specify vectype for
760 vect_is_simple_cond when SLP vectorizing.
762 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
765 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
766 define __ibm128 as long double.
767 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
768 as a distinct type when IEEE 128-bit support is enabled.
769 (init_float128_ieee): Fix up conversions between IFmode and IEEE
770 128-bit types to use the correct functions.
771 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
772 convert between 128-bit floating point types that have different
773 modes but the same representation, instead of using gen_lowpart to
775 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
777 (IFKF_reg): New attributes to give the register constraints for
779 (extend<mode>tf2_internal): New insns to mark an explicit
780 conversion between 128-bit floating point types that have a
781 different mode but share the same representation.
783 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
785 PR tree-optimization/85814
786 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
787 a null return from get_strinfo when unsharing the next
788 strinfo in the chain.
790 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
793 * varasm.c (weak_finish): Clean up weak_decls.
795 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
797 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
798 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
799 UNSPEC_UADALP values.
800 * config/aarch64/iterators.md (ABAL): New int iterator.
803 (sur): Add mappings for the above.
804 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
806 (aarch64_<sur>abal<mode>_4): Likewise.
807 (aarch64_<sur>adalp<mode>_3): Likewise.
808 (<sur>sadv16qi): New define_expand.
810 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
812 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
813 (*movdf_internal): Ditto.
814 (*rcpsf2_sse): Ditto.
815 (*rsqrtsf2_sse): Ditto.
816 (*sqrt<mode>2_sse): Ditto.
818 2018-05-21 Tamar Christina <tamar.christina@arm.com>
820 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
822 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
823 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
824 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
825 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
827 * config/aarch64/arm_neon.h: Likewise.
828 * config/aarch64/iterators.md (VQ_I): New.
830 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
832 * config.gcc: Add arc/t-multilib-linux to tmake_file for
834 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
837 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
839 * config/nds32/constraints.md (S): New constraint.
840 * config/nds32/nds32.md (call_internal): Use constraint S.
841 (call_value_internal): Likewise.
842 (sibcall_internal): Likewise.
843 (sibcall_value_internal): Likewise.
845 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
846 Chung-Ju Wu <jasonwucj@gmail.com>
848 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
851 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
852 Chung-Ju Wu <jasonwucj@gmail.com>
854 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
855 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
856 (nds32_rtx_costs_impl): Simplify.
857 (nds32_address_cost_impl): Simplify.
858 (nds32_init_rtx_costs): New function.
859 (nds32_rtx_costs_speed_prefer): Likewise.
860 (nds32_rtx_costs_size_prefer): Likewise.
861 (nds32_address_cost_speed_prefer): Likewise.
862 (nds32_address_cost_speed_fwprop): Likewise.
863 (nds32_address_cost_size_prefer): Likewise.
864 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
865 * config/nds32/nds32.c (nds32_option_override): Use
866 nds32_init_rtx_costs function.
868 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
870 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
871 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
872 (TARGET_PIPELINE_N8): Likewise.
873 (TARGET_PIPELINE_N10): Likewise.
874 (TARGET_PIPELINE_N13): Likewise.
875 (TARGET_PIPELINE_GRAYWOLF): Likewise.
877 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
879 * config/nds32/nds32-fpu.md: Update copyright year.
881 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
883 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
885 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
887 * config/nds32/nds32.c
888 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
889 * config/nds32/nds32.opt (minline-asm-r15): New option.
891 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
893 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
895 * config/nds32/nds32.md (abssi2): New pattern.
897 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
899 * config/i386/i386.md (rex64namesuffix): New mode attribute.
900 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
901 Merge insn pattern from sse_cvtsi2ss<round_name> and
902 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
903 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
904 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
905 using SWI48 mode iterator.
906 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
907 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
908 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
909 pattern from sse_cvttss2si<round_saeonly_name>
910 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
911 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
912 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
913 using SWI48 mode iterator.
914 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
915 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
916 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
917 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
918 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
919 using SWI48 mode iterator.
920 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
921 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
922 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
923 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
924 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
926 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
927 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
928 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
929 pattern from sse_cvttsd2si<round_saeonly_name>
930 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
932 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
934 * config/nds32/nds32-md-auxiliary.c
935 (nds32_valid_smw_lwm_base_p): Refine.
936 (nds32_output_smw_single_word): Refine.
937 (nds32_output_smw_double_word): New.
938 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
940 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
942 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
943 (nds32_output_stack_pop): Refine.
944 (nds32_expand_unaligned_load): Refine.
945 (nds32_expand_unaligned_store): Refine.
947 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
948 Chung-Ju Wu <jasonwucj@gmail.com>
950 * config/nds32/constants.md: Add TP_REGNUM constant.
951 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
952 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
954 * config/nds32/nds32-doubleword.md: Consider flag_pic.
955 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
956 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
957 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
958 and PIC code generation.
959 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
961 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
963 * config/nds32/nds32.md: Support TLS and PIC.
964 * config/nds32/nds32.c: Support TLS and PIC.
965 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
966 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
969 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
971 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
974 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
975 Chung-Ju Wu <jasonwucj@gmail.com>
977 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
978 * config/nds32/nds32-md-auxiliary.c
979 (symbolic_reference_mentioned_p): New.
980 (nds32_legitimize_ict_address): New.
981 (nds32_expand_ict_move): New.
982 (nds32_indirect_call_referenced_p): New.
983 (nds32_symbol_binds_local_p): Delete.
984 (nds32_long_call_p): Modify.
985 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
986 * config/nds32/nds32-protos.h
987 (symbolic_reference_mentioned_p): Declare.
988 (nds32_legitimize_ict_address): Declare.
989 (nds32_expand_ict_move): Declare.
990 (nds32_indirect_call_referenced_p): Declare.
991 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
992 (nds32_relax_group): Use nds32_ict_const_p as condition.
993 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
994 (nds32_asm_file_start): Output ict_model directive in asm code.
995 (nds32_legitimate_address_p): Consider indirect call.
996 (nds32_print_operand): Consider indirect call.
997 (nds32_print_operand_address): Consider indirect call.
998 (nds32_insert_attributes): Handle "indirect_call" attribute.
999 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1000 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1001 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1002 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1003 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1004 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1005 (TARGET_ICT_MODEL_SMALL): Define.
1006 (TARGET_ICT_MODEL_LARGE): Define.
1007 * config/nds32/nds32.md (movsi): Consider ict model.
1008 (call, call_value): Consider ict model.
1009 (sibcall, sibcall_value): Consider ict model.
1010 * config/nds32/nds32.opt (mict-model): New option.
1011 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1014 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1015 Monk Chiang <sh.chiang04@gmail.com>
1016 Jim Wilson <jimw@sifive.com>
1018 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1019 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1020 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1021 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1022 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1023 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1024 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1025 compute save_libcall_adjustment properly.
1026 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1027 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1028 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1029 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1030 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1031 (ABI_SPEC): Handle mabi=ilp32e.
1032 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1033 (RVE): Add RVE mask.
1034 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1035 <-march>: Add rv32e as an example.
1037 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1040 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1041 (intra_create_variable_infos): Handle C++ constructors.
1043 2018-05-18 Martin Liska <mliska@suse.cz>
1045 * passes.def: Remove a redundant pass.
1047 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1050 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1052 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1054 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1056 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1058 (armv2, armv2a, armv3, armv3m): Delete architectures.
1059 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1060 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1061 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1063 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1064 (*mulsidi3adddi): Likewise.
1065 (mulsidi3): Likewise.
1066 (*mulsidi3_nov6): Likewise.
1067 (umulsidi3): Likewise.
1068 (umulsidi3_nov6): Likewise.
1069 (umaddsidi4): Likewise.
1070 (*umulsidi3adddi): Likewise.
1071 (smulsi3_highpart): Likewise.
1072 (*smulsi3_highpart_nov6): Likewise.
1073 (umulsi3_highpart): Likewise.
1074 (*umulsi3_highpart_nov6): Likewise.
1075 * config/arm/arm.h (arm_arch3m): Delete.
1076 * config/arm/arm.c (arm_arch3m): Delete.
1077 (arm_option_override_internal): Update armv3-related comment.
1078 (arm_configure_build_target): Delete use of isa_bit_mode32.
1079 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1080 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1081 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1083 (mulusa3): Likewise.
1084 * config/arm/arm-protos.h (arm_arch3m): Delete.
1085 * config/arm/arm-tables.opt: Regenerate.
1086 * config/arm/arm-tune.md: Likewise.
1087 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1088 deleted architectures.
1090 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1092 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1093 (armv5t, armv5te): New features.
1094 (ARMv5, ARMv5e): Delete fgroups.
1095 (ARMv5t, ARMv5te): Adjust for above changes.
1097 (armv5, armv5e): Delete arches.
1098 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1100 (*call_reg_arm): Likewise.
1101 (*call_value_reg_armv5): Likewise.
1102 (*call_value_reg_arm): Likewise.
1103 (*call_symbol): Likewise.
1104 (*call_value_symbol): Likewise.
1105 (*sibcall_insn): Likewise.
1106 (*sibcall_value_insn): Likewise.
1108 (prefetch): Likewise.
1109 (define_split and define_peephole2 dependent on arm_arch5):
1111 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1113 (TARGET_ARM_QBIT): Likewise.
1114 (TARGET_DSP_MULTIPLY): Likewise.
1115 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1116 (arm_arch5, arm_arch5e): Delete.
1117 (arm_arch5t, arm_arch5te): Declare.
1118 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1119 (arm_arch5t): Declare.
1120 (arm_option_reconfigure_globals): Update for the above.
1121 (arm_options_perform_arch_sanity_checks): Update comment, replace
1122 use of arm_arch5 with arm_arch5t.
1123 (use_return_insn): Likewise.
1124 (arm_emit_call_insn): Likewise.
1125 (output_return_instruction): Likewise.
1126 (arm_final_prescan_insn): Likewise.
1127 (arm_coproc_builtin_available): Likewise.
1128 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1129 arm_arch5e with arm_arch5t and arm_arch5te.
1130 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1131 (arm_arch5t, arm_arch5te): Declare.
1132 * config/arm/arm-tables.opt: Regenerate.
1133 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1134 * config/arm/t-multilib: Likewise.
1135 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1136 instead of arm_arch5.
1137 (*call_reg_thumb1): Likewise.
1138 (*call_value_reg_thumb1_v5): Likewise.
1139 (*call_value_reg_thumb1): Likewise.
1140 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1142 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1144 2018-05-18 Martin Liska <mliska@suse.cz>
1146 PR gcov-profile/84846
1147 * doc/gcov.texi: Document -t option of gcov tool.
1149 2018-05-18 Martin Liska <mliska@suse.cz>
1151 PR gcov-profile/84846
1152 * gcov.c (print_usage): Add new -t option.
1153 (process_args): Handle the option.
1154 (generate_results): Use stdout as output when requested by
1157 2018-05-18 Martin Liska <mliska@suse.cz>
1159 PR gcov-profile/84846
1160 * coverage.c (coverage_init): Write PWD to .gcno file.
1161 * doc/gcov.texi: Document how working directory is printed.
1162 * gcov-dump.c (dump_gcov_file): Print PWD.
1163 * gcov.c (output_intermediate_file): Likewise.
1164 (read_graph_file): Read PWD string.
1165 (output_lines): Print PWD.
1167 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1170 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1171 for retval and return false if all args to phi are zero.
1173 2018-05-18 Richard Biener <rguenther@suse.de>
1175 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1177 (evrp_dom_walker::before_dom_children): Call it.
1179 2018-05-18 Richard Biener <rguenther@suse.de>
1181 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1182 results when processing array refs with variable index.
1184 2018-05-18 Toon Moene <toon@moene.org>
1186 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1187 directly after that of -floop-interchange. Indicate that both
1188 options are enabled by default when specifying -O3.
1190 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1192 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1193 iterator. Delete separate integer-mode vec_set<mode> expander.
1194 (aarch64_simd_vec_setv2di): Delete.
1195 (vec_setv2di): Delete.
1196 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1197 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1198 the "w, r" alternative.
1200 2018-05-18 Martin Liska <mliska@suse.cz>
1202 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1203 * tree-pass.h (make_pass_lower_switch_O0): New function.
1204 * tree-switch-conversion.c (node_has_low_bound): Remove.
1205 (node_has_high_bound): Likewise.
1206 (node_is_bounded): Likewise.
1207 (class pass_lower_switch): Make it a template type and create
1209 (pass_lower_switch::execute): Add template argument.
1210 (make_pass_lower_switch): New function.
1211 (make_pass_lower_switch_O0): New function.
1212 (do_jump_if_equal): Remove.
1213 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1214 all the hard work to tree optimization passes.
1216 2018-05-18 Martin Liska <mliska@suse.cz>
1218 * dbgcnt.c (limit_low): Renamed from limit.
1219 (limit_high): New variable.
1220 (dbg_cnt_is_enabled): Check for upper limit.
1221 (dbg_cnt): Adjust dumping.
1222 (dbg_cnt_set_limit_by_index): Add new argument for high
1224 (dbg_cnt_set_limit_by_name): Likewise.
1225 (dbg_cnt_process_single_pair): Parse new format.
1226 (dbg_cnt_process_opt): Use strtok.
1227 (dbg_cnt_list_all_counters): Remove 'value' and add
1229 * doc/invoke.texi: Document changes.
1231 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1233 * doc/sourcebuild.texi (scalar_all_fma): Document.
1234 * tree.def (FMA_EXPR): Delete.
1235 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1236 * internal-fn.c (ternary_direct): New macro.
1237 (expand_ternary_optab_fn): Likewise.
1238 (direct_ternary_optab_supported_p): Likewise.
1239 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1240 * builtins.c (fold_builtin_fma): Delete.
1241 (fold_builtin_3): Don't call it.
1242 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1243 * expr.c (expand_expr_real_2): Likewise.
1244 * fold-const.c (operand_equal_p): Likewise.
1245 (fold_ternary_loc): Likewise.
1246 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1247 * gimple.c (DEFTREECODE): Likewise.
1248 * gimplify.c (gimplify_expr): Likewise.
1249 * optabs-tree.c (optab_for_tree_code): Likewise.
1250 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1251 * tree-eh.c (operation_could_trap_p): Likewise.
1252 (stmt_could_throw_1_p): Likewise.
1253 * tree-inline.c (estimate_operator_cost): Likewise.
1254 * tree-pretty-print.c (dump_generic_node): Likewise.
1255 (op_code_prio): Likewise.
1256 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1257 * tree-ssa-operands.c (get_expr_operands): Likewise.
1258 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1259 * fold-const-call.h (fold_fma): Delete.
1260 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1261 CFN_FNMA and CFN_FNMS.
1263 * genmatch.c (combined_fn): New enum.
1264 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1265 (commutative_op): New function.
1266 (commutate): Use it. Handle more than 2 operands.
1267 (dt_operand::gen_gimple_expr): Use commutative_op.
1268 (parser::parse_expr): Allow :c to be used with non-binary
1269 operators if the commutative operand is known.
1270 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1271 CFN_FMS, CFN_FNMA and CFN_FNMS.
1272 (backprop::process_assign_use): Remove FMA_EXPR handling.
1273 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1274 (gen_hsa_fma): New function.
1275 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1276 IFN_FNMA and IFN_FNMS.
1277 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1278 * gimple-fold.h (follow_all_ssa_edges): Declare.
1279 * gimple-fold.c (follow_all_ssa_edges): New function.
1280 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1281 gimple_build interface and use follow_all_ssa_edges to fold the result.
1282 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1283 instead of checking for optabs directly.
1284 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1285 rather than FMA_EXPRs.
1286 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1287 call to IFN_FMA instead of an FMA_EXPR.
1289 2018-05-17 Jim Wilson <jimw@sifive.com>
1291 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1292 sign extended promoted subreg, and the range does not have the sign bit
1293 set, then do a sign extend.
1295 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1296 test, check for sign extended subreg and/or constant operands, and
1297 do a sign extend in that case.
1299 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1301 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1302 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1304 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1305 Change logics_shift_reg to logics_shift_imm.
1306 (thunderx2t99_fp_loadpair_basic): Delete.
1307 (thunderx2t99_fp_storepair_basic): Delete.
1308 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1309 (thunderx2t99_asimd_polynomial): Delete.
1310 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1311 and neon_fp_mul_d_scalar_q.
1312 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1313 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1314 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1315 (thunderx2t99_asimd_lut): Add missing tbl types.
1316 (thunderx2t99_asimd_ext): Delete.
1317 (thunderx2t99_asimd_load1_1_mult): Delete.
1318 (thunderx2t99_asimd_load1_2_mult): Delete.
1319 (thunderx2t99_asimd_load1_ldp): New.
1320 (thunderx2t99_asimd_load1): New.
1321 (thunderx2t99_asimd_load2): Add missing *load2* types.
1322 (thunderx2t99_asimd_load3): New.
1323 (thunderx2t99_asimd_load4): New.
1324 (thunderx2t99_asimd_store1_1_mult): Delete.
1325 (thunderx2t99_asimd_store1_2_mult): Delete.
1326 (thunderx2t99_asimd_store2_mult): Delete.
1327 (thunderx2t99_asimd_store2_onelane): Delete.
1328 (thunderx2t99_asimd_store_stp): New.
1329 (thunderx2t99_asimd_store1): New.
1330 (thunderx2t99_asimd_store2): New.
1331 (thunderx2t99_asimd_store3): New.
1332 (thunderx2t99_asimd_store4): New.
1334 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1336 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1337 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1339 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1340 Segher Boessenkool <segher@kernel.crashing.org>
1343 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1346 2018-05-17 Richard Biener <rguenther@suse.de>
1348 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1349 for pruning loop and prune defs feeding only already visited PHIs.
1351 2018-05-17 Richard Biener <rguenther@suse.de>
1353 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1355 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1356 Richard Biener <rguenther@suse.de>
1358 PR tree-optimization/85793
1359 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1360 for VMAT_ELEMENTWISE.
1362 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1364 * internal-fn.h (lookup_internal_fn): Declare
1365 * internal-fn.c (lookup_internal_fn): New function.
1366 * gimple.c (gimple_build_call_from_tree): Handle calls to
1368 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1369 internal function names.
1370 * tree-pretty-print.c (dump_generic_node): Likewise.
1371 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1373 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1375 * gimple-fold.h (gimple_build): Make the function forms take
1376 combined_fn rather than built_in_function.
1377 (gimple_simplify): Likewise.
1378 * gimple-match-head.c (gimple_simplify): Likewise.
1379 * gimple-fold.c (gimple_build): Likewise.
1380 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1381 rather than gimple_build_call_internal.
1382 (get_initial_defs_for_reduction): Likewise.
1383 (vect_create_epilog_for_reduction): Likewise.
1384 (vectorizable_live_operation): Likewise.
1386 2018-05-17 Martin Liska <mliska@suse.cz>
1388 * gimple-ssa-sprintf.c (format_directive): Do not use
1389 space in between 'G_' and '('.
1391 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1394 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1395 even if the mask is not all ones.
1398 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1400 (ix86_gimple_fold_builtin): Likewise.
1403 * config/i386/i386.c: Include tree-vector-builder.h.
1404 (ix86_vector_shift_count): New function.
1405 (ix86_fold_builtin): Fold shift builtins by scalar count.
1406 (ix86_gimple_fold_builtin): Likewise.
1408 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1409 _mm512_setzero): New intrinsics.
1411 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1412 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1414 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1415 code generation for cases where splatting a value is not useful.
1416 * simplify-rtx.c (simplify_ternary_operation): Simplify
1417 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1418 mode to a vec_concat.
1420 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1422 * config.gcc: Support "goldmont-plus".
1423 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1425 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1426 PROCESSOR_GOLDMONT_PLUS.
1427 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1428 (processor_target_table): Add "goldmont-plus".
1429 (PTA_GOLDMONT_PLUS): Define.
1430 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1431 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1432 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1433 (fold_builtin_cpu): Add "goldmont-plus".
1434 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1435 (ix86_option_override_internal): Add "goldmont-plus".
1436 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1437 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1438 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1439 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1441 2018-05-17 Richard Biener <rguenther@suse.de>
1443 PR tree-optimization/85757
1444 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1445 remove defs that only feed that PHI from further processing.
1447 2018-05-16 Jim Wilson <jimw@sifive.com>
1449 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1451 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1453 2018-05-16 Mark Wielaard <mark@klomp.org>
1455 * dwarf2out.c (count_index_strings): New function.
1456 (output_indirect_strings): Call count_index_strings and generate
1457 header for dwarf_version >= 5.
1459 2018-05-16 Mark Wielaard <mark@klomp.org>
1461 * dwarf2out.c (dwarf_FORM): New function.
1462 (set_indirect_string): Use dwarf_FORM.
1463 (reset_indirect_string): Likewise.
1464 (size_of_die): Likewise.
1465 (value_format): Likewise.
1466 (output_die): Likewise.
1467 (add_skeleton_AT_string): Likewise.
1468 (output_macinfo_op): Likewise.
1469 (index_string): Likewise.
1470 (output_index_string_offset): Likewise.
1471 (output_index_string): Likewise.
1472 (count_index_strings): Likewise.
1474 2018-05-16 Carl Love <cel@us.ibm.com>
1476 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1477 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1479 2018-05-16 Martin Jambor <mjambor@suse.cz>
1481 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1482 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1484 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1486 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1487 (fnma<mode>4): Likewise.
1488 (fms<mode>4): Likewise.
1489 (fnms<mode>4): Likewise.
1490 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1491 (aarch64_fnma<mode>4): Likewise.
1492 (aarch64_fms<mode>4): Likewise.
1493 (aarch64_fnms<mode>4): Likewise.
1494 (aarch64_fnmadd<mode>4): Likewise.
1496 2018-05-16 Jason Merrill <jason@redhat.com>
1498 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1500 2018-05-16 Richard Biener <rguenther@suse.de>
1502 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1503 (dump_stmt_cost): Declare.
1504 (add_stmt_cost): Dump cost we add.
1505 (add_stmt_costs): New function.
1506 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1508 (vect_analyze_stmt): Adjust prototype.
1509 (vectorizable_condition): Likewise.
1510 (vectorizable_live_operation): Likewise.
1511 (vectorizable_reduction): Likewise.
1512 (vectorizable_induction): Likewise.
1513 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1514 cost vector to pass to vectorizable_ and record afterwards.
1515 (vect_model_reduction_cost): Take cost vector argument and adjust.
1516 (vect_model_induction_cost): Likewise.
1517 (vectorizable_reduction): Likewise.
1518 (vectorizable_induction): Likewise.
1519 (vectorizable_live_operation): Likewise.
1520 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1521 SLP_TREE_NUMBER_OF_VEC_STMTS.
1522 (vect_analyze_slp_cost_1): Remove.
1523 (vect_analyze_slp_cost): Likewise.
1524 (vect_slp_analyze_node_operations): Take visited args and
1525 a target cost vector. Avoid processing already visited stmt sets.
1526 (vect_slp_analyze_operations): Use a local cost vector to gather
1527 costs and register those of non-discarded instances.
1528 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1529 (vect_schedule_slp_instance): Remove copying of
1530 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1532 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1533 adding cost. Record cost entry location.
1534 (vect_prologue_cost_for_slp_op): Function to compute cost of
1535 a constant or invariant generated for SLP vect in the prologue,
1536 split out from vect_analyze_slp_cost_1.
1537 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1538 (vect_model_promotion_demotion_cost): Likewise.
1539 (vect_model_store_cost): Likewise, make static.
1540 (vect_model_load_cost): Likewise.
1541 (vectorizable_bswap): Add cost vector arg and adjust.
1542 (vectorizable_call): Likewise.
1543 (vectorizable_simd_clone_call): Likewise.
1544 (vectorizable_conversion): Likewise.
1545 (vectorizable_assignment): Likewise.
1546 (vectorizable_shift): Likewise.
1547 (vectorizable_operation): Likewise.
1548 (vectorizable_store): Likewise.
1549 (vectorizable_load): Likewise.
1550 (vectorizable_condition): Likewise.
1551 (vectorizable_comparison): Likewise.
1552 (can_vectorize_live_stmts): Likewise.
1553 (vect_analyze_stmt): Likewise.
1554 (vect_transform_stmt): Adjust calls to vectorizable_*.
1555 * tree-vectorizer.c: Include gimple-pretty-print.h.
1556 (dump_stmt_cost): New function.
1558 2018-05-16 Richard Biener <rguenther@suse.de>
1560 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1561 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1562 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1563 (check_name): New callback.
1564 (dse_classify_store): Track cycles via a visited bitmap of PHI
1565 defs and simplify handling of in-loop and across loop dead stores
1566 and properly fail for loop-variant refs. Handle byte-tracking with
1567 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1570 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1572 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1573 (vect_get_mask_type_for_stmt): Likewise.
1574 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1576 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1577 to determine the statement's vector type and the vector type that
1578 should be used for calculating nunits. Deal with cases in which
1579 the type has to be deferred.
1580 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1581 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1582 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1583 (vect_determine_vf_for_stmt): New functions, split out from...
1584 (vect_determine_vectorization_factor): ...here.
1585 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1586 (vect_get_mask_type_for_stmt): New functions, split out from
1587 vect_determine_vectorization_factor.
1589 2018-05-16 Richard Biener <rguenther@suse.de>
1591 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1592 verify the [VEC_]COND_EXPR embedded comparison.
1594 2018-05-15 Martin Sebor <msebor@redhat.com>
1596 PR tree-optimization/85753
1597 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1598 RECORD_TYPE in addition to ARRAY_TYPE.
1600 2018-05-15 Martin Sebor <msebor@redhat.com>
1603 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1605 2018-05-15 Richard Biener <rguenther@suse.de>
1607 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1608 add by_clobber_p one. Change algorithm to collect all defs
1609 representing uses we need to walk and try reducing them to
1610 a single one before failing.
1611 (dse_dom_walker::dse_optimize_stmt): Adjust.
1613 2018-05-13 Mark Wielaard <mark@klomp.org>
1615 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1616 (size_of_loc_descr): Likewise.
1617 (output_loc_operands): Likewise.
1618 (output_loc_operands_raw): Likewise.
1619 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1620 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1621 (hash_loc_operands): Likewise.
1622 (compare_loc_operands): Likewise.
1624 2018-05-14 Mark Wielaard <mark@klomp.org>
1626 * dwarf2out.c (count_index_addrs): New function.
1627 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1629 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1631 PR tree-optimization/83648
1632 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1633 return value as malloc candidate.
1635 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1638 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1639 as true in call to suggest_attribute.
1641 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1643 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1644 -mreadonly-in-sdata.
1646 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1648 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1650 (aarch64_crypto_aesd_fused): Likewise.
1652 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1654 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1655 (movsi_aarch64): Likewise.
1656 (load_pairsi): Likewise.
1657 (load_pairdi): Likewise.
1658 (store_pairsi): Likewise.
1659 (store_pairdi): Likewise.
1660 (load_pairsf): Likewise.
1661 (load_pairdf): Likewise.
1662 (store_pairsf): Likewise.
1663 (store_pairdf): Likewise.
1664 (zero_extend): Likewise.
1665 (trunc): Swap alternatives.
1666 (fcvt_target): Add '?' to prefer w over r.
1668 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1671 * config/i386/i386.md: Disallow non-commutative arithmetics in
1672 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1673 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1674 in the peephole2 before it.
1676 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1678 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1679 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1680 (ix86_handle_option): Handle -mcldemote.
1681 * config.gcc: New header.
1682 * config/i386/cldemoteintrin.h: New file.
1683 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1684 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1686 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1687 OPTION_MASK_ISA_CLDEMOTE.
1688 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1689 (ix86_valid_target_attribute_inner_p): Ditto.
1690 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1691 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1692 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1693 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1694 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1696 * config/i386/i386.opt: Add -mcldemote.
1697 * config/i386/x86intrin.h: New header.
1698 * doc/invoke.texi: Add -mcldemote.
1700 2018-05-14 Richard Biener <rguenther@suse.de>
1702 * doc/match-and-simplify.texi: Adjust :s documentation.
1704 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1706 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1707 intended memcpy size.
1708 (REORDER_45): Likewise.
1710 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1712 * sort.cc: New file.
1713 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1714 * vec.c (qsort_chk): Use gcc_qsort.
1715 * Makefile.in (OBJS-libcommon): Add sort.o.
1716 (build/sort.o): New target. Use it...
1717 (BUILD_RTL): ... here, and...
1718 (build/gencfn-macros): ... here, and...
1719 (build/genmatch): ... here.
1721 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1722 Chung-Ju Wu <jasonwucj@gmail.com>
1724 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1725 * config/nds32/nds32-graywolf.md: New file.
1726 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1727 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1729 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1730 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1731 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1732 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1733 * config/nds32/pipelines.md: Include n15 settings.
1735 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1736 Chung-Ju Wu <jasonwucj@gmail.com>
1738 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1739 * config/nds32/nds32-n13.md: New file.
1740 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1741 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1743 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1744 * config/nds32/nds32.md (pipeline_model): Add n13.
1745 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1746 * config/nds32/pipelines.md: Include n13 settings.
1748 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1749 Chung-Ju Wu <jasonwucj@gmail.com>
1751 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1752 * config/nds32/nds32-n10.md: New file.
1753 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1754 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1756 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1757 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1758 * config/nds32/nds32.md (pipeline_model): Add n10.
1759 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1760 * config/nds32/pipelines.md: Include n10 settings.
1762 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1763 Kito Cheng <kito.cheng@gmail.com>
1764 Chung-Ju Wu <jasonwucj@gmail.com>
1766 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1767 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1768 Add enum values for DSP extension instructions.
1769 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1771 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1772 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1774 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1775 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1776 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1777 * config/nds32/nds32-intrinsic.md: Likewise.
1778 * config/nds32/nds32_intrinsic.h: Likewise.
1779 * config/nds32/nds32-md-auxiliary.c: Likewise.
1780 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1781 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1782 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1783 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1784 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1785 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1786 TYPE_DMAC in switch statement.
1787 * config/nds32/nds32.c: New checking and implementation for DSP
1788 extension instructions.
1789 * config/nds32/nds32.h: Likewise.
1790 * config/nds32/nds32.md: Likewise.
1791 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1792 * config/nds32/predicates.md: Implement new predicates for DSP
1795 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1797 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1798 Reformat alternatives and attributes so it is easier to identify
1799 which constraints/attributes go with which instruction.
1800 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1801 (mov<mode>_softfloat32, FMOVE64): Likewise.
1802 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1803 (mov<mode>_softfloat64, FMOVE64): Likewise.
1805 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1807 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1809 (Basic PowerPC Built-in Functions): The new name of the
1810 subsection previously known as "PowerPC Built-in Functions".
1811 (Basic PowerPC Built-in Functions Available on all Configurations):
1813 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1814 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1815 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1816 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1818 2018-05-11 Martin Jambor <mjambor@suse.cz>
1821 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1824 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1827 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1829 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1831 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1832 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1833 (ix86_handle_option): Handle -mwaitpkg.
1834 * config.gcc: New header.
1835 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1836 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1837 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1839 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1840 OPTION_MASK_ISA_WAITPKG.
1841 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1842 (ix86_option_override_internal): Add PTA_WAITPKG.
1843 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1844 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1845 IX86_BUILTIN_TPAUSE.
1846 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1847 __builtin_ia32_umwait and __builtin_ia32_tpause.
1848 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1849 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1850 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1851 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1852 UNSPECV_TPAUSE): New.
1853 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1854 * config/i386/i386.opt: Add -mwaitpkg.
1855 * config/i386/waitpkgintrin.h: New file.
1856 * config/i386/x86intrin.h: New header.
1857 * doc/invoke.texi: Add -mwaitpkg.
1859 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1862 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1864 (cortex-m0): Use armv6s-m isa.
1865 (cortex-m0plus): Likewise.
1866 (cortex-m1): Likewise.
1867 (cortex-m0.small-multiply): Likewise.
1868 (cortex-m0plus.small-multiply): Likewise.
1869 (cortex-m1.small-multiply): Likewise.
1871 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1872 Jakub Jelinek <jakub@redhat.com>
1874 PR tree-optimization/85692
1875 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1876 source permute as well.
1878 2018-05-11 Martin Liska <mliska@suse.cz>
1881 * doc/extend.texi: Document LLVM style format for no_sanitize
1884 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1886 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1887 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1888 (mode_supports_vsx_dform_quad): Likewise.
1889 (mode_supports_vmx_dform): Move these functions to be next to the
1890 other mode_supports functions.
1891 (mode_supports_dq_form): Likewise.
1892 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1893 mode_supports_dq_form.
1894 (reg_offset_addressing_ok_p): Likewise.
1895 (offsettable_ok_by_alignment): Likewise.
1896 (rs6000_legitimate_offset_address_p): Likewise.
1897 (legitimate_lo_sum_address_p): Likewise.
1898 (rs6000_legitimize_address): Likewise.
1899 (rs6000_legitimize_reload_address): Likewise.
1900 (rs6000_secondary_reload_inner): Likewise.
1901 (rs6000_preferred_reload_class): Likewise.
1902 (rs6000_output_move_128bit): Likewise.
1904 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1906 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1907 Generate SImode target register for null target.
1908 <case IX86_BUILTIN_XGETBV>: Ditto.
1909 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1910 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1912 2018-05-10 Carl Love <cel@us.ibm.com>
1914 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1915 dcbtt and dcbtstt if operands[2] is 0.
1917 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1920 * config/i386/sse.md (usadv64qi): New expander.
1922 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1924 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1925 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1926 -maltivec=be support.
1927 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1928 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1929 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1930 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1931 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1932 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1933 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1934 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1935 altivec_vsumsws): Adjust.
1936 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1937 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1938 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1940 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1941 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1942 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1943 (altivec_lve<VI_char>x): Delete expand.
1944 (*altivec_lve<VI_char>x_internal): Rename to...
1945 (altivec_lve<VI_char>x): ... this.
1946 (altivec_lvxl_<mode>): Delete expand.
1947 (*altivec_lvxl_<mode>_internal): Rename to ...
1948 (altivec_lvxl_<mode>): ... this.
1949 (altivec_stvxl_<mode>): Delete expand.
1950 (*altivec_stvxl_<mode>_internal): Rename to ...
1951 (altivec_stvxl_<mode>): ... this.
1952 (altivec_stve<VI_char>x): Delete expand.
1953 (*altivec_stve<VI_char>x_internal): Rename to ...
1954 (altivec_stve<VI_char>x): ... this.
1955 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1956 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1957 reduc_plus_scal_<mode>): Adjust.
1958 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1960 (rs6000_cpu_cpp_builtins): Adjust.
1961 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1962 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1963 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1964 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1965 -maltivec=be support.
1966 (rs6000_split_vec_extract_var): Adjust.
1967 (rs6000_split_v4si_init): Adjust.
1968 (swap_selector_for_mode): Delete.
1969 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1970 altivec_expand_stvex_be): Delete.
1971 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1972 -maltivec=be support.
1973 (rs6000_gimple_fold_builtin): Ditto.
1974 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1976 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1977 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1978 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1979 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1980 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1981 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1982 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1983 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1984 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1985 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1986 anonymous split): Adjust.
1987 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1988 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1990 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1992 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1993 when --with-gxx-include-dir is also specified.
1994 * configure: Regenerate.
1996 2018-05-09 Jim Wilson <jimw@sifive.com>
1999 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2000 * config/riscv/t-withmultilib: New.
2001 * config/riscv/withmultilib.h: New.
2002 * doc/install.texi: Document RISC-V --with-multilib-list support.
2004 2018-05-09 Richard Biener <rguenther@suse.de>
2006 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2008 (vect_bb_vectorization_profitable_p): Adjust. Compute
2009 actual scalar cost using the cost vector and the add_stmt_cost
2012 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2014 PR rtl-optimization/85645
2015 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2016 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2018 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2020 PR rtl-optimization/85645
2021 * shrink-wrap.c (spread_components): Return a boolean saying if
2022 anything was changed.
2023 (try_shrink_wrapping_separate): Iterate spread_components until
2024 nothing changes anymore.
2026 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2028 PR rtl-optimization/85645
2029 * regrename.c (build_def_use): Also kill the chains that include the
2030 destination of a REG_CFA_REGISTER note.
2032 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2034 PR rtl-optimization/85645
2035 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2036 insn that has a REG_CFA_REGISTER note.
2038 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2040 * cfgexpand.c (expand_clobber): New function.
2041 (expand_gimple_stmt_1): Use it.
2042 * tree-vect-stmts.c (vect_clobber_variable): New function,
2044 (vectorizable_simd_clone_call): ...here.
2045 (vectorizable_store): Emit a clobber either side of an
2046 IFN_STORE_LANES sequence.
2047 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2049 2018-05-09 Tom de Vries <tom@codesourcery.com>
2052 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2053 (define_insn "trap_if_false"): Add exit after trap.
2055 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2057 PR rtl-optimization/85638
2058 * bb-reorder.c: Include common/common-target.h.
2059 (create_forwarder_block): New function extracted from...
2060 (fix_up_crossing_landing_pad): ...here. Rename into...
2061 (dw2_fix_up_crossing_landing_pad): ...this.
2062 (sjlj_fix_up_crossing_landing_pad): New function.
2063 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2064 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2065 from both partitions and exit the loop after one iteration.
2067 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2070 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2072 (Basic PowerPC Built-in Functions): The new name of the
2073 subsection previously known as "PowerPC Built-in Functions".
2074 (Basic PowerPC Built-in Functions Available on all Configurations):
2076 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2078 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2080 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2082 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2085 2018-05-08 Jim Wilson <jimw@sifive.com>
2087 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2088 (LD_EMUL_SUFFIX): New.
2089 (LINK_SPEC): Use it.
2091 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2093 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2095 (Basic PowerPC Built-in Functions): The new name of the
2096 subsection previously known as "PowerPC Built-in Functions".
2097 (Basic PowerPC Built-in Functions Available on all Configurations):
2099 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2101 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2103 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2105 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2108 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2111 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2112 after cmpelim optimization.
2114 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2116 * config.gcc: Support "goldmont".
2117 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2118 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2120 * config/i386/i386.c (m_GOLDMONT): Define.
2121 (processor_target_table): Add "goldmont".
2122 (PTA_GOLDMONT): Define.
2123 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2124 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2125 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2126 (fold_builtin_cpu): Add "goldmont".
2127 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2128 (ix86_option_override_internal): Add "goldmont".
2129 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2130 (processor_type): Add PROCESSOR_GOLDMONT.
2131 * config/i386/i386.md: Add CPU "glm".
2132 * config/i386/glm.md: New file.
2133 * config/i386/x86-tune.def: Add m_GOLDMONT.
2134 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2136 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2139 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2141 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2142 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2143 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2146 * config/i386/i386.c (ix86_fold_builtin): Handle
2147 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2150 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2151 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2153 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2156 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2157 (check_arch): Likewise.
2158 (check_fpu): Return the result rather than printing it.
2159 (end arch): Fix operator precedence.
2160 (end cpu): Likewise.
2161 (END): Print the result from check_fpu.
2163 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2164 Alan Hayward <alan.hayward@arm.com>
2165 David Sherwood <david.sherwood@arm.com>
2167 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2168 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2169 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2170 (*fcmuo<mode>_and): New patterns.
2172 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2174 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2175 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2176 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2177 (cmp_op, sve_imm_con): New code attributes.
2178 (SVE_COND_INT_CMP, imm_con): Delete.
2179 (cmp_op): Remove above unspecs from int attribute.
2180 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2182 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2183 comparison-specific unspecs.
2184 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2185 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2186 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2187 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2188 (*vec_fcm<cmp_op><mode>): Rename to...
2189 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2190 (*vec_fcmuo<mode>): Rename to...
2191 (*fcmuo<mode>): ...this and adjust likewise.
2192 (*pred_fcm<cmp_op><mode>): New pattern.
2193 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2194 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2196 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2198 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2199 (aarch64_emit_sve_predicated_cond): New function.
2200 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2201 (aarch64_emit_unspec_cond_or): Replace with...
2202 (aarch64_emit_sve_or_conds): ...this new function. Use
2203 aarch64_emit_sve_ptrue_op for the individual comparisons and
2204 aarch64_emit_binop to OR them together.
2205 (aarch64_emit_inverted_unspec_cond): Replace with...
2206 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2207 aarch64_emit_sve_ptrue_op for the comparison and
2208 aarch64_emit_unop to invert the result.
2209 (aarch64_expand_sve_vec_cmp_float): Update after the above
2210 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2212 2018-05-07 Nathan Sidwell <nathan@acm.org>
2214 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2215 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2216 (Backwards Compatibility): Likewise.
2218 2018-05-07 Luis Machado <luis.machado@linaro.org>
2222 2018-05-07 Luis Machado <luis.machado@linaro.org>
2224 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2225 <prefetch_dynamic_strides>: New const bool field.
2226 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2227 prefetch_dynamic_strides.
2228 (exynosm1_prefetch_tune): Likewise.
2229 (thunderxt88_prefetch_tune): Likewise.
2230 (thunderx_prefetch_tune): Likewise.
2231 (thunderx2t99_prefetch_tune): Likewise.
2232 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2233 (aarch64_override_options_internal): Update to set
2234 PARAM_PREFETCH_DYNAMIC_STRIDES.
2235 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2236 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2237 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2238 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2239 prefetch-dynamic-strides setting.
2241 2018-05-07 Luis Machado <luis.machado@linaro.org>
2243 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2244 <minimum_stride>: New const int field.
2245 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2246 minimum_stride field.
2247 (exynosm1_prefetch_tune): Likewise.
2248 (thunderxt88_prefetch_tune): Likewise.
2249 (thunderx_prefetch_tune): Likewise.
2250 (thunderx2t99_prefetch_tune): Likewise.
2251 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2252 (aarch64_override_options_internal): Update to set
2253 PARAM_PREFETCH_MINIMUM_STRIDE.
2254 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2255 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2256 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2257 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2258 stride is constant and is below the minimum stride threshold.
2260 2018-05-07 Luis Machado <luis.machado@linaro.org>
2262 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2265 2018-05-07 Luis Machado <luis.machado@linaro.org>
2267 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2268 <prefetch_dynamic_strides>: New const bool field.
2269 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2270 prefetch_dynamic_strides.
2271 (exynosm1_prefetch_tune): Likewise.
2272 (thunderxt88_prefetch_tune): Likewise.
2273 (thunderx_prefetch_tune): Likewise.
2274 (thunderx2t99_prefetch_tune): Likewise.
2275 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2276 (aarch64_override_options_internal): Update to set
2277 PARAM_PREFETCH_DYNAMIC_STRIDES.
2278 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2279 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2280 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2281 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2282 prefetch-dynamic-strides setting.
2284 2018-05-07 Luis Machado <luis.machado@linaro.org>
2286 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2287 <minimum_stride>: New const int field.
2288 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2289 minimum_stride field.
2290 (exynosm1_prefetch_tune): Likewise.
2291 (thunderxt88_prefetch_tune): Likewise.
2292 (thunderx_prefetch_tune): Likewise.
2293 (thunderx2t99_prefetch_tune): Likewise.
2294 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2295 (aarch64_override_options_internal): Update to set
2296 PARAM_PREFETCH_MINIMUM_STRIDE.
2297 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2298 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2299 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2300 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2301 stride is constant and is below the minimum stride threshold.
2303 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2306 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2307 the type is addressable. Don't force op into register if it has
2310 2018-05-05 Roland McGrath <mcgrathr@google.com>
2313 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2314 any section for which we don't know a specific type it should have,
2315 regardless of name. Previously this was done only for the exact
2316 names ".init_array", ".fini_array", and ".preinit_array".
2317 (default_elf_asm_named_section): Add comment about
2318 relationship with default_section_type_flags and SECTION_NOTYPE.
2319 (get_section): Don't consider it a type conflict if one side has
2320 SECTION_NOTYPE and the other doesn't, as long as neither has the
2321 SECTION_BSS et al used in the default_section_type_flags logic.
2323 2018-05-05 Tom de Vries <tom@codesourcery.com>
2326 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2327 (workaround_barsyncs): New function.
2328 (nvptx_reorg): Use workaround_barsyncs.
2329 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2330 (define_expand "nvptx_membar_cta"): New define_expand.
2331 (define_insn "*nvptx_membar_cta"): New insn.
2333 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2335 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2336 To improve optimization opportunities.
2337 * builtin-types.def: The new needed builtin types for the above.
2339 2018-05-04 Richard Biener <rguenther@suse.de>
2341 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2342 * gimple-ssa-store-merging.c
2343 (imm_store_chain_info::output_merged_store): Remove redundant create,
2344 release split_store vector contents on failure.
2345 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2346 scalar stmt vector on cache hit.
2348 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2350 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2352 * config.gcc (powerpc-xilinx-eabi*): Remove.
2353 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2355 (fusion_addis_mem_combo_load): Ditto.
2356 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2358 (rs6000_cpu_cpp_builtins): Ditto.
2359 * config/rs6000/rs6000-linux.c
2360 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2361 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2362 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2364 (rs6000_setup_reg_addr_masks): Ditto.
2365 (rs6000_init_hard_regno_mode_ok): Ditto.
2366 (rs6000_option_override_internal): Ditto.
2367 (legitimate_lo_sum_address_p): Ditto.
2368 (rs6000_legitimize_address): Ditto.
2369 (rs6000_legitimize_reload_address): Ditto.
2370 (rs6000_legitimate_address_p): Ditto.
2371 (abi_v4_pass_in_fpr): Ditto.
2372 (setup_incoming_varargs): Ditto.
2373 (rs6000_gimplify_va_arg): Ditto.
2374 (rs6000_split_multireg_move): Ditto.
2375 (rs6000_savres_strategy): Ditto.
2376 (rs6000_emit_prologue_components): Ditto.
2377 (rs6000_emit_epilogue_components): Ditto.
2378 (rs6000_emit_prologue): Ditto.
2379 (rs6000_emit_epilogue): Ditto.
2380 (rs6000_elf_file_end): Ditto.
2381 (rs6000_function_value): Ditto.
2382 (rs6000_libcall_value): Ditto.
2383 * config/rs6000/rs6000.h: Ditto.
2384 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2385 (TARGET_MINMAX): ... this. New.
2386 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2387 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2388 (*movsi_internal1_single): Delete.
2389 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2390 mfpu=, mxilinx-fpu): Delete.
2391 * config/rs6000/singlefp.h: Delete.
2392 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2393 * config/rs6000/t-rs6000: Ditto.
2394 * config/rs6000/t-xilinx: Delete.
2395 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2396 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2397 (VStype_simple): Delete.
2398 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2399 * config/rs6000/xfpu.h: Delete.
2400 * config/rs6000/xfpu.md: Delete.
2401 * config/rs6000/xilinx.h: Delete.
2402 * config/rs6000/xilinx.opt: Delete.
2403 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2404 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2406 2018-05-04 Tom de Vries <tom@codesourcery.com>
2409 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2412 2018-05-04 Richard Biener <rguenther@suse.de>
2415 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2416 (expand_complex_div_wide): Likewise.
2417 (expand_complex_operations_1): Likewise.
2418 (expand_complex_libcall): Preserve EH info of the original stmt.
2419 (tree_lower_complex): Handle removed blocks.
2420 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2421 on complex multiplication and division libcall builtins.
2423 2018-05-04 Richard Biener <rguenther@suse.de>
2426 * fold-const.c (negate_expr_p): Restrict negation of operand
2427 zero of a division to when we know that can happen without
2429 (fold_negate_expr_1): Likewise.
2431 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2434 * real.h (real_nextafter): Declare.
2435 * real.c (real_nextafter): New function.
2436 * fold-const-call.c (fold_const_nextafter): New function.
2437 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2438 CASE_CFN_NEXTTOWARD.
2439 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2440 even when arg1_mode is different from arg0_mode.
2442 2018-05-03 Nathan Sidwell <nathan@acm.org>
2444 * doc/extend.texi (Deprecated Features): Remove
2446 (Backwards Compatibility): Likewise.
2447 * doc/invoke.texi (C++ Language Options): Likewise.
2448 (C++ Dialect Options): Likewise.
2450 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2453 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2454 _mm512_mask_mullox_epi64): New intrinsics.
2456 2018-05-03 Tom de Vries <tom@codesourcery.com>
2459 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2460 dump files): Add offload-tree.
2462 2018-05-03 Richard Biener <rguenther@suse.de>
2464 PR tree-optimization/85615
2465 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2466 to loops not nested in BBs loop father to avoid creating multi-entry
2469 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2471 PR tree-optimization/70291
2472 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2473 arguments. Change return type to tree. Emit libcall as a new
2474 statement rather than replacing existing one when inplace_p is true.
2475 (expand_complex_multiplication_components): New function.
2476 (expand_complex_multiplication): Expand floating-point complex
2477 multiplication using the above.
2478 (expand_complex_division): Rename inner_type parameter to type.
2479 Update expand_complex_libcall call-site.
2480 (expand_complex_operations_1): Update expand_complex_multiplication
2481 and expand_complex_division call-sites.
2483 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2486 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2487 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2488 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2489 the highest significant bit of the shift count mask is clear. In
2490 check whether and[sq]i3 is needed verify that all significant bits
2491 of the shift count other than the highest are set.
2493 2018-05-02 Tom de Vries <tom@codesourcery.com>
2496 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2497 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2498 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2499 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2500 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2501 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2502 __builtin_goacc_parlevel_size.
2504 2018-05-02 Richard Biener <rguenther@suse.de>
2506 PR tree-optimization/85597
2507 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2508 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2511 2018-05-02 Tom de Vries <tom@codesourcery.com>
2514 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2515 dump files): Add ltrans-tree.
2517 2018-05-02 Tom de Vries <tom@codesourcery.com>
2520 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2521 dump files): Add wpa-ipa.
2523 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2525 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2526 powerpc*-*-linux*paired* target.
2527 * config/rs6000/750cl.h: Delete.
2528 * config/rs6000/paired.h: Delete.
2529 * config/rs6000/paired.md: Delete.
2530 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2532 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2533 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2534 comment. Remove paired float support.
2535 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2536 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2538 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2539 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2541 * config/rs6000/rs6000.c: Remove paired float support.
2542 (paired_expand_vector_init, paired_expand_vector_move,
2543 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2544 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2545 paired_expand_builtin, paired_expand_predicate_builtin,
2546 paired_init_builtins): Delete.
2547 * config/rs6000/rs6000.h: Remove paired float support.
2548 * config/rs6000/rs6000.md: Remove paired float support.
2549 (move_from_CR_ov_bit): Delete.
2550 * config/rs6000/rs6000.opt (mpaired): Delete.
2551 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2552 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2554 2018-05-02 Richard Biener <rguenther@suse.de>
2557 * gimplify.c (gimplify_save_expr): When in SSA form allow
2558 SAVE_EXPRs to compute to SSA vars.
2560 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2563 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2564 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2565 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2566 clobber operands[2], instead use a new pseudo. Formatting fixes.
2568 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2570 PR tree-optimization/85586
2571 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2572 exit early for statements in the same group if the accesses are
2575 2018-05-02 Tom de Vries <tom@codesourcery.com>
2578 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2581 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2583 PR tree-optimization/85143
2584 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2586 2018-05-01 Tom de Vries <tom@codesourcery.com>
2589 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2590 not found" error message.
2592 2018-05-01 Tom de Vries <tom@codesourcery.com>
2595 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2596 * vec.c (test_ordered_remove_if): New function.
2597 (vec_c_tests): Call test_ordered_remove_if.
2598 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2599 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2600 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2601 VEC_ORDERED_REMOVE_IF.
2603 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2605 PR tree-optimization/82665
2606 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2607 pointer subtraction where arguments come from a memchr call.
2609 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2611 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2612 --push-state --as-needed and --pop-state instead of --as-needed and
2613 --no-as-needed if ld supports it.
2614 * configure: Regenerated.
2617 * doc/install.texi2html: Replace _002d with - and _002a with * in
2618 generated html files using sed.
2620 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2623 * gcc-rich-location.c (blank_line_before_p): New function.
2624 (use_new_line): New function.
2625 (gcc_rich_location::add_fixit_insert_formatted): New function.
2626 * gcc-rich-location.h
2627 (gcc_rich_location::add_fixit_insert_formatted): New function.
2629 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2631 * selftest.c (assert_streq): Rename "expected" and "actual" to
2632 "val1" and "val2". Extend NULL-handling to cover both inputs
2633 symmetrically, while still requiring both to be non-NULL for a pass.
2634 * selftest.h (assert_streq): Rename "expected" and "actual" to
2636 (ASSERT_EQ): Likewise.
2637 (ASSERT_EQ_AT): Likewise.
2638 (ASSERT_KNOWN_EQ): Likewise.
2639 (ASSERT_KNOWN_EQ_AT): Likewise.
2640 (ASSERT_NE): Likewise.
2641 (ASSERT_MAYBE_NE): Likewise.
2642 (ASSERT_MAYBE_NE_AT): Likewise.
2643 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2644 the assertion to pass.
2645 (ASSERT_STREQ_AT): Likewise.
2647 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2649 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2650 interaction with -pie.
2652 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2654 * selftest.h: Fix alphabetization of per-source-file selftest
2657 2018-04-30 Jason Merrill <jason@redhat.com>
2659 PR c++/61982 - dead stores to destroyed objects.
2660 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2663 2018-04-30 Jason Merrill <jason@redhat.com>
2665 * tree.c (build_clobber): New.
2666 * tree.h: Declare it.
2667 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2669 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2671 * diagnostic-show-locus.c (layout::layout): Update for
2672 location_get_source_line returning a char_span.
2673 (struct char_span): Move to input.h.
2674 (struct correction): Update for fields in char_span becoming
2676 (struct source_line): Update for location_get_source_line
2677 returning a char_span.
2678 (layout::print_line): Likewise.
2679 * edit-context.c (edited_file::print_content): Likewise.
2680 (edited_file::print_diff_hunk): Likewise.
2681 (edited_file::print_run_of_changed_lines): Likewise.
2682 (edited_file::get_num_lines): Likewise.
2683 (edited_line::edited_line): Likewise.
2684 * final.c (asm_show_source): Likewise.
2685 * input.c (location_get_source_line): Convert return type
2686 from const char * to char_span, losing the final "line_len"
2688 (dump_location_info): Update for the above.
2689 (get_substring_ranges_for_loc): Likewise. Use a char_span
2690 when handling the literal within the line.
2691 (test_reading_source_line): Update for location_get_source_line
2692 returning a char_span.
2693 * input.h (class char_span): Move here from
2694 diagnostic-show-locus.c, converting from a struct to a class.
2695 Make data members private.
2696 (char_span::operator bool): New.
2697 (char_span::length): New.
2698 (char_span::get_buffer): New.
2699 (char_span::operator[]): New.
2700 (char_span::subspan): Make const.
2701 (char_span::xstrdup): New.
2702 (location_get_source_line): Convert return type from const char *
2703 to char_span, losing the final "line_size" param.
2705 2018-04-30 Jan Hubicka <jh@suse.cz>
2707 * lto-wrapper.c (ltrans_priorities): New static var.
2708 (cmp_priority): New.
2709 (run_gcc): Read priorities and if doing parallel build order
2710 the Makefile by them.
2712 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2714 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2716 2018-04-30 Richard Biener <rguenther@suse.de>
2718 * tree-cfg.c (verify_address): Remove base argument, add
2719 flag whether to check TREE_ADDRESSABLE and do that.
2720 (verify_expr): Remove.
2721 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2722 (verify_gimple_assign_single): Likewise.
2723 (verify_gimple_switch): Likewise.
2724 (verify_expr_location_1): Dereference tp once. Add (disabled)
2725 piece from verify_expr.
2726 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2728 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2730 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2732 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2734 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2735 (small_data_pattern): Likewise.
2736 (arc_rewrite_small_data): Likewise.
2737 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2738 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2739 (get_symbol_alignment): New function.
2740 (legitimate_small_data_address_p): Likewise.
2741 (legitimate_scaled_address): Update, call
2742 legitimate_small_data_address_p.
2743 (output_sdata): New static variable.
2744 (arc_print_operand): Update how we handle small data operands.
2745 (arc_print_operand_address): Likewise.
2746 (arc_legitimate_address_p): Update, use
2747 legitimate_small_data_address_p.
2748 (arc_rewrite_small_data_p): Remove.
2749 (arc_rewrite_small_data_1): Likewise.
2750 (arc_rewrite_small_data): Likewise.
2751 (small_data_pattern): Likewise.
2752 (compact_sda_memory_operand): Update to use
2753 legitimate_small_data_address_p and get_symbol_alignment.
2754 (prepare_move_operands): Don't rewite sdata pattern.
2755 (prepare_extend_operands): Remove.
2756 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2758 (zero_extendqisi2): Likewise.
2759 (zero_extendhisi2): Likewise.
2760 (extendqihi2): Likewise.
2761 (extendqisi2): Likewise.
2762 (extendhisi2): Likewise.
2766 * config/arc/constraints.md (Usd): Change it to memory constraint.
2768 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2770 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2771 as source of std instructions.
2772 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2773 allow 6-bit constants as source for store instructions.
2774 (movdi_insn): Update instruction pattern to allow 6-bit constants
2775 as source for store instructions.
2777 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2779 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2781 2018-04-30 Nathan Sidwell <nathan@acm.org>
2782 Sandra Loosemore <sandra@codesourcery.com>
2784 * dumpfile.c (dump_open): Allow '-' for stdout.
2785 * doc/invoke.texi (Developer Options): Document dump filename
2786 determination early. Document stdin/stdout selection.
2788 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2790 Microblaze Target: PIC data text relative
2792 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2793 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2795 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2796 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2797 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2798 New addressing mode for data-text relative position indepenedent code.
2799 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2800 'ADDRESS_SYMBOLIC_TXT_REL'.
2801 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2802 (microblaze_legitimate_pic_operand): Exclude function calls from
2803 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2804 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2806 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2807 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2808 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2809 for 'address + offset'.
2810 (microblaze_expand_prologue): Add new function prologue call for
2812 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2813 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2814 table in case of TARGET_PIC_DATA_TEXT_REL.
2815 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2816 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2817 Add new macros 'UNSPEC_TEXT',
2818 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2819 + exclude function calls from 'UNSPEC_PLT' in case of data text
2821 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2822 new target hook for generating address diff vector tables in case of
2824 * doc/tm.texi : Regenerate.
2825 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2826 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2827 of addr diff vector generation.
2828 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2829 target hook definition.
2830 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2831 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2832 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2833 option for data text relative.
2835 2018-04-30 Richard Biener <rguenther@suse.de>
2837 * tree-chrec.h (evolution_function_is_constant_p): Remove
2839 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2841 2018-04-30 Richard Biener <rguenther@suse.de>
2844 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2846 2018-04-30 Richard Biener <rguenther@suse.de>
2848 PR tree-optimization/28364
2849 PR tree-optimization/85275
2850 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2851 copying first exit test.
2853 2018-04-28 Mark Wielaard <mark@klomp.org>
2855 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2857 (dwarf_AT): Handle DW_AT_addr_base.
2858 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2860 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2863 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2864 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2865 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2866 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2868 2018-04-28 Richard Biener <rguenther@suse.de>
2870 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2871 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2872 to reflect use. Only add interesting stmts.
2874 2018-04-27 Martin Jambor <mjambor@suse.cz>
2877 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2878 the jump function allows for passing through aggregate values.
2880 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2882 * input.h (in_system_header_at): Convert from macro to inline
2884 (from_macro_expansion_at): Likewise.
2885 (from_macro_definition_at): Likewise.
2887 2018-04-27 Jeff Law <law@redhat.com>
2889 * config.gcc: Mark tile* targets as deprecated/obsolete.
2891 2018-04-27 Richard Biener <rguenther@suse.de>
2893 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2896 2018-04-27 Richard Biener <rguenther@suse.de>
2898 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2900 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2902 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2903 with Yd constraint. Set "preferred_for_speed" attribute from
2904 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2906 (*movdi_internal): Ditto.
2907 (movti_interunit splitters): Remove
2908 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2909 (movdi_interunit splitters): Ditto.
2910 * config/i386/constraints.md (Ye): Remove.
2911 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2913 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2916 * config/aarch64/constraints.md (Usg): Limit to 31.
2919 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2921 PR tree-optimization/85529
2922 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2923 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2924 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2925 zero extension or masking of the MSB bit.
2926 (optimize_range_tests): Add FIRST_BB argument, pass it through
2927 to optimize_range_tests_var_bound.
2928 (maybe_optimize_range_tests, reassociate_bb): Adjust
2929 optimize_range_tests callers.
2931 2018-04-26 Richard Biener <rguenther@suse.de>
2932 Jakub Jelinek <jakub@redhat.com>
2934 * cgraph.h (symbol_table): Just declare debug method here.
2935 * symtab.c (symbol_table::debug): Define.
2937 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2939 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2941 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2943 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2944 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2945 (*movdi_internal): Substitute Yi and Yj constraint with x
2946 and Ym and Yn constraint with y constraint. Update "isa"
2947 attribute and set "preferred_for_speed" attribute from
2948 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2949 (*movsi_internal): Ditto.
2950 (*movdf_internal): Ditto.
2951 (*movsf_internal): Ditto.
2952 (*zero_extendsidi2): Ditto.
2953 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2954 (sse2_loadld): Ditto.
2955 (*vec_extract<ssevecmodelower>_0): Ditto.
2956 (*vec_extractv4si_0_zext_sse4): Ditto.
2957 (vec_concatv2di): Ditto.
2958 (*vec_dup<mode>): Ditto.
2959 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2960 * config/i386/constraints.md (Yi): Remove.
2965 2018-04-26 Nathan Sidwell <nathan@acm.org>
2967 * dumpfile.c (dump_open): New.
2968 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2969 (dump_finish): Detect stdio/stderr by value not name.
2971 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2973 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2975 2018-04-26 Tom de Vries <tom@codesourcery.com>
2978 * config/nvptx/nvptx.c (verify_neutering_jumps)
2979 (verify_neutering_labels): New function
2980 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2982 2018-04-26 Tom de Vries <tom@codesourcery.com>
2985 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2986 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2989 2018-04-26 Richard Biener <rguenther@suse.de>
2990 Tom de Vries <tom@codesourcery.com>
2993 * lto-streamer-out.c (output_function): Fixup loops if required to match
2994 discovery done in the reader.
2996 2018-04-26 Richard Biener <rguenther@suse.de>
2998 PR tree-optimization/85116
2999 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3000 have a loop exit from the single latch predecessor. Remove
3001 case of header with just condition.
3002 (ch_base::copy_headers): Exclude infinite loops from any
3004 (pass_ch::execute): Record exits.
3006 2018-04-26 Richard Biener <rguenther@suse.de>
3008 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3009 prologue cost vector and pass it to vect_get_load_cost.
3010 (vect_get_peeling_costs_all_drs): Likewise.
3011 (vect_peeling_hash_get_lowest_cost): Likewise.
3012 (vect_enhance_data_refs_alignment): Likewise.
3014 2018-04-26 Richard Biener <rguenther@suse.de>
3017 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3018 checking of integer<->pointer conversions.
3019 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3020 sign-/zero-extending pointer types.
3021 (expand_omp_for_static_chunk): Likewise.
3023 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3024 Jean Lee <xiaoyur347@gmail.com>
3026 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3027 (TARGET_ASAN_SHADOW_OFFSET): Define.
3028 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3029 true for -fsanitize=address.
3031 2018-04-25 Mark Wielaard <mark@klomp.org>
3033 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3036 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3038 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3039 than "alu", remove explicit "memory" and "imm_disp" attributes.
3040 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3043 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3044 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3045 gen_lowpart_no_emit.
3047 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3050 * config/i386/i386.c (ix86_expand_builtin): Change memory
3051 operand to XI, extend p0 to Pmode.
3052 * config/i386/i386.md: Change unspec volatile and operand
3053 1 mode to XI, change operand 0 mode to P.
3055 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3057 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3058 GET_MODE_MASK before any checking.
3059 (nds32_can_use_bset_p): Likewise.
3060 (nds32_can_use_btgl_p): Likewise.
3062 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3064 * config/nds32/nds32-doubleword.md: New define_split pattern for
3065 illegal register number.
3067 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3069 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3071 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3073 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3075 2018-04-25 Richard Biener <rguenther@suse.de>
3077 * lto-streamer.h (LTO_major_version): Bump to 8.
3079 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3081 * BASE-VER: Set to 9.0.0.
3083 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3085 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3086 in __abskf2 and __powikf2.
3088 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3091 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3092 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3093 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3094 Use the above on operand 2. Reindent.
3095 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3097 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3100 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3102 * config/i386/i386.opt (mcet): Removed.
3103 * doc/install.texi: Remove -mcet documentation.
3104 * doc/invoke.texi: Likewise.
3106 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3109 * doc/install.texi: Remove -mcet from bootstrap-cet.
3111 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3114 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3115 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3119 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3120 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3121 containing a CONST_VECTOR.
3123 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3125 * doc/install.texi: Update newlib dependency for nvptx.
3127 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3130 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3131 instead of INTVAL when shifting x left.
3133 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3135 PR tree-optimization/85478
3136 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3137 vect_grouped_store_supported for single element vectors.
3139 2018-04-24 Richard Biener <rguenther@suse.de>
3142 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3143 load cost increase to the case of non-constant step.
3145 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3148 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3149 destination if any_malformed_asm.
3151 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3154 * expr.c (store_field): In the bitfield case, if the value comes from
3155 a function call and is returned in registers by means of a PARALLEL,
3156 do not change the mode of the temporary unless BLKmode and VOIDmode.
3158 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3160 PR rtl-optimization/85423
3161 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3162 dependencies to debug insns when the previous insn is non-debug.
3164 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3166 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3167 enums into a single definition.
3168 (fls): Fix predicates and printing.
3171 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3173 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3174 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3175 and short u6 immediate.
3176 (check_if_valid_sleep_operand): Remove.
3177 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3179 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3181 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3182 flag_always_save_lp condition.
3183 * config/nds32/nds32.opt (malways-save-lp): New option.
3185 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3187 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3188 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3189 * config/nds32/nds32.h
3190 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3191 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3193 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3195 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3196 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3198 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3199 Chung-Ju Wu <jasonwucj@gmail.com>
3201 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3203 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3204 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3206 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3208 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3210 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3212 * config/nds32/nds32-protos.h (nds32_data_alignment,
3213 nds32_local_alignment): Declare.
3214 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3215 nds32_local_alignment): New functions.
3216 (TARGET_CONSTANT_ALIGNMENT): Define.
3217 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3219 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3221 * config/nds32/nds32.c
3222 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3223 (TARGET_MODES_TIEABLE_P): Likewise.
3225 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3227 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3230 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3231 Chung-Ju Wu <jasonwucj@gmail.com>
3233 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3234 for unaligned access.
3235 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3237 * config/nds32/nds32-intrinsic.md: Likewise.
3238 * config/nds32/nds32_intrinsic.h: Likewise.
3239 * config/nds32/nds32.h (nds32_builtins): Likewise.
3240 * config/nds32/nds32.opt (munaligned-access): New option.
3241 * config/nds32/nds32.c (nds32_asm_file_start): Display
3242 flag_unaligned_access status.
3244 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3246 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3247 -mno-relax is present.
3248 * config/riscv/linux.h (LINK_SPEC): Ditto.
3250 2018-04-20 Martin Sebor <msebor@redhat.com>
3253 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3255 (gimple_fold_builtin_stxcpy_chk): Same.
3256 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3258 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3261 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3262 __powikf2 when long double is IEEE 128-bit.
3264 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3266 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3267 step to make sure stack always aligned.
3269 2018-04-20 Carl Love <cel@us.ibm.com>
3272 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3273 size check for arg0.
3275 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3276 Tom de Vries <tom@codesourcery.com>
3279 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3280 Emit insns for calls too.
3281 (nvptx_find_par): Always look for worker-level predecessor insn.
3282 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3284 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3285 (nvptx_process_pars): Propagate frames for calls.
3287 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3290 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3292 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3293 (ix86_handle_option): Don't handle OPT_mibt.
3294 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3296 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3298 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3299 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3300 (ix86_target_macros): Define __CET__ with flag_cf_protection
3301 for -fcf-protection.
3302 * config/i386/i386.c (isa2_opts): Remove -mibt.
3303 * config/i386/i386.h (TARGET_IBT): Removed.
3304 (TARGET_IBT_P): Likewise.
3305 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3306 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3307 * config/i386/i386.opt (mcet): Update help message.
3310 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3311 -mcet as an alias for -mshstk.
3313 2018-04-20 Richard Biener <rguenther@suse.de>
3316 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3317 complexity by forcing a single use of the multiply operand.
3319 2018-04-20 Martin Jambor <mjambor@suse.cz>
3322 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3323 recursion dependency to only apply to non-clones.
3325 2018-04-20 Martin Jambor <mjambor@suse.cz>
3328 * ipa-cp.c (create_specialized_node): Check that clones of
3329 self-recursive edges exist during IPA-CP.
3331 2018-04-19 Toon Moene <toon@moene.org>
3333 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3336 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3338 PR tree-optimization/85467
3339 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3340 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3341 VECTOR_CST element to type.
3343 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3346 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3347 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3348 (builtin_longjmp): Likewise.
3349 (save_stack_nonlocal): New pattern.
3350 (restore_stack_nonlocal): Likewise.
3352 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3355 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3356 Replace ASM_OUTPUT_LABEL with fprintf.
3358 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3361 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3362 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3363 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3364 define __IBT__ and __SHSTK__ for -fcf-protection.
3365 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3367 (ix86_trampoline_init): Likewise.
3368 (x86_output_mi_thunk): Likewise.
3369 (ix86_notrack_prefixed_insn_p): Likewise.
3370 (ix86_option_override_internal): Don't disallow -fcf-protection.
3371 * config/i386/i386.md (rdssp<mode>): Also enable for
3373 (incssp<mode>): Likewise.
3374 (nop_endbr): Likewise.
3375 * config/i386/i386.opt (mcet): Change help message to built-in
3379 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3380 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3381 enable CET built-in functions.
3383 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3385 * common/config/i386/i386-common.c
3386 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3387 OPTION_MASK_ISA_MOVDIRI_UNSET,
3388 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3389 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3390 * config.gcc (movdirintrin.h): New header.
3391 * config/i386/cpuid.h (bit_MOVDIRI,
3392 bit_MOVDIR64B): New bits.
3393 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3395 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3396 (VOID, PVOID, PCVOID)): New function types.
3397 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3398 __builtin_ia32_directstoreu_u64,
3399 __builtin_ia32_movdir64b): New builtins.
3400 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3401 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3403 (ix86_valid_target_attribute_inner_p): Ditto.
3404 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3405 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3406 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3407 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3408 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3409 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3410 (movdiri<mode>, movdir64b_<mode>): New.
3411 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3412 * config/i386/immintrin.h: Include movdirintrin.h.
3413 * config/i386/movdirintrin.h: New file.
3414 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3416 2018-04-19 Richard Biener <rguenther@suse.de>
3419 * cfg.c (clear_bb_flags): When loop state says we have
3420 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3422 2018-04-19 Richard Biener <rguenther@suse.de>
3424 PR tree-optimization/84737
3425 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3426 copying restrict info.
3427 (vect_setup_realignment): Use it.
3428 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3429 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3430 the first DR to all generated stores.
3431 (vectorizable_load): Likewise for loads.
3433 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3435 PR tree-optimization/85446
3436 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3437 the integral and pointer types to have the same precision.
3439 * doc/install.texi: Document --disable-cet being the default and
3442 2018-04-18 Martin Liska <mliska@suse.cz>
3444 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3447 2018-04-18 Martin Liska <mliska@suse.cz>
3450 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3453 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3454 arguments if they are comparable.
3456 2018-04-18 Martin Liska <mliska@suse.cz>
3459 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3462 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3465 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3468 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3469 ENDBR after calling __morestack.
3471 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3474 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3475 by using gcc_base_ver to generate a gcc_driver_version, and use
3476 it when generating GCC_DRIVER_NAME.
3477 * configure: Regenerate.
3479 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3482 * config.gcc: Obsolete powerpc*-*-*spe*.
3484 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3487 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3488 (stabstr_D): Change type of unum from unsigned int to
3489 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3492 2018-04-17 Jim Wilson <jimw@sifive.com>
3495 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3496 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3497 Set arg_pointer_offset after using pretend_args_size.
3499 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3501 PR rtl-optimization/85431
3502 * dse.c (record_store): Ignore zero width stores.
3505 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3506 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3507 __builtin_stack_restore rather than after it.
3508 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3509 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3510 argument instead of virtual_dynamic_stack_rtx.
3512 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3514 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3516 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3517 Add note to error message to explain internal mapping of overloaded
3518 built-in function name to non-overloaded built-in function name.
3519 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3522 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3525 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3526 where the inputs overlap with the output.
3528 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3530 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3531 (=v, v) alternative and explicit "memory" attribute.
3532 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3533 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3535 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3536 "sselog1" type instead of "sselog".
3537 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3538 "sselog". Remove explicit "memory" attribute.
3539 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3540 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3542 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3543 "sselog1" type instead of "sselog". Remove explicit "memory"
3545 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3546 use "sselog1" type instead of "sselog". Remove explicit "memory"
3548 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3549 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3551 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3552 "sselog1" type instead of "sselog". Remove explicit "memory"
3554 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3555 use "sselog1" type instead of "sselog". Remove explicit "memory"
3559 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3562 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3565 2018-04-17 Martin Jambor <mjambor@suse.cz>
3568 * ipa-cp.c (create_specialized_node): Call
3569 expand_all_artificial_thunks if necessary.
3571 2018-04-17 Martin Liska <mliska@suse.cz>
3574 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3575 in message, remote space in between '_G' and '('.
3577 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3580 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3581 avx512f_vmcmp<mode>3<round_saeonly_name>,
3582 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3583 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3584 avx512f_rndscale<mode><round_saeonly_name>,
3585 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3586 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3587 Use %<iptr>2 instead of %2 for -masm=intel.
3588 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3589 avx512f_vcvttss2usi<round_saeonly_name>,
3590 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3592 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3593 avx512f_vcvttsd2usi<round_saeonly_name>,
3594 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3595 Use %q1 instead of %1 for -masm=intel.
3596 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3597 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3598 of %3 for -masm=intel.
3599 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3601 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3603 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3605 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3606 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3608 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3610 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3611 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3612 %g1 and one with %0 and %1.
3613 (avx512er_vmrcp28<mode><round_saeonly_name>,
3614 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3616 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3617 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3618 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3619 of %0 and %{%4%} for -masm=intel.
3620 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3621 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3622 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3623 order of %0 and %{%5%}%{z%} for -masm=intel.
3625 2018-04-17 Jan Hubicka <jh@suse.cz>
3628 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3630 2018-04-17 Martin Liska <mliska@suse.cz>
3633 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3634 for target_clone error message. Make default implementation
3635 clone to be a local declaration.
3636 (separate_attrs): Add new argument and check for an empty
3638 (expand_target_clones): Handle it.
3639 (ipa_target_clone): Make redirection just for target_clones
3642 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3643 Tom de Vries <tom@codesourcery.com>
3646 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3647 tiled basic blocks without omp continue statements.
3649 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3652 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3653 vec_extract expression as having side effects to make sure it gets
3656 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3659 * config/i386/i386.c (get_builtin_code_for_version): Check
3662 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3665 * gcc/config.gcc: Support "skylake".
3666 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3668 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3669 (processor_target_table): Add "skylake".
3670 (ix86_option_override_internal): Add "skylake".
3671 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3672 PROCESSOR_CANNONLAKE.
3673 (get_builtin_code_for_version): Fix priority for
3674 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3675 PROCESSOR_SKYLAKE-AVX512.
3676 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3677 (processor_type): Add PROCESSOR_SKYLAKE.
3679 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3680 Jason Merrill <jason@redhat.com>
3683 * convert.c (convert_to_integer_1): Use direct recursion for
3684 enumeral types and types with a precision less than the number
3685 of bits in their mode.
3687 2018-04-16 Julia Koval <julia.koval@intel.com>
3690 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3691 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3693 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3696 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3697 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3698 and -mno-direct-move.
3700 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3703 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3704 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3705 Ensure negative shifts result in {0}.
3707 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3709 PR rtl-optimization/79916
3710 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3711 regs (if any) to define how to gnerate SD moves when LRA is in
3714 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3716 PR rtl-optimization/85393
3717 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3718 * except.c (expand_dw2_landing_pad_for_region): Make static.
3719 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3720 a label and unconditional jump to old_bb, rather than
3721 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3724 PR rtl-optimization/85376
3725 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3726 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3727 instead of a specific value.
3729 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3730 Bin Cheng <bin.cheng@arm.com>
3732 PR tree-optimization/82965
3733 PR tree-optimization/83991
3734 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3735 by_profile_only parameter.
3736 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3737 information if the loop was predicted to iterate too many times.
3738 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3740 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3743 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3746 2018-04-13 Martin Liska <mliska@suse.cz>
3747 Jakub Jelinek <jakub@redhat.com>
3750 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3751 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3752 * builtins.c (expand_builtin_memory_copy_args): Use
3753 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3754 handle dest_addr == pc_rtx.
3756 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3759 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3760 asked to not generate direct moves.
3761 (fix_trunc<mode>si2_stfiwx): Similar.
3762 (fix_trunc<mode>si2_internal): Similar.
3764 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3767 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3768 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3769 lookup if dest in some wider mode is known to be const0_rtx and
3770 if so, record permanent equivalence for it to be ZERO_EXTEND of
3771 the narrower mode destination.
3773 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3775 * lto-streamer-out.c (output_function): Revert 259346.
3776 * omp-expand.c (expand_oacc_for): Likewise.
3778 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3780 PR rtl-optimization/85354
3781 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3782 * sel-sched.c (sel_global_init): ... here.
3784 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3787 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3788 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3789 mode for PE-COFF targets.
3790 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3791 (i386_pe_asm_lto_end): Likewise.
3792 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3793 (TARGET_ASM_LTO_END): Likewise.
3794 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3795 (i386_pe_asm_lto_start): New function.
3796 (i386_pe_asm_lto_end): Likewise.
3798 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3799 Richard Biener <rguenther@suse.de>
3802 * lto-streamer-out.c (output_function): Fix CFG loop state before
3804 * omp-expand.c (expand_oacc_for): Handle calls to internal
3805 functions like regular functions.
3807 2018-04-12 Richard Biener <rguenther@suse.de>
3810 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3811 for the early LTO debug to properly generate references to it
3812 during DIE emission. Do not re-use that for the skeleton for
3814 (dwarf2out_early_finish): Likewise.
3816 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3819 * config/i386/sse.md
3820 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3821 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3822 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3823 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3824 and output is a reg, avoid creating invalid lowpart subreg, but
3825 instead split into a 512-bit move. Don't split if not AVX512VL,
3826 input is xmm16+ reg and output is a mem.
3827 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3828 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3829 xmm16+ reg and output is a mem.
3831 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3833 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3834 also for flag_dwarf2_cfi_asm.
3836 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3838 PR rtl-optimization/85342
3839 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3840 a bool scalar var inside of the loop instead. Don't try to update
3841 recog_data.operand after failed apply_change_group.
3843 2018-04-12 Tom de Vries <tom@codesourcery.com>
3846 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3847 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3848 array with flexible array member as array without given dimension.
3849 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3850 argument for undefined param to true.
3852 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3855 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3856 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3857 from PowerPC section.
3858 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3859 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3860 help text that is too long.
3861 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3862 help text that is too long.
3863 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3864 help text that is too long.
3866 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3868 * config/alpha/alpha.md (stack_probe_internal): Rename
3869 from "probe_stack". Update all callers.
3871 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3873 PR rtl-optimization/84566
3874 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3875 sched_macro_fuse_insns.
3877 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3880 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3881 (compute_block_dependences): ... from here.
3883 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3885 PR tree-optimization/85331
3886 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3887 from int to HOST_WIDE_INT.
3889 2018-04-11 Martin Jambor <mjambor@suse.cz>
3892 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3893 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3894 not the same as the source val.
3895 (cgraph_edge_brings_value_p): New parameter.
3896 (gather_edges_for_value): Pass destination value to
3897 cgraph_edge_brings_value_p.
3898 (perhaps_add_new_callers): Likewise.
3899 (get_info_about_necessary_edges): Likewise and exclude values brought
3900 only by self-recursive edges.
3901 (create_specialized_node): Redirect only clones of self-calling edges.
3902 (+self_recursive_pass_through_p): New function.
3903 (find_more_scalar_values_for_callers_subset): Use it.
3904 (find_aggregate_values_for_callers_subset): Likewise.
3905 (known_aggs_to_agg_replacement_list): Removed.
3906 (decide_whether_version_node): Re-calculate known constants for all
3907 remaining context clones.
3909 2018-04-11 Richard Biener <rguenther@suse.de>
3912 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3913 from early DWARF output.
3914 (dwarf2out_early_finish): Output line info unconditionally into
3915 early DWARF and add reference to it.
3917 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3920 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3921 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3922 other than V2DFmode using iptr mode attribute.
3923 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3925 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3927 PR rtl-optimization/84659
3928 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3930 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3933 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3935 (output_loc_list): Pass address of a dummy size variable even in the
3936 locview handling loop.
3937 (index_location_lists): Add comment on why skip_loc_list_entry can't
3940 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3943 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3946 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3949 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3950 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3951 and -mstring-compare-inline-limit.
3953 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3956 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3957 for stack clash protection in a register whenever we need it to be in
3960 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3962 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3963 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3965 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3968 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3970 (mlong-double-): Ditto.
3971 * config/rs6000/sysv4.opt (msdata=): Ditto.
3972 (mtls-size=): Ditto.
3974 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3976 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3977 erroneous entries for
3978 "vector int vec_ldl (int, long int *)", and
3979 "vector unsigned int vec_ldl (int, unsigned long int *)".
3980 Add comments and entries for
3981 "vector bool char vec_ldl (int, bool char *)",
3982 "vector bool short vec_ldl (int, bool short *)",
3983 "vector bool int vec_ldl (int, bool int *)",
3984 "vector bool long long vec_ldl (int, bool long long *)",
3985 "vector pixel vec_ldl (int, pixel *)",
3986 "vector long long vec_ldl (int, long long *)",
3987 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3988 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3989 type tree bool_long_long_type_node and correct definition of
3990 bool_V2DI_type_node to make reference to this new type tree.
3991 (rs6000_mangle_type): Replace erroneous reference to
3992 bool_long_type_node with bool_long_long_type_node.
3993 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3994 comments to emphasize sign distinctions for char and int types and
3995 replace RS6000_BTI_bool_long constant with
3996 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3997 use of RS6000_BTI_pixel.
3998 (bool_long_type_node): Remove this macro definition.
3999 (bool_long_long_type_node): New macro definition
4001 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4003 PR rtl-optimization/85300
4004 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4005 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4006 simplify_unary_operation fails.
4008 2018-04-10 Martin Liska <mliska@suse.cz>
4010 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4011 cgraph_edge and ipa_ref.
4013 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4017 * config/i386/sse.md
4018 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4019 computation of the VEC_MERGE selector from mask.
4020 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4021 Fix decoding of the VEC_MERGE selector into mask.
4023 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4025 PR tree-optimization/85286
4026 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4028 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4030 * final.c (final_1): Set insn_last_address as well as
4031 insn_current_address.
4033 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4036 * explow.c (emit_stack_probe): Call validize_mem on memory location
4037 before passing it to gen_probe_stack. Create address operand and
4038 legitimize it for the probe_stack_address case.
4040 2018-04-09 Jan Hubicka <jh@suse.cz>
4043 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4044 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4045 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4047 (free_lang_data): Rebuild type inheritance graph.
4049 2018-04-09 Martin Sebor <msebor@redhat.com>
4051 * invoke.texi (-finline-small-functions): Mention other optimization
4053 (-findirect-inlining, -fpartial-inlining): Same.
4054 (-finline-functions-called-once): Same.
4055 (-freorder-blocks-and-partition): Same.
4057 2018-04-09 Jan Hubicka <jh@suse.cz>
4060 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4061 jumps; choose last target that matches the criteria (i.e.
4062 no partition changes for non-crossing jumps).
4063 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4064 support for redirecting crossing jumps to non-crossing.
4066 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4068 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4069 also for naked functions.
4071 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4073 * config/arc/arc.md (add_shift): New pattern.
4074 (add_shift2): Likewise.
4075 (sub_shift): Likewise.
4076 (sub_shift_cmp0_noout): Likewise.
4077 (compare_si_ashiftsi): Likewise.
4078 (xbfu_cmp0_noout): New combine pattern.
4079 (xbfu_cmp0"): Likewise.
4080 (movsi_set_cc_insn): Place the predicable variant first.
4081 (commutative_binary_cmp0_noout): Remove clobber.
4082 (commutative_binary_cmp0): New pattern.
4083 (noncommutative_binary_cmp0): Likewise.
4084 (noncommutative_binary_cmp0_noout): Likewise.
4085 (noncommutative_binary_comparison_result_used): Removed.
4086 (rsub_cmp0): New pattern.
4087 (rsub_cmp0_noout): Likewise.
4088 (extzvsi): Changed, keep only meaningful variants.
4089 (SQH, SEZ): New iterators.
4090 (SQH_postfix): New mode attribute.
4091 (SEZ_prefix): New code attribute.
4092 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4093 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4094 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4096 (noncommutative_operator): Check the availability of barrel
4099 2018-04-09 Richard Biener <rguenther@suse.de>
4101 PR tree-optimization/85284
4102 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4103 Only use the niter constraining form of simple_iv when the exit
4106 2018-04-09 Tom de Vries <tom@codesourcery.com>
4109 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4110 (define_expand "*memory_barrier"): New define_expand.
4111 (define_insn "memory_barrier"): New insn.
4113 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4115 PR rtl-optimization/80463
4116 PR rtl-optimization/83972
4117 PR rtl-optimization/83480
4119 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4120 correct producer for the insn.
4121 (tidy_control_flow): Fixup seqnos in case of debug insns.
4123 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4125 PR rtl-optimization/83913
4127 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4128 different sched-times when merging exprs.
4130 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4132 PR rtl-optimization/83962
4134 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4135 tidy_fallthru_edge and tidy_control_flow.
4137 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4139 PR rtl-optimization/83530
4141 * sel-sched.c (force_next_insn): New global variable.
4142 (remove_insn_for_debug): When force_next_insn is true, also leave only
4143 next insn in the ready list.
4144 (sel_sched_region): When the region wasn't scheduled, make another pass
4145 over it with force_next_insn set to 1.
4147 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4149 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4151 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4152 for interrupt control.
4153 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4154 functions for interrupt control.
4155 * config/nds32/nds32-intrinsic.md: Likewise.
4156 * config/nds32/nds32_intrinsic.h: Likewise.
4157 * config/nds32/nds32.h (nds32_builtins): Likewise.
4159 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4161 * config/nds32/nds32.c (nds32_init_machine_status,
4162 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4163 strict_aligned_p field.
4164 (nds32_expand_to_rtl_hook): New function.
4165 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4166 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4168 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4169 Chung-Ju Wu <jasonwucj@gmail.com>
4171 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4172 * config/nds32/nds32-n7.md: New file.
4173 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4174 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4176 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4177 * config/nds32/nds32.md (pipeline_model): Add n7.
4178 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4179 * config/nds32/pipelines.md: Include n7 settings.
4181 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4182 Chung-Ju Wu <jasonwucj@gmail.com>
4184 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4185 * config/nds32/nds32-e8.md: New file.
4186 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4187 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4189 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4190 * config/nds32/nds32.md (pipeline_model): Add e8.
4191 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4192 * config/nds32/pipelines.md: Include e8 settings.
4194 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4195 Chung-Ju Wu <jasonwucj@gmail.com>
4197 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4198 * config/nds32/nds32-n8.md: New file.
4199 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4200 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4202 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4203 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4204 * config/nds32/nds32.md (pipeline_model): Add n8.
4205 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4206 * config/nds32/pipelines.md: Include n8 settings.
4208 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4209 Chung-Ju Wu <jasonwucj@gmail.com>
4211 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4212 * config/nds32/nds32-n9-2r1w.md: New file.
4213 * config/nds32/nds32-n9-3r2w.md: New file.
4214 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4215 nds32_register_ports): New or modify for cpu n9.
4216 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4218 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4219 * config/nds32/nds32-utils.c: New file.
4220 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4221 TARGET_MUL_SLOW): Define.
4222 * config/nds32/nds32.md (pipeline_model): New attribute.
4223 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4224 New options that support cpu n9.
4225 * config/nds32/pipelines.md: Include n9 settings.
4226 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4228 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4230 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4231 information if necessary.
4232 (output_cond_branch_compare_zero): Likewise.
4233 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4234 (nds32_target_alignment): Refine for alignment.
4235 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4236 (FUNCTION_BOUNDARY): Modify.
4237 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4239 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4241 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4243 * config/nds32/constants.md (unspec_volatile_element): Add values for
4244 TLB operation and data prefetch.
4245 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4246 functions for TLB operation and data prefetch.
4247 * config/nds32/nds32-intrinsic.md: Likewise.
4248 * config/nds32/nds32_intrinsic.h: Likewise.
4249 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4250 (nds32_print_operand): Likewise.
4251 * config/nds32/nds32.h (nds32_builtins): Likewise.
4253 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4254 Andrew Pinski <pinsika@gcc.gnu.org>
4257 * match.pd: Use constant_boolean_node of correct type instead of
4258 boolean_true_node or boolean_false_node for simplifying
4259 pointer comparisons to zero.
4261 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4263 PR tree-optimization/80021
4264 * tree.c (verify_type_variant): Make error call in verify_variant_match
4265 translatable and remove final full stop.
4267 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4269 * config/nds32/constants.md (unspec_volatile_element): Add
4270 UNSPEC_VOLATILE_EH_RETURN.
4271 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4272 nds32_output_stack_pop): Support dwarf exception handling process.
4273 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4274 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4275 exception handling process.
4276 (nds32_compute_stack_frame): Likewise.
4277 (nds32_return_addr_rtx): Likewise.
4278 (nds32_initial_elimination_offset): Likewise.
4279 (nds32_expand_prologue): Likewise.
4280 (nds32_expand_epilogue): Likewise.
4281 (nds32_dynamic_chain_address): New function.
4282 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4284 (DYNAMIC_CHAIN_ADDRESS): Define.
4285 (EH_RETURN_DATA_REGNO): Define.
4286 (EH_RETURN_STACKADJ_RTX): Define.
4287 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4288 patterns for dwarf exception handling.
4290 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4292 * config/nds32/nds32.h: Clean up obsolete macros.
4294 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4296 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4297 Add enum values for particular instructions.
4298 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4299 particular intrinsic functions.
4300 * config/nds32/nds32-intrinsic.md: Likewise.
4301 * config/nds32/nds32_intrinsic.h: Likewise.
4302 * config/nds32/nds32.h (nds32_builtins): Likewise.
4303 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4304 (btst, ave): New patterns for particular instructions.
4306 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4308 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4309 Add enum values for atomic load/store and memory sync.
4310 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4312 * config/nds32/nds32-intrinsic.md: Likewise.
4313 * config/nds32/nds32_intrinsic.h: Likewise.
4314 * config/nds32/nds32.h (nds32_builtins): Likewise.
4316 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4318 PR tree-optimization/85257
4319 * fold-const.c (native_encode_vector): If not all elts could fit
4320 and off is -1, return 0 rather than offset.
4321 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4322 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4323 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4324 adjust buffer in native_interpret_expr call.
4326 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4328 * config/nds32/constants.md (unspec_volatile_element): Add cache
4329 control enum values.
4330 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4331 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4332 * config/nds32/nds32.c (nds32_cctl_names): New.
4333 (nds32_print_operand): Handle cache control register names.
4334 * config/nds32/nds32.h (nds32_builtins): New enum values.
4335 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4337 * config/nds32/nds32.md (type): Add mmu.
4338 * config/nds32/pipelines.md (simple_insn): Add mmu.
4340 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4342 * config/nds32/nds32.md (type): Remove call.
4343 * config/nds32/pipelines.md (simple_insn): Likewise.
4345 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4347 * config/nds32/constants.md (unspec_volatile_element): Add
4348 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4349 UNSPEC_VOLATILE_FMFCFG.
4350 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4351 description for fmfcfg and fmfcsr.
4352 (bdesc_1arg): Add fmtcsr.
4353 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4354 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4355 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4356 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4357 unspec_fmfcfg): New patterns.
4358 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4359 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4360 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4361 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4362 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4363 __nds32__fmfcfg): Define.
4365 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4367 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4368 intrinsic register names.
4369 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4370 intrinsic register enum values and macros.
4372 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4374 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4375 for load/store addressing form.
4376 (nds32_print_operand_address): Likewise.
4378 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4381 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4382 based on LABEL_REF. Remove useless assertion.
4383 (pic_address_needs_scratch): Fix formatting.
4384 (sparc_legitimize_pic_address): Minor tweaks.
4385 (sparc_delegitimize_address): Adjust assertion accordingly.
4386 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4387 into symbolic_operand.
4388 (movsi_high_pic_label_ref): Likewise.
4389 (movsi_lo_sum_pic_label_ref): Likewise.
4390 (movdi_pic_label_ref): Likewise.
4391 (movdi_high_pic_label_ref): Likewise.
4392 (movdi_lo_sum_pic_label_ref): Likewise.
4394 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4396 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4397 custom LIB_SPEC setup.
4399 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4400 Kito Cheng <kito.cheng@gmail.com>
4402 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4403 * config/riscv/freebsd.h: New.
4405 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4407 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4408 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4411 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4412 Kito Cheng <kito.cheng@gmail.com>
4414 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4415 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4416 * config/nds32/nds32-protos.h (nds32_output_call,
4417 nds32_output_return): Declare.
4418 * config/nds32/nds32.md: Refine all the call and return patterns.
4420 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4423 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4424 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4426 PR rtl-optimization/84872
4427 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4428 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4431 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4433 * expr.c (copy_blkmode_to_reg): Revert 254862.
4434 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4436 2018-04-06 Richard Biener <rguenther@suse.de>
4439 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4440 after seeing a component reference with an adjacent field. Treat
4441 refs to arrays at struct end of external decls similar to
4442 refs to unconstrained commons.
4444 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4447 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4448 look through SAVE_EXPRs with non-side-effects argument. Adjust
4450 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4453 2018-04-06 Richard Biener <rguenther@suse.de>
4456 * alias.c (find_base_term): New wrapper around find_base_term
4457 unwinding CSELIB_VAL_PTR changes.
4458 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4461 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4463 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4465 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4466 constant definitions.
4467 ("nop"): lr 0,0 -> nopr r0
4468 ("nop_lr0", "nop_lr1"): New insn definitions.
4470 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4472 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4473 NDS32_V3PUSH_AVAILABLE_P macro.
4475 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4476 Chung-Ju Wu <jasonwucj@gmail.com>
4478 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4479 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4480 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4481 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4482 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4483 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4484 * config/nds32/constraints.md: New constraints and checking for hard
4485 float configuration.
4486 * config/nds32/iterators.md: New mode iterator and attribute for hard
4487 float configuration.
4488 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4490 * config/nds32/nds32-fpu.md: New file.
4491 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4492 deal with hard float code generation.
4493 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4495 (abi_type, float_reg_number): New enum type.
4496 * config/nds32/nds32-predicates.c: New predicates for hard float.
4497 * config/nds32/nds32-protos.h: Declare functions for hard float.
4498 * config/nds32/nds32.c: Implementation for hard float configuration.
4499 * config/nds32/nds32.h: Definitions for hard float configuration.
4500 * config/nds32/nds32.md: Include hard float machine description and
4501 modify patterns for hard float configuration.
4502 * config/nds32/nds32.opt: New options for hard float configuration.
4503 * config/nds32/predicates.md: New predicates for hard float
4506 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4508 * common/config/nds32/nds32-common.c
4509 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4511 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4514 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4515 CONSTRUCTOR_ELT (ctor, ...)->value.
4517 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4520 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4522 2018-04-05 Tom de Vries <tom@codesourcery.com>
4525 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4528 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4529 Kito Cheng <kito.cheng@gmail.com>
4531 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4532 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4533 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4534 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4536 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4537 Kito Cheng <kito.cheng@gmail.com>
4539 * config/nds32/constraints.md (Ufe): New memory constraint.
4540 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4541 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4542 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4544 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4545 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4547 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4549 * config/nds32/nds32.md: Use optimize_size in the condition for
4550 alu-shift instructions.
4552 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4554 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4556 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4558 * config/nds32/nds32.md (negsi2): Refine pattern.
4560 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4561 Chung-Ju Wu <jasonwucj@gmail.com>
4563 * config/nds32/iterators.md (shift_rotate): New code iterator.
4564 (shift): New code attribute.
4565 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4566 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4567 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4568 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4569 bit-wise operations.
4570 (andsi3, *andsi3): Ditto.
4571 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4572 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4573 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4574 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4575 nds32_ior_operand, nds32_xor_operand): New predicates.
4577 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4579 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4580 (addsi3, subsi3): ... this.
4582 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4584 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4586 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4588 * config/nds32/nds32.md: Adjust indention.
4590 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4592 * config/nds32/nds32.md (feature): New attribute.
4594 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4596 * config/nds32/nds32.md (subtype): New attribute.
4598 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4601 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4602 expansion to perform a bitwise AND of the argument followed by a
4603 boolean negation of the result.
4605 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4607 PR rtl-optimization/84878
4608 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4609 the basic block. Assert the use reference is not artificial and that
4610 it has an associated insn.
4612 2018-04-04 Michael Matz <matz@suse.de>
4614 * builtins.c (compute_objsize): Pass correct operand
4615 to array_at_struct_end_p.
4617 2018-04-04 Richard Biener <rguenther@suse.de>
4620 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4621 from contexts for DINFO_LEVEL_TERSE and below.
4623 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4625 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4626 resiter_operand condition.
4627 * config/nds32/nds32.md (*move<mode>): Ditto.
4629 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4630 Monk Chiang <sh.chiang04@gmail.com>
4632 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4634 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4636 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4638 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4639 Kito Cheng <kito.cheng@gmail.com>
4641 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4642 nds32_cond_code_str, output_cond_branch,
4643 output_cond_branch_compare_zero, nds32_expand_cbranch,
4644 nds32_expand_cstore, nds32_expand_movcc,
4645 nds32_output_cbranchsi4_equality_zero,
4646 nds32_output_cbranchsi4_equality_reg,
4647 nds32_output_cbranchsi4_equality_reg_or_const_int,
4648 nds32_output_cbranchsi4_greater_less_zero: New functions.
4649 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4650 nds32_expand_cstore, nds32_expand_movcc,
4651 nds32_output_cbranchsi4_equality_zero,
4652 nds32_output_cbranchsi4_equality_reg,
4653 nds32_output_cbranchsi4_equality_reg_or_const_int,
4654 nds32_output_cbranchsi4_greater_less_zero): Declare.
4655 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4656 nds32_rimm11s_operand): New predicates.
4657 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4658 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4661 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4663 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4664 * config/nds32/nds32.md: Ditto.
4665 * config/nds32/pipelines.md: Ditto.
4667 2018-04-04 Richard Biener <rguenther@suse.de>
4669 PR tree-optimization/85168
4670 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4671 propagating abnormals.
4673 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4675 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4677 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4678 Kito Cheng <kito.cheng@gmail.com>
4680 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4681 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4682 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4683 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4684 * config/nds32/nds32.md (sibcall_internal): New.
4685 (sibcall_register): Remove.
4686 (sibcall_immediate): Remove.
4687 (sibcall_value_internal): New.
4688 (sibcall_value_register): Remove.
4689 (sibcall_value_immediate): Remove.
4690 * config/nds32/predicates.md (nds32_general_register_operand): New.
4691 (nds32_call_address_operand): New.
4693 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4695 PR rtl-optimization/85167
4696 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4697 bb_defs if *split_p, instead preinitialize it to NULL.
4699 PR tree-optimization/85156
4700 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4701 evaluating the argument multiple times.
4703 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4705 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4707 (_mm_cvtpd_ps): Likewise.
4708 (_mm_cvttpd_epi32): Likewise.
4709 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4710 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4711 vector, pixel, and bool following altivec.h include.
4713 2018-04-03 Martin Sebor <msebor@redhat.com>
4715 * doc/extend.texi (Common Function Attributes): Clarify.
4716 (const attribute): Likewise.
4717 (pure attribute): Likewise.
4719 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4722 * config/i386/i386.c (ix86_expand_vector_set): Use
4723 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4725 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4727 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4728 instructions when changing rounding bits to preserve precision bits
4729 in the x87 control word.
4731 2018-04-03 Martin Liska <mliska@suse.cz>
4733 PR tree-optimization/82491
4734 * rtl.h (strip_offset_and_add): Replace += suboffset with
4735 poly_uint64 () + suboffset.
4737 2018-03-29 Martin Liska <mliska@suse.cz>
4738 Martin Jambor <mjambor@suse.cz>
4741 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4742 param_type is not an integral or pointer type.
4744 2018-04-03 Richard Biener <rguenther@suse.de>
4746 * sese.h (recompute_all_dominators): Remove.
4748 2018-04-02 Martin Sebor <msebor@redhat.com>
4750 * doc/invoke.texi (-Wrestrict): Fix typos.
4752 2018-04-02 Jim Wilson <jimw@sifive.com>
4754 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4755 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4756 (<optab>di3, <optab>si3_extend): Likewise.
4757 (<optab>si3_mask, <optab>si3_mask_1): New.
4758 (<optab>di3_mask, <optab>di3_mask_1): New.
4759 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4760 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4761 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4763 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4765 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4768 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4770 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4771 (nds32_canonicalize_comparison): New function.
4773 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4774 Kito Cheng <kito.cheng@gmail.com>
4775 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4777 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4778 * config/nds32/constants.md (unspec_volatile_element): Add
4779 UNSPEC_VOLATILE_RELAX_GROUP.
4780 * config/nds32/nds32-relax-opt.c: New file.
4781 * config/nds32/nds32-predicates.c
4782 (nds32_symbol_load_store_p): New function.
4783 * config/nds32/nds32-protos.h
4784 (nds32_symbol_load_store_p): Declare function.
4785 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4786 * config/nds32/nds32.c
4787 (nds32_register_pass): New function to register pass.
4788 (nds32_register_passes): New function to register passes.
4789 * config/nds32/nds32.md (relax_group): New pattern.
4790 * config/nds32/nds32.opt (mrelax-hint): New option.
4791 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4793 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4795 * config/nds32/t-nds32: Modify files dependency.
4797 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4799 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4800 (PROFILE_HOOK): Define its implementation.
4802 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4804 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4805 type and 32-bit size.
4807 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4810 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4811 (V_128_256): New mode iterator.
4812 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4813 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4814 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4816 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4817 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4819 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4822 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4823 NaN inputs correctly.
4825 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4828 * config/rs6000/vsx.md (??r): New mode attribute.
4829 (*vsx_mov<mode>_64bit): Use it.
4830 (*vsx_mov<mode>_32bit): Likewise.
4832 2018-03-30 Martin Sebor <msebor@redhat.com>
4834 PR tree-optimization/84818
4835 * builtins.c (check_access): Use warning_n.
4837 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4840 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4842 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4845 2018-03-30 Julia Koval <julia.koval@intel.com>
4848 * x86-tune.def (movx, partial_reg_dependency): Enable for
4851 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4854 * lra-constraints.c (process_alt_operands): Move setting
4855 this_alternative_matches below.
4857 2018-03-29 Martin Liska <mliska@suse.cz>
4860 * doc/invoke.texi: Document how LTO works with debug info.
4861 Describe auto-load support of binutils. Mention 'x86-64'
4862 as valid option value of -march option.
4864 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4866 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4869 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4870 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4871 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4874 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4877 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4878 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4879 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4880 for RS6000_BTM_POWERPC64.
4881 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4882 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4883 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4888 2018-03-28 Carl Love <cel@us.ibm.com>
4891 2017-09-27 Carl Love <cel@us.ibm.com>
4893 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4894 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4895 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4898 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4900 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4901 instead of __vector bool.
4902 (_mm_max_pu8): Likewise.
4903 (_mm_min_pi16): Likewise.
4905 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4908 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4909 (DIVWEUO): Likewise.
4911 (DIVDEUO): Likewise.
4912 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4913 DIVWEUO and DIVDEUO.
4914 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4915 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4916 (div_extend): Likewise.
4917 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4919 (__builtin_divweuo): Likewise.
4920 (__builtin_divdeo): Likewise.
4921 (__builtin_divdeuo): Likewise.
4923 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4926 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4927 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4929 PR tree-optimization/82004
4930 * gimple-match-head.c (optimize_pow_to_exp): New function.
4931 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4932 Don't fold to exp if optimize_pow_to_exp is false.
4934 2018-03-28 Martin Liska <mliska@suse.cz>
4937 * calls.c (initialize_argument_information): Fix trailing space.
4938 * common.opt: Fix typo and provide better explanation for
4939 -fsanitize-coverage option.
4940 * config/i386/i386.opt: Fix typo.
4942 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4943 Martin Liska <mliska@suse.cz>
4946 * gimplify.c (asan_poison_variable): Don't do the check for
4947 gimplify_omp_ctxp here.
4948 (gimplify_decl_expr): Do it here.
4949 (gimplify_target_expr): Likewise.
4951 2018-03-28 Martin Liska <mliska@suse.cz>
4954 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4955 chkp_type_bounds_count if MPX is not enabled.
4957 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4959 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4961 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4964 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4965 function to create the function decl for complex long double
4966 multiply and divide for -mabi=ieeelongdouble.
4967 (init_float128_ieee): Call it.
4969 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4972 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4973 -fcf-protection=branch -mibt.
4974 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4976 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4979 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4981 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4984 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4985 extern array declarations.
4987 2018-03-27 Richard Biener <rguenther@suse.de>
4990 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4991 explicit single_use checks.
4993 2018-03-27 Richard Biener <rguenther@suse.de>
4995 PR tree-optimization/85082
4996 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4999 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5001 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5002 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5003 Turn on fasynchronous-unwind-tables and funwind-tables.
5005 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5008 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5009 (*bmi_blsr_<mode>_ccz): Ditto.
5011 2018-03-26 Tom de Vries <tom@codesourcery.com>
5013 PR tree-optimization/85063
5014 * omp-general.c (offloading_function_p): New function. Factor out
5016 * omp-offload.c (pass_omp_target_link::gate): ... here.
5017 * omp-general.h (offloading_function_p): Declare.
5018 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5019 with attribute omp declare target for offloading functions.
5021 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5023 PR tree-optimization/84005
5024 * tree-data-ref.h (get_base_for_alignment): Declare.
5025 * tree-data-ref.c (get_base_for_alignment_1): New function.
5026 (get_base_for_alignment): Likewise.
5027 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5028 get_base_for_alignment to find a suitable base object, instead
5029 of always using drb->base_address.
5031 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5034 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5035 known size by default.
5037 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5040 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5041 and non BLKmode operands.
5043 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5046 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5047 Clean up attributes.
5049 2018-03-23 Richard Biener <rguenther@suse.de>
5052 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5053 we are going to emit early debug for LTO.
5055 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5058 * function.c (match_asm_constraints_1): Don't optimize if input
5059 doesn't satisfy general_operand predicate for output's mode.
5062 * alias.c (write_dependence_p): Don't require for x_canonicalized
5063 non-VOIDmode if x has VOIDmode.
5066 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5067 just don't try to optimize it rather than assert it never happens.
5069 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5071 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5072 macro expansions for definition of ST_INTERNAL_<mode> and
5073 LD_INTERNAL_<mode> builtins.
5074 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5076 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5078 (altivec_expand_st_builtin): Likewise.
5079 (altivec_expand_builtin): Remove calls to deleted functions.
5080 (rs6000_address_for_altivec): Delete this function.
5081 * config/rs6000/vector.md: Remove expands for
5082 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5084 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5087 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5088 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5089 re-computing once computed.
5090 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5091 (arm_init_machine_status): Initialize
5092 machine->static_chain_stack_bytes.
5094 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5097 * doc/extend.texi: Add four new prototypes for vec_ld.
5098 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5099 definitions for more logical presentation.
5100 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5101 entries for V1TI variants of __builtin_altivec_ld builtin.
5102 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5103 handling of V1TI variant of LVX icode pattern.
5104 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5105 (rs6000_gimple_fold_builtin): Likewise.
5106 (altivec_init_builtins): Add code to define
5107 __builtin_altivec_lvx_v1ti function.
5109 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5112 * function.c (match_asm_constraints_1): Don't do the optimization
5113 if input isn't a REG, SUBREG, MEM or constant.
5115 2018-03-22 Tom de Vries <tom@codesourcery.com>
5117 PR tree-optimization/84956
5118 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5119 bb_has_abnormal_pred.
5121 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5124 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5125 DECL_INITIAL (decl) to decl at the end.
5126 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5129 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5131 * doc/extend.texi (__builtin_tgmath): Document when complex
5132 integer types are treated as _Complex _Float64.
5134 2018-03-21 Tom de Vries <tom@codesourcery.com>
5136 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5138 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5140 PR tree-optimization/84960
5141 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5142 if it is ENTRY block, move them into single succ of ENTRY in that case.
5144 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5146 PR tree-optimization/84811
5147 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5148 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5149 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5150 (known_subrange_p): Update accordingly. Cast each value involved
5151 in the size comparison, rather than casting the result of the
5154 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5156 PR tree-optimization/84982
5157 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5158 by flipping the least significant bit rather than all bits from
5159 bitpos to bitpos + bitsize - 1.
5161 2018-03-21 Nathan Sidwell <nathan@acm.org>
5163 * doc/extend.texi (Deprecated Features): Remove mention of
5164 long-deleted deprecations.
5166 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5169 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5170 * configure: Regenerate.
5172 2018-03-21 Tom de Vries <tom@codesourcery.com>
5174 PR tree-optimization/83126
5175 * tree-parloops.c (num_phis): New function.
5176 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5178 2018-03-21 Nathan Sidwell <nathan@acm.org>
5180 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5181 mention anon-struct/union members and trailing attributes.
5183 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5185 PR tree-optimization/84969
5186 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5187 builtin memset partitions if they set different rhs values.
5189 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5191 PR rtl-optimization/84989
5192 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5193 VEC_DUPLICATE with scalar result mode.
5195 2018-03-21 Martin Liska <mliska@suse.cz>
5198 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5199 not intended return statement.
5201 2018-03-21 Martin Liska <mliska@suse.cz>
5204 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5205 (chkp_find_bound_slots_1): Limit number of iterations.
5207 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5210 * Minor grammar fixes for x86 options.
5212 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5215 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5216 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5218 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5221 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5222 (altivec_lvx_<mode>_1op): Likewise.
5223 (altivec_stvx_<mode>_2op): Likewise.
5224 (altivec_stvx_<mode>_1op): Likewise.
5225 (altivec_lvx_<VM2:mode>): New define_expand.
5226 (altivec_stvx_<VM2:mode>): Likewise.
5227 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5228 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5229 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5230 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5231 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5232 (rs6000_gen_lvx): Likewise.
5233 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5234 (altivec_expand_stv_builtin): Likewise.
5235 (altivec_expand_builtin): Likewise.
5236 * config/rs6000/vector.md: Likewise.
5238 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5241 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5244 2018-03-20 Richard Biener <rguenther@suse.de>
5247 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5248 sign-conversions as zero, fall back to standard scalar_stmt
5251 2018-03-20 Martin Liska <mliska@suse.cz>
5254 * predict.c (rebuild_frequencies): Handle case when we have
5255 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5257 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5260 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5261 flag_section_anchors.
5262 * varasm.c (use_blocks_for_decl_p): Remove hack for
5263 dw2_force_const_mem.
5266 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5268 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5269 be created, use lowpart_subreg of operands[0] rather than operands[0]
5271 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5272 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5273 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5274 and n constraint instead of aarch64_shift_imm_di and Usd.
5275 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5276 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5278 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5281 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5282 to favor GPR over NEON registers.
5283 (<shift>di3_neon): Likewise.
5285 2018-03-20 Tom de Vries <tom@codesourcery.com>
5288 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5289 (nvptx_process_pars): Emit bar.sync asap and alap.
5291 2018-03-20 Tom de Vries <tom@codesourcery.com>
5294 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5295 seen_label if seen_label is already set.
5297 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5300 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5301 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5302 Use 1U instead of 1. Formatting fixes.
5305 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5306 instead of TREE_TYPE (s1) for the return value.
5308 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5310 PR tree-optimization/84946
5311 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5312 bitsize + bitsize in poly_uint64 rather than poly_int64.
5315 * dwarf2asm.c: Include fold-const.c.
5316 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5317 of decl rather than decl itself.
5319 PR rtl-optimization/84643
5320 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5322 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5325 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5326 calling assemble_variable.
5328 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5331 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5332 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5334 2018-03-19 Jim Wilson <jimw@sifive.com>
5337 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5338 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5339 (riscv_first_stack_step): Likewise.
5340 (riscv_option_override): Use STACK_BOUNDARY instead of
5342 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5344 (BIGGEST_ALIGNMENT): Set to 128.
5345 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5346 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5349 2018-03-19 Richard Biener <rguenther@suse.de>
5351 PR tree-optimization/84933
5352 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5353 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5355 2018-03-19 Richard Biener <rguenther@suse.de>
5357 PR tree-optimization/84859
5358 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5359 (cond_if_else_store_replacement): Perform sinking operation on
5360 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5361 Generalize what a BB with a single eligible store is.
5363 2018-03-19 Richard Biener <rguenther@suse.de>
5365 PR tree-optimization/84929
5366 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5367 chrec_is_positive against non-chrec arg.
5369 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5372 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5374 2018-03-18 Martin Liska <mliska@suse.cz>
5376 PR rtl-optimization/84635
5377 * regrename.c (build_def_use): Use matches_mode only when
5380 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5382 PR tree-optimization/84913
5383 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5384 vectorize chains of COND_EXPRs.
5386 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5388 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5390 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5392 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5394 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5396 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5398 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5399 Kito Cheng <kito.cheng@gmail.com>
5401 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5402 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5403 (nds32_adjust_reg_alloc_order): New function.
5404 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5406 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5408 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5409 nds32_print_operand, nds32_print_operand_address): Use
5410 HOST_WIDE_INT_PRINT_DEC instead.
5412 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5414 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5416 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5419 * config/i386/i386.c (initial_ix86_tune_features,
5420 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5422 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5423 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5424 rather than 1u << ix86_tune. Formatting fix.
5425 (ix86_option_override_internal): Change ix86_arch_mask from
5426 unsigned int to unsigned HOST_WIDE_INT, initialize to
5427 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5428 (ix86_function_specific_restore): Likewise.
5430 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5433 * postreload.c (reload_combine_recognize_pattern): Perform
5434 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5435 truncate_int_for_mode the result for the destination's mode.
5438 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5439 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5441 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5444 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5447 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5448 regno_allocno_class_array and sorted_pseudos.
5449 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5450 insns where regno is used.
5452 2018-03-16 Martin Liska <mliska@suse.cz>
5455 * multiple_target.c (create_dispatcher_calls): Redirect
5456 reference in the symbol table.
5458 2018-03-16 Martin Liska <mliska@suse.cz>
5461 * multiple_target.c (create_dispatcher_calls): Redirect also
5464 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5468 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5469 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5470 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5472 2018-03-16 Julia Koval <julia.koval@intel.com>
5474 * doc/invoke.texi (Skylake Server): Add CLWB.
5475 Cannonlake): Remove CLWB.
5477 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5479 PR tree-optimization/84841
5480 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5482 (FLOAT_ONE_CONST_TYPE): Define.
5483 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5484 (sort_by_operand_rank): Put entries with higher constant_type last
5485 rather than first to match comments.
5487 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5489 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5492 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5495 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5496 check and use address of target always.
5498 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5501 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5502 (indirect_thunk_bnd_needed): Likewise.
5503 (indirect_thunks_used): Likewise.
5504 (indirect_thunks_bnd_used): Likewise.
5505 (indirect_return_needed): New.
5506 (indirect_return_bnd_needed): Likewise.
5507 (output_indirect_thunk_function): Add a bool argument for
5509 (output_indirect_thunk_function): Don't generate alias for
5510 function return thunk.
5511 (ix86_code_end): Call output_indirect_thunk_function to generate
5512 function return thunks.
5513 (ix86_output_function_return): Set indirect_return_bnd_needed
5514 and indirect_return_needed instead of indirect_thunk_bnd_needed
5515 and indirect_thunk_needed.
5517 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5519 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5520 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5521 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5523 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5524 Paul Hua <paul.hua.gm@gmail.com>
5527 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5529 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5531 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5532 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5535 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5538 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5539 instead of GET_MODE_SIZE when comparing Units.
5541 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5544 * varasm.c (hash_section): Return an unchangeble hash value
5545 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5546 Return !aarch64_can_use_per_function_literal_pools_p ().
5548 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5551 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5552 rather than address of cmode as last argument to prepare_cmp_insn.
5554 2018-03-15 Julia Koval <julia.koval@intel.com>
5556 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5557 F_AVX512VNNI, F_AVX512BITALG): New.
5559 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5562 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5563 insn for floating-point loads and stores.
5565 2018-03-14 Carl Love <cel@us.ibm.com>
5567 * config/rs6000/rs6000-c.c: Add macro definitions for
5568 ALTIVEC_BUILTIN_VEC_PERMXOR.
5569 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5570 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5571 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5572 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5574 * config/doc/extend.texi: Add prototypes for vec_permxor.
5576 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5579 * diagnostic-show-locus.c (class layout_point): Convert m_line
5580 from int to linenum_type.
5581 (line_span::comparator): Use linenum "compare" function when
5582 comparing line numbers.
5583 (test_line_span): New function.
5584 (layout_range::contains_point): Convert param "row" from int to
5586 (layout_range::intersects_line_p): Likewise.
5587 (layout::will_show_line_p): Likewise.
5588 (layout::print_source_line): Likewise.
5589 (layout::should_print_annotation_line_p): Likewise.
5590 (layout::print_annotation_line): Likewise.
5591 (layout::print_leading_fixits): Likewise.
5592 (layout::annotation_line_showed_range_p): Likewise.
5593 (struct line_corrections): Likewise for field m_row.
5594 (line_corrections::line_corrections): Likewise for param "row".
5595 (layout::print_trailing_fixits): Likewise.
5596 (layout::get_state_at_point): Likewise.
5597 (layout::get_x_bound_for_row): Likewise.
5598 (layout::print_line): Likewise.
5599 (diagnostic_show_locus): Likewise for locals "last_line" and
5601 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5602 * input.c (selftest::test_linenum_comparisons): New function.
5603 (selftest::input_c_tests): Call it.
5604 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5605 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5606 * selftest.h (ASSERT_GT): New macro.
5607 (ASSERT_GT_AT): New macro.
5608 (ASSERT_LT): New macro.
5609 (ASSERT_LT_AT): New macro.
5611 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5613 PR rtl-optimization/84780
5614 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5616 2018-03-14 Martin Liska <mliska@suse.cz>
5618 * tree.c (record_node_allocation_statistics): Use
5619 get_stats_node_kind.
5620 (get_stats_node_kind): New function extracted from
5621 record_node_allocation_statistics.
5622 (free_node): Use get_stats_node_kind.
5624 2018-03-14 Richard Biener <rguenther@suse.de>
5626 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5627 that the value-set of ANTIC_IN doesn't grow.
5630 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5632 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5633 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5635 2018-03-14 Julia Koval <julia.koval@intel.com>
5637 * config.gcc (icelake-client, icelake-server): New.
5639 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5640 (initial_ix86_arch_features): Ditto.
5641 (PTA_SKYLAKE): Add SGX.
5642 (PTA_ICELAKE): Remove.
5643 (PTA_ICELAKE_CLIENT): New.
5644 (PTA_ICELAKE_SERVER): New.
5645 (ix86_option_override_internal): Split up icelake on icelake client and
5647 (get_builtin_code_for_version): Ditto.
5648 (fold_builtin_cpu): Ditto.
5649 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5650 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5651 * config/i386/i386.h (processor_type): Ditto.
5652 * doc/invoke.texi: Ditto.
5654 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5657 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5658 INTEGER_CST offset, add it together with bitpos / 8 and
5659 sign extend based on POINTER_SIZE.
5663 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5666 * config/i386/constraints.md (Yc): New register constraint.
5667 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5668 Use Yc constraint for alternative 2 of operand 0. Remove
5669 preferred_for_speed attribute.
5671 2018-03-14 Richard Biener <rguenther@suse.de>
5673 PR tree-optimization/84830
5674 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5675 with the old one to avoid oscillations.
5677 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5680 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5682 (assign_by_spills): Return a flag of reload assignment failure.
5683 Do not process the reload assignment failures. Do not spill other
5684 reload pseudos if they has the same reg class. Update n if
5686 (lra_assign): Add a return arg. Set up from the result of
5687 assign_by_spills call.
5688 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5689 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5690 usage_insns if it is not NULL.
5691 (spill_hard_reg_in_range): New function.
5692 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5693 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5694 function prototypes.
5695 (lra_assign): Change prototype.
5696 * lra.c (lra): Add code to deal with fails by splitting hard reg
5699 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5701 * config/riscv/riscv.opt (mrelax): New option.
5702 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5703 "norelax" when riscv_mrelax is disabled.
5704 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5706 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5709 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5710 reassociation for int modes.
5712 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5714 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5715 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5717 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5718 * config/aarch64/aarch64-sve.md
5719 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5720 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5721 (*extend<mode><Vwide>2): Rename to...
5722 (aarch64_sve_extend<mode><Vwide>2): ...this.
5723 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5724 renaming the old pattern to...
5725 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5727 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5728 define_expand, renaming the old pattern to...
5729 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5730 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5731 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5732 account when deciding which SVE instruction the optab should use.
5733 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5735 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5737 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5738 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5739 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5740 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5741 (tlsdesc_small_<mode>): Turn a define_expand and use
5742 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5743 (tlsdesc_small_advsimd_<mode>): ...this.
5744 (tlsdesc_small_sve_<mode>): New pattern.
5746 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5748 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5749 (UNSPEC_UMUL_HIGHPART): New constants.
5750 (MUL_HIGHPART): New int iteraor.
5751 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5752 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5754 (*<su>mul<mode>3_highpart): New define_insn.
5756 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5759 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5762 2018-03-13 Martin Liska <mliska@suse.cz>
5765 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5767 (sem_item_optimizer::~sem_item_optimizer): Release it.
5768 (sem_item_optimizer::merge_classes): Register variable aliases.
5769 (sem_item_optimizer::fixup_pt_set): New function.
5770 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5771 * ipa-icf.h: Declare new variables and functions.
5773 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5776 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5777 integer_pow2p@2 and test integer_pow2p in condition.
5778 (A < 0 ? C : 0): Similarly for @1.
5781 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5782 characters starting at p contain '\0' character, don't look beyond
5786 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5787 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5790 * reg-stack.c (change_stack): Change update_end var from int to
5791 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5792 also call set_block_for_insn on the newly added insns and rescan.
5795 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5796 on the last operand.
5799 * tree.c (stabilize_reference_1): Return save_expr (e) for
5800 STATEMENT_LIST even if it doesn't have side-effects.
5802 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5804 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5806 2018-03-12 Renlin Li <renlin.li@arm.com>
5808 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5809 aarch64_output_scalar_simd_mov_immediate.
5811 2018-03-12 Martin Sebor <msebor@redhat.com>
5813 PR tree-optimization/83456
5814 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5815 for perfectly overlapping calls to memcpy.
5816 (gimple_fold_builtin_memory_chk): Same.
5817 (gimple_fold_builtin_strcpy): Handle no-warning.
5818 (gimple_fold_builtin_stxcpy_chk): Same.
5819 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5821 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5823 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5824 parameter. Use it for SFmode.
5825 (rs6000_function_arg_advance_1): Adjust.
5826 (rs6000_function_arg): Adjust.
5827 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5829 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5831 PR rtl-optimization/84169
5832 PR rtl-optimization/84780
5833 * combine.c (can_combine_p): Check for a 2-insn combination whether
5834 the destination register is used between the two insns, too.
5836 2018-03-12 Richard Biener <rguenther@suse.de>
5838 PR tree-optimization/84803
5839 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5840 for refs DR analysis didn't process.
5842 2018-03-12 Richard Biener <rguenther@suse.de>
5844 PR tree-optimization/84777
5845 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5846 force-vectorize loops ignore whether we are optimizing for size.
5848 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5850 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5851 (TARGET_MD_ASM_ADJUST): Define.
5853 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5854 Kito Cheng <kito.cheng@gmail.com>
5855 Chung-Ju Wu <jasonwucj@gmail.com>
5857 * config/nds32/nds32.c (nds32_compute_stack_frame,
5858 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5859 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5860 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5861 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5862 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5863 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5864 * config/nds32/nds32.md (prologue, epilogue): Use macro
5865 NDS32_V3PUSH_AVAILABLE_P to do checking.
5867 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5870 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5871 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5872 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5873 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5874 addition of most attributes on !orig_type_die or the attribute not
5875 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5877 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5878 Chung-Ju Wu <jasonwucj@gmail.com>
5880 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5882 * config/nds32/nds32.opt (mvh): New option.
5884 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5885 Chung-Ju Wu <jasonwucj@gmail.com>
5887 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5889 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5890 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5893 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5894 Chung-Ju Wu <jasonwucj@gmail.com>
5896 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5898 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5899 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5901 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5902 Kito Cheng <kito.cheng@gmail.com>
5903 Chung-Ju Wu <jasonwucj@gmail.com>
5905 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5906 UNSPEC_FFMISM and UNSPEC_FLMISM.
5907 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5908 for ffb, ffmism and flmism.
5909 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5910 (unspec_ffmism): Ditto.
5911 (unspec_flmism): Ditto.
5912 (nds32_expand_builtin_impl): Check if string extension is available.
5913 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5914 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5916 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5919 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5922 * lra-assigns.c (assign_by_spills): Return a flag of reload
5923 assignment failure. Do not process the reload assignment
5924 failures. Do not spill other reload pseudos if they has the same
5926 (lra_assign): Add a return arg. Set up from the result of
5927 assign_by_spills call.
5928 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5929 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5930 usage_insns if it is not NULL.
5931 (spill_hard_reg_in_range): New function.
5932 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5933 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5934 function prototypes.
5935 (lra_assign): Change prototype.
5936 * lra.c (lra): Add code to deal with fails by splitting hard reg
5939 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5942 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5944 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5947 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5948 (dw_val_node): Add val_symbolic_view.
5949 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5950 (symview_upper_bound): New.
5951 (new_line_info_table): Initialize symviews_since_reset.
5952 (dwarf2out_source_line): Count symviews_since_reset and set
5953 symview_upper_bound.
5954 (dw_val_equal_p): Handle symview.
5955 (add_AT_symview): New.
5956 (print_dw_val): Handle symview.
5957 (attr_checksum, attr_checksum_ordered): Likewise.
5958 (same_dw_val_p, size_of_die): Likewise.
5959 (value_format, output_die): Likewise.
5960 (add_high_low_attributes): Use add_AT_symview for entry_view.
5961 (dwarf2out_finish): Reset symview_upper_bound, clear
5964 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5967 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5968 Add strict argument and use it.
5969 (rs6000_split_multireg_move): Update for new strict argument.
5970 (mem_operand_gpr): Disallow all non-offsettable addresses.
5971 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5973 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5976 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5977 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5978 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5981 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5982 decl, use remap_type if we want to use the type.
5984 2018-03-09 Martin Sebor <msebor@redhat.com>
5986 PR tree-optimization/84526
5987 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5989 (builtin_access::generic_overlap): Be prepared to handle non-array
5992 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5994 PR rtl-optimization/84682
5995 * lra-constraints.c (process_address_1): Check is_address flag
5996 for address constraints.
5997 (process_alt_operands): Likewise.
5998 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5999 preprocess_constraints.
6000 * recog.h (preprocess_constraints): Add oploc parameter.
6002 * recog.c (preprocess_constraints): Test address_operand for
6003 CT_ADDRESS constraints.
6005 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6008 * lra-assigns.c (assign_by_spills): Return a flag of reload
6009 assignment failure. Do not process the reload assignment
6010 failures. Do not spill other reload pseudos if they has the same
6012 (lra_assign): Add a return arg. Set up from the result of
6013 assign_by_spills call.
6014 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6015 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6016 usage_insns if it is not NULL.
6017 (spill_hard_reg_in_range): New function.
6018 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6019 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6020 function prototypes.
6021 (lra_assign): Change prototype.
6022 * lra.c (lra): Add code to deal with fails by splitting hard reg
6025 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6028 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6029 Accept complain bool parameter. Only emit errors if it is true.
6030 (arm_parse_cpu_option_name): Likewise.
6031 (arm_target_thumb_only): Adjust callers of the above.
6032 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6033 prototype to take a default true bool parameter.
6034 (arm_parse_arch_option_name): Likewise.
6036 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6037 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6041 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6042 * configure: Regenerate.
6043 * configure.ac ("linker --version-script option"): New.
6044 ("linker soname option"): New.
6046 2018-03-09 Richard Biener <rguenther@suse.de>
6048 PR tree-optimization/84775
6049 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6050 immediate uses of predicate stmts and mark them modified.
6053 PR tree-optimization/84178
6054 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6056 (version_loop_for_if_conversion): Delay update_ssa call.
6057 (tree_if_conversion): Delay update_ssa until after predicate
6060 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6063 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6064 when the function accesses prior frames.
6066 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6069 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6070 gen_llsym, otherwise call maybe_gen_llsym.
6073 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6074 has ',' character inside of it.
6076 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6079 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6080 as clobbering CC_REGNUM.
6082 2018-03-08 Richard Biener <rguenther@suse.de>
6085 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6086 (follow_copies_to_constant): Do not follow SSA names registered
6089 2018-03-08 Richard Biener <rguenther@suse.de>
6091 PR tree-optimization/84178
6092 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6094 (version_loop_for_if_conversion): Delay update_ssa call.
6095 (tree_if_conversion): Delay update_ssa until after predicate
6098 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6100 PR tree-optimization/84178
6101 * tree-if-conv.c (release_bb_predicate): Remove the
6102 the assertion that the stmts have NULL use_ops.
6103 Discard the statements, asserting that they haven't
6104 yet been added to a BB.
6106 2018-03-08 Richard Biener <rguenther@suse.de>
6108 PR tree-optimization/84746
6109 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6110 (phi_translate): Pass in destination ANTIC_OUT set.
6111 (phi_translate_1): Likewise. For a simplified result lookup
6112 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6113 (phi_translate_set): Adjust.
6114 (do_pre_regular_insertion): Likewise.
6115 (do_pre_partial_partial_insertion): Likewise.
6117 2018-03-08 Martin Liska <mliska@suse.cz>
6119 PR gcov-profile/84735
6120 * doc/gcov.texi: Document usage of profile files.
6121 * gcov-io.h: Document changes in the format.
6123 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6127 * dwarf2out.c (struct dw_line_info_table): Update comments for
6129 (FORCE_RESET_NEXT_VIEW): New.
6130 (FORCE_RESETTING_VIEW_P): New.
6131 (RESETTING_VIEW_P): Check for -1 too.
6132 (ZERO_VIEW_P): Likewise.
6133 (new_line_info_table): Force-reset next view.
6134 (dwarf2out_begin_function): Likewise.
6135 (dwarf2out_source_line): Simplify zero_view_p initialization.
6136 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6137 view directly. Omit view when omitting .loc at line 0.
6139 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6141 PR tree-optimization/84740
6142 * tree-switch-conversion.c (process_switch): Call build_constructors
6143 only if info.phi_count is non-zero.
6145 PR tree-optimization/84739
6146 * tree-tailcall.c (find_tail_calls): Check call arguments against
6147 DECL_ARGUMENTS (current_function_decl) rather than
6148 DECL_ARGUMENTS (func) when checking for tail recursion.
6150 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6152 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6153 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6154 Volker Reichelt's entry and add entries for people that perform
6155 GCC fuzzy testing and report numerous bugs.
6157 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6160 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6161 readonly data in sdata, if that is disabled.
6162 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6163 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6164 -mreadonly-in-sdata option.
6166 2018-03-07 Martin Sebor <msebor@redhat.com>
6168 PR tree-optimization/84468
6169 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6170 basic block when looking for nul assignment.
6172 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6175 * except.h (output_function_exception_table): Adjust prototype.
6176 * except.c (output_function_exception_table): Remove FNNAME parameter
6177 and add SECTION parameter. Ouput one part of the table at a time.
6178 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6179 the first part of the exception table and emit unwind directives.
6180 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6181 (i386_pe_seh_cold_init): Likewise.
6182 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6183 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6184 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6185 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6186 * config/i386/winnt.c: Include except.h.
6187 (struct seh_frame_state): Add reg_offset, after_prologue and
6188 in_cold_section fields.
6189 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6190 (i386_pe_seh_cold_init): New function.
6191 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6192 to seh->in_cold_section.
6193 (seh_emit_push): Record the offset of the push.
6194 (seh_emit_save): Record the offet of the save.
6195 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6196 Test seh->after_prologue to disregard the epilogue.
6197 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6198 (i386_pe_end_cold_function): New function.
6200 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6203 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6204 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6207 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6209 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6210 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6213 * multiple_target.c: Include tree-inline.h and intl.h.
6214 (expand_target_clones): Diagnose and fail if node->definition and
6215 !tree_versionable_function_p (node->decl).
6217 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6219 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6221 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6222 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6223 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6225 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6228 * combine.c (try_combine): Use reg_or_subregno instead of handling
6229 just paradoxical SUBREGs and REGs.
6231 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6233 * config/arc/arc.c (arc_finalize_pic): Remove function.
6234 (arc_must_save_register): We use single base PIC register, remove
6235 checks to save/restore the PIC register.
6236 (arc_expand_prologue): Likewise.
6237 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6239 (arc_verify_short): Likewise.
6240 (arc_attr_type): Likewise.
6241 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6242 (walk_stores): Likewise.
6243 (arc_address_cost): Make it static.
6244 (arc_verify_short): Likewise.
6245 (branch_dest): Likewise.
6246 (arc_attr_type): Likewise.
6247 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6248 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6249 (arc_final_prescan_insn): Remove inserting the nops due to
6250 hardware hazards. It is done in reorg step.
6251 (insn_length_variant_t): Remove.
6252 (insn_length_parameters_t): Likewise.
6253 (arc_insn_length_parameters): Likewise.
6254 (arc_get_insn_variants): Likewise.
6255 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6257 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6260 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6263 PR tree-optimization/84687
6264 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6266 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6268 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6270 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6271 Rename to ppc_speculation_barrier.
6272 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6273 __builtin_ppc_speculation_barrier.
6275 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6278 * combine.c (combine_simplify_rtx): Don't try to simplify if
6279 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6282 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6284 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6285 to 32 bytes when compiling for POWER9.
6287 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6290 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6291 regparm >= 3 with no arg reg available also for calls with
6292 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6295 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6297 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6299 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6302 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6304 2018-03-05 Richard Biener <rguenther@suse.de>
6306 PR tree-optimization/84486
6307 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6308 When inserting a __builtin_assume_aligned call set the LHS
6309 SSA name alignment info accordingly.
6311 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6313 PR tree-optimization/84114
6314 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6315 Avoid reassociation of FLOAT_MODE addition.
6317 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6319 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6320 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6321 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6322 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6323 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6324 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6325 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6327 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6328 __builtin_ia32_wbinvd): New builtins.
6329 (SPECIAL_ARGS2): New.
6330 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6331 (SPECIAL_ARGS2): New.
6332 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6333 (ix86_valid_target_attribute_inner_p): Ditto.
6334 (ix86_init_mmx_sse_builtins): Add special_args2.
6335 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6336 TARGET_WBNOINVD_P): New.
6337 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6338 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6339 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6340 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6341 * config/i386/pconfigintrin.h: New file.
6342 * config/i386/wbnoinvdintrin.h: Ditto.
6343 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6344 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6346 2018-03-05 Richard Biener <rguenther@suse.de>
6348 PR tree-optimization/84670
6349 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6351 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6352 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6353 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6354 doesn't grow if all successors (recursively) were visited at least
6357 2018-03-05 Richard Biener <rguenther@suse.de>
6359 PR tree-optimization/84650
6360 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6361 if executed in the loop pipeline.
6363 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6365 * doc/configfiles.texi (Configuration Files): Move info about
6366 conditionalizing $target-protos.h to...
6367 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6368 differs from $target-protos.h.
6370 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6371 Chung-Ju Wu <jasonwucj@gmail.com>
6373 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6374 * config/nds32/nds32-multiple.md (setmemsi): Define.
6375 * config/nds32/nds32-memory-manipulation.c
6376 (nds32_gen_dup_4_byte_to_word_value): New.
6377 (emit_setmem_word_loop): New.
6378 (emit_setmem_byte_loop): New.
6379 (nds32_expand_setmem_loop): New.
6380 (nds32_expand_setmem_loop_v3m): New.
6381 (nds32_expand_setmem_unroll): New.
6382 (nds32_expand_setmem): New.
6384 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6385 Chung-Ju Wu <jasonwucj@gmail.com>
6387 * config/nds32/nds32-memory-manipulation.c
6388 (nds32_emit_load_store): New.
6389 (nds32_emit_post_inc_load_store): New.
6390 (nds32_emit_mem_move): New.
6391 (nds32_emit_mem_move_block): New.
6392 (nds32_expand_movmemsi_loop_unknown_size): New.
6393 (nds32_expand_movmemsi_loop_known_size): New.
6394 (nds32_expand_movmemsi_loop): New.
6395 (nds32_expand_movmemsi_unroll): New.
6396 (nds32_expand_movmemqi): Rename ...
6397 (nds32_expand_movmemsi): ... to this.
6398 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6399 (movmemsi): ... to this.
6400 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6401 (nds32_expand_movmemsi): ... to this.
6403 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6404 Monk Chiang <sh.chiang04@gmail.com>
6405 Chung-Ju Wu <jasonwucj@gmail.com>
6407 * config/nds32/nds32-protos.h
6408 (nds32_expand_load_multiple): New arguments.
6409 (nds32_expand_store_multiple): Ditto.
6410 (nds32_valid_multiple_load_store): Rename ...
6411 (nds32_valid_multiple_load_store_p): ... to this.
6412 * config/nds32/nds32-memory-manipulation.c
6413 (nds32_expand_load_multiple): Refine implementation.
6414 (nds32_expand_store_multiple): Ditto.
6415 * config/nds32/nds32-multiple.md
6416 (load_multiple): Update nds32_expand_load_multiple interface.
6417 (store_multiple): Update nds32_expand_store_multiple interface.
6418 * config/nds32/nds32-predicates.c
6419 (nds32_valid_multiple_load_store): Rename ...
6420 (nds32_valid_multiple_load_store_p): ... to this and refine
6422 * config/nds32/predicates.md
6423 (nds32_load_multiple_and_update_address_operation): New predicate.
6424 (nds32_store_multiple_and_update_address_operation): New predicate.
6426 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6427 Chung-Ju Wu <jasonwucj@gmail.com>
6429 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6430 (combo): New attribute.
6431 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6433 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6435 * config/nds32/nds32.opt: Change -mcmodel= default value.
6437 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6438 Monk Chiang <sh.chiang04@gmail.com>
6439 Chung-Ju Wu <jasonwucj@gmail.com>
6441 * config/nds32/constants.md (unspec_element): New enum.
6442 * config/nds32/constraints.md (Umw): New constraint.
6443 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6444 * config/nds32/nds32-intrinsic.md: Likewise.
6445 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6446 (nds32_valid_smw_lwm_base_p): New.
6447 (nds32_output_smw_single_word): New.
6448 (nds32_output_lmw_single_word): New.
6449 (nds32_expand_unaligned_load): New.
6450 (nds32_expand_unaligned_store): New.
6451 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6452 (nds32_output_smw_single_word): Declare.
6453 (nds32_output_lmw_single_word): Declare.
6454 (nds32_expand_unaligned_load): Declare.
6455 (nds32_expand_unaligned_store): Declare.
6456 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6457 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6458 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6459 NDS32_BUILTIN_UASTORE_DW.
6460 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6463 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6464 Kito Cheng <kito.cheng@gmail.com>
6465 Chung-Ju Wu <jasonwucj@gmail.com>
6467 * config/nds32/nds32-intrinsic.c
6468 (nds32_expand_builtin_null_ftype_reg): Delete.
6469 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6470 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6471 (nds32_read_argument): New.
6472 (nds32_legitimize_target): Ditto.
6473 (nds32_legitimize_argument): Ditto.
6474 (nds32_check_constant_argument): Ditto.
6475 (nds32_expand_unop_builtin): Ditto.
6476 (nds32_expand_unopimm_builtin): Ditto.
6477 (nds32_expand_binop_builtin): Ditto.
6478 (nds32_builtin_decl_impl): Ditto.
6479 (builtin_description): Ditto.
6480 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6481 (nds32_init_builtins_impl): Ditto.
6482 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6483 (nds32_builtin_decl): New.
6484 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6485 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6487 2018-03-02 Jeff Law <law@redhat.com>
6489 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6490 (redundant_insn, fill_simple_delay_slots): Likewise.
6491 (fill_slots_from_thread): Likewise.
6492 * resource.c (mark_referenced_resources): Likewise.
6493 (mark_set_resources, find_dead_or_set_registers): Likewise.
6495 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6497 * substring-locations.h (format_warning_va): Formatting fix for
6499 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6501 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6502 * substring-locations.c: Include intl.h.
6503 (format_warning_va): Turned into small wrapper around
6504 format_warning_n_va, renamed to ...
6505 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6506 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6508 (format_warning_at_substring_n): New function.
6509 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6510 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6511 format_warning_at_substring with just a shorter name instead of
6512 const function pointer.
6513 (fmtwarn_n): New function.
6514 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6515 appropriate, get rid of all the fmtstr temporaries, move conditionals
6516 with G_() wrapped string literals directly into fmtwarn arguments,
6517 cast dir.len to (int), formatting fixes.
6519 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6521 * doc/invoke.texi: Remove "Cilk Plus" references.
6523 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6524 Richard Biener <rguenther@suse.de>
6527 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6528 for error or warning attributes if CALL_FROM_THUNK_P is set.
6531 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6534 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6535 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6538 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6539 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6541 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6542 instead of -1U in last predictors element's probability member.
6544 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6547 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6548 arguments if they are comparable.
6550 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6552 PR tree-optimization/84634
6553 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6554 masks and masked_loop_p with a single loop_masks, making sure it's
6555 null for bb vectorization.
6557 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6559 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6560 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6561 loop->force_vectorize to check whether there is no alias.
6563 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6566 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6568 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6570 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6571 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6572 instead of a loop around prev_real_insn.
6573 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6577 * config/i386/i386.c (ix86_print_operand): Use conditional
6578 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6581 2018-03-02 Richard Biener <rguenther@suse.de>
6583 PR tree-optimization/84427
6584 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6585 (bitmap_set_subtract_values): Rewrite to handle multiple
6588 (prune_clobbered_mems): Likewise.
6589 (phi_translate): Take edge instead of pred/phiblock.
6590 (phi_translate_1): Likewise.
6591 (phi_translate_set): Likewise. Insert all translated
6592 exprs for a value into the set, keeping possibly multiple
6593 expressions per value.
6594 (compute_antic_aux): Adjust for phi_translate changes.
6595 When intersecting union the expressions and prune those
6596 not in the final value set, keeping possibly multiple
6597 expressions per value. Do not use value-insertion
6598 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6599 all expressions. Add verification that the value-sets
6600 only shrink during iteration.
6601 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6602 (do_pre_regular_insertion): Likewise.
6603 (do_pre_partial_partial_insertion): Likewise.
6605 2018-03-02 Richard Biener <rguenther@suse.de>
6608 * config/darwin.c (saved_debug_info_level): New static global.
6609 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6610 (darwin_asm_lto_end): Restore debug info generation settings.
6612 2018-03-01 Martin Liska <mliska@suse.cz>
6615 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6618 2018-03-01 Richard Biener <rguenther@suse.de>
6621 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6622 type annotation with LTO when debug was disabled at compile-time.
6624 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6626 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6628 (mips_final_postscan_insn): Likewise.
6630 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6632 PR rtl-optimization/84528
6633 * alias.c (init_alias_target): Add commentary.
6634 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6635 a unique base value if the frame pointer is not eliminated
6636 to the stack pointer.
6638 2018-03-01 Tom de Vries <tom@codesourcery.com>
6640 PR rtl-optimization/83327
6641 * lra-int.h (hard_regs_spilled_into): Declare.
6642 * lra.c (hard_regs_spilled_into): Define.
6643 (init_reg_info): Init hard_regs_spilled_into.
6644 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6645 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6646 (process_bb_lives): Handle hard_regs_spilled_into.
6647 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6648 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6650 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6652 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6653 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6654 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6655 * config/rs6000/aix72.h: New file.
6657 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6659 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6660 instead of warning_at with conditional singular and plural messages
6664 * stor-layout.c (update_alignment_for_field): For
6665 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6666 && !DECL_PACKED (field), do the alignment update, just use
6667 only desired_align instead of MAX (type_align, desired_align)
6669 (place_field): Don't do known_align < desired_align handling
6670 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6671 is non-NULL, instead do it after rli->prev_field handling and
6672 only if not within a bitfield word. For DECL_PACKED (field)
6673 use type_align of BITS_PER_UNIT.
6675 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6677 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6678 superfluous parentheses and trailing spaces.
6680 2018-02-28 Richard Biener <rguenther@suse.de>
6682 PR tree-optimization/84584
6683 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6684 SCoPs with fake exit edge.
6686 2018-02-28 Martin Liska <mliska@suse.cz>
6689 * timevar.c (timer::print): Fix format to properly print 100%
6692 2018-02-28 Richard Biener <rguenther@suse.de>
6695 * genmatch.c (capture_info::walk_match): Do not mark
6696 captured expressions without operands as expr_p given
6697 they act more like predicates and should be subject to
6698 "lost tail" side-effect preserving.
6700 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6702 PR rtl-optimization/81611
6703 * auto-inc-dec.c (attempt_change): Move dead note from
6704 mem_insn if it's the next use of regno
6705 (find_address): Take address use of reg holding
6706 non-incremented value. Add parm to limit search to the named
6708 (merge_in_block): Attempt to use a mem insn that is the next
6709 use of the original regno.
6711 2018-02-27 Martin Sebor <msebor@redhat.com>
6714 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6715 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6717 2018-02-27 Martin Sebor <msebor@redhat.com>
6719 PR translation/84207
6720 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6721 n argument to unsigned HOST_WIDE_INT.
6722 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6723 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6724 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6725 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6727 2018-02-27 Richard Biener <rguenther@suse.de>
6729 PR tree-optimization/84512
6730 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6731 Do not use the estimate returned from record_stmt_cost for
6732 the scalar iteration cost but sum properly using add_stmt_cost.
6734 2018-02-27 Richard Biener <rguenther@suse.de>
6736 PR tree-optimization/84466
6737 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6738 Adjust last change to less strictly validate use operands.
6740 2018-02-27 Martin Liska <mliska@suse.cz>
6742 PR gcov-profile/84548
6743 * gcov.c (process_file): Allow partial overlap and consider it
6744 also as group functions.
6745 (output_lines): Properly calculate range of lines for a group.
6747 2018-02-27 Martin Liska <mliska@suse.cz>
6749 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6750 'ggc' suffixes. Change first column width.
6751 (timer::print): Fix formatting of the column.
6753 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6755 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6756 preserve inline entry blocks for the sake of debug inline
6757 entry point markers alone.
6758 (remove_unused_locals): Suggest in comments a better place to
6759 force the preservation of inline entry blocks that are
6760 otherwise unused, but do not preserve them.
6762 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6764 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6766 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6769 * config/i386/constraints.md (Bs): Replace
6770 ix86_indirect_branch_register with
6771 TARGET_INDIRECT_BRANCH_REGISTER.
6773 * config/i386/i386.md (indirect_jump): Likewise.
6774 (tablejump): Likewise.
6775 (*sibcall_memory): Likewise.
6776 (*sibcall_value_memory): Likewise.
6777 Peepholes of indirect call and jump via memory: Likewise.
6778 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6779 (*sibcall_value_GOT_32): Likewise.
6780 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6781 (GOT_memory_operand): Likewise.
6782 (call_insn_operand): Likewise.
6783 (sibcall_insn_operand): Likewise.
6784 (GOT32_symbol_operand): Likewise.
6785 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6787 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6789 PR rtl-optimization/83496
6790 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6791 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6792 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6793 redundant insn, if any.
6794 (relax_delay_slots): Likewise.
6795 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6797 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6799 PR tree-optimization/83965
6800 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6801 that grouped statements are part of a reduction chain. Return
6802 true if the statement is not marked as a reduction itself but
6804 (vect_recog_dot_prod_pattern): Don't check whether the statement
6805 is part of a group here.
6806 (vect_recog_sad_pattern): Likewise.
6807 (vect_recog_widen_sum_pattern): Likewise.
6809 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6812 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6814 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6817 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6819 (ix86_output_indirect_function_return): New prototype.
6820 (ix86_split_simple_return_pop_internal): Likewise.
6821 * config/i386/i386.c (indirect_return_via_cx): New.
6822 (indirect_return_via_cx_bnd): Likewise.
6823 (indirect_thunk_name): Handle return va CX_REG.
6824 (output_indirect_thunk_function): Create alias for
6825 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6826 (ix86_output_indirect_jmp): Remove the bool argument.
6827 (ix86_output_indirect_function_return): New function.
6828 (ix86_split_simple_return_pop_internal): Likewise.
6829 * config/i386/i386.md (*indirect_jump): Don't pass false
6830 to ix86_output_indirect_jmp.
6831 (*tablejump_1): Likewise.
6832 (simple_return_pop_internal): Change it to define_insn_and_split.
6833 Call ix86_split_simple_return_pop_internal to split it for
6835 (simple_return_indirect_internal): Call
6836 ix86_output_indirect_function_return instead of
6837 ix86_output_indirect_jmp.
6839 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6842 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6843 memset and value initialization afterwards.
6845 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6847 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6849 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6852 * common/config/aarch64/aarch64-common.c
6853 (aarch_option_optimization_table[]): Switch
6854 off fomit-frame-pointer
6856 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6857 Chung-Ju Wu <jasonwucj@gmail.com>
6859 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6861 (store_multiple): Ditto.
6863 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6865 * config.gcc: Add --with-cpu support for nds32 target.
6866 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6867 * config/nds32/nds32.opt: Add -mcpu= option.
6869 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6871 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6872 isel=yes): Warn for these deprecated options.
6874 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6876 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6877 ISA_2_5_MASKS_EMBEDDED.
6879 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6881 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6882 p->max as pointers rather than using iterative_hash_expr.
6884 2018-02-23 Carl Love <cel@us.ibm.com>
6886 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6887 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6889 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6890 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6891 P8V_BUILTIN_VEC_VUNSIGNED2.
6893 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6896 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6897 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6899 * lra-constraints.c (curr_insn_transform): Set up
6900 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6902 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6903 * lra-eliminations.c (spill_pseudos): Ditto.
6904 (process_insn_for_elimination): Ditto.
6905 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6906 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6907 LRA_NON_CLOBBERED_ALT.
6909 2018-02-22 Martin Sebor <msebor@redhat.com>
6911 PR tree-optimization/84480
6912 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6913 to maybe_diag_stxncpy_trunc. Call it.
6914 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6915 from gimple_fold_builtin_strcpy. Print inlining stack.
6916 (handle_builtin_stxncpy): Print inlining stack.
6917 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6919 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6922 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6923 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6924 and -fcheck-pointer-bounds are used together.
6925 (indirect_thunk_prefix): New enum.
6926 (indirect_thunk_need_prefix): New function.
6927 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6928 "_nt" instead of "_bnd" for NOTRACK prefix.
6929 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6930 (output_indirect_thunk_function): Likewise.
6932 (ix86_code_end): Update output_indirect_thunk_function calls.
6933 (ix86_output_indirect_branch_via_reg): Replace
6934 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6935 (ix86_output_indirect_branch_via_push): Likewise.
6936 (ix86_output_function_return): Likewise.
6937 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6938 incompatible with -fcf-protection=branch and
6939 -fcheck-pointer-bounds.
6941 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6944 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6945 Change gcc_assert call to output_operand_lossage.
6947 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6949 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6951 2018-02-22 DJ Delorie <dj@redhat.com>
6952 Sebastian Perta <sebastian.perta@renesas.com>
6953 Oleg Endo <olegendo@gcc.gnu.org>
6955 * config/rx/rx.c (rx_rtx_costs): New function.
6956 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6958 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6960 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6962 2018-02-22 Martin Liska <mliska@suse.cz>
6965 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6966 Add "native" as a possible value.
6968 2018-02-22 Martin Liska <mliska@suse.cz>
6971 * config/i386/i386.c (ix86_option_override_internal):
6972 Add "native" as a possible value for -march and -mtune.
6974 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6977 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6978 to all type variants.
6980 PR tree-optimization/84503
6981 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6982 width as info->bitpos + info->bitsize - start.
6983 (merged_store_group::merge_overlapping): Simplify width computation.
6984 (check_no_overlap): New function.
6985 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6986 start + width and last_order of the group, fail if check_no_overlap
6988 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6989 to group if check_no_overlap fails.
6991 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6993 * config/rs6000/altivec.md: Delete contraint arguments to
6994 define_expand, define_split, and define_peephole2, and in
6995 define_insn_and_split if always unused.
6996 * config/rs6000/darwin.md: Ditto.
6997 * config/rs6000/dfp.md: Ditto.
6998 * config/rs6000/rs6000.md: Ditto.
6999 * config/rs6000/sync.md: Ditto.
7000 * config/rs6000/vector.md: Ditto.
7001 * config/rs6000/vsx.md: Ditto.
7003 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7005 * config/rs6000/altivec.md: Write output control strings as braced
7006 blocks instead of double-quoted strings.
7007 * config/rs6000/darwin.md: Ditto.
7008 * config/rs6000/rs6000.md: Ditto.
7009 * config/rs6000/vector.md: Ditto.
7010 * config/rs6000/vsx.md: Ditto.
7012 2018-02-21 Jason Merrill <jason@redhat.com>
7014 PR c++/84314 - ICE with templates and fastcall attribute.
7015 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7017 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7019 * ipa-cp.c (determine_versionability): Fix comment typos.
7021 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7024 * ipa-cp.c (determine_versionability): Do not version functions caling
7027 2018-02-21 Martin Liska <mliska@suse.cz>
7030 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7031 Add "native" as a possible value.
7032 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7033 the macro when native cpu detection is available.
7035 2018-02-21 Martin Liska <mliska@suse.cz>
7038 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7039 Add "native" as a possible value.
7040 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7041 when native cpu detection is available.
7043 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7044 Martin Sebor <msebor@redhat.com>
7046 PR tree-optimization/84478
7047 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7049 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7050 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7051 support which is conservatively correct, for 2 only stay conservative
7052 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7053 argument to the 2 argument get_range_strlen, adjust 6 arg
7054 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7056 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7057 (gimple_fold_builtin_strlen): Pass true as last argument to
7060 2018-02-20 Martin Sebor <msebor@redhat.com>
7063 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7064 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7065 (builtin_memref::builtin_memref): Factor out parts into
7066 set_base_and_offset and call it.
7068 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7071 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7072 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7073 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7074 search at the associated MODE_INT.
7076 2018-02-20 Jeff Law <law@redhat.com>
7079 PR tree-optimization/81592
7081 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7082 for range data rather than using global data.
7083 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7084 range data rather than using global data.
7085 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7086 pass it to children as needed.
7087 (struct directive::fmtresult): Similarly.
7088 (struct directive::set_width): Similarly.
7089 (struct directive::set_precision): Similarly.
7090 (format_integer, format_directive, parse_directive): Similarly.
7091 (format_none): Accept unnamed vr_values parameter.
7092 (format_percent, format_floating, format_character): Similarly.
7093 (format_string, format_plain): Similarly.
7094 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7095 the EVRP range analyzer for range data rather than using global data.
7096 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7097 gimple-ssa-evrp-analyze.h
7098 (class sprintf_dom_walker): Add after_dom_children member function.
7099 Add evrp_range_analyzer member.
7100 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7101 range analyzer as needed.
7102 (sprintf_dom_walker::after_dom_children): New member function.
7103 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7105 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7106 (evrp_range_analyzer::pop_to_marker): Likewise.
7108 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7110 PR tree-optimization/84419
7111 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7112 with the required type if its current type is compatible but
7115 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7118 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7119 after vectorization.
7121 2018-02-20 Martin Liska <mliska@suse.cz>
7124 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7125 possible values if we don't have a hint.
7127 2018-02-20 Martin Liska <mliska@suse.cz>
7131 * final.c (shorten_branches): Build align_tab array with one
7133 * opts.c (finish_options): Add alignment option limit check.
7134 (MAX_CODE_ALIGN): Likewise.
7135 (MAX_CODE_ALIGN_VALUE): Likewise.
7136 * doc/invoke.texi: Document maximum allowed option value for
7137 all -falign-* options.
7139 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7142 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7143 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7144 * var-tracking.c (emit_note_insn_var_location): Remove all references
7145 to NOTE_INSN_CALL_ARG_LOCATION.
7146 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7147 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7148 Use copy_rtx_if_shared.
7149 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7150 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7151 (dwarf2out_var_location): Remove handling of
7152 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7154 * final.c (final_scan_insn): Remove all references to
7155 NOTE_INSN_CALL_ARG_LOCATION.
7156 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7157 before dumping final insns.
7158 * except.c (emit_note_eh_region_end): Remove all references to
7159 NOTE_INSN_CALL_ARG_LOCATION.
7160 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7161 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7162 * config/arc/arc.c (hwloop_optimize): Likewise.
7163 * config/arm/arm.c (create_fix_barrier): Likewise.
7164 * config/s390/s390.c (s390_chunkify_start): Likewise.
7165 * config/sh/sh.c (find_barrier): Likewise.
7166 * config/i386/i386.c (rest_of_insert_endbranch,
7167 ix86_seh_fixup_eh_fallthru): Likewise.
7168 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7169 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7170 * config/frv/frv.c (frv_function_prologue): Likewise.
7171 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7173 (note_outside_basic_block_p): Remove all references to
7174 NOTE_INSN_CALL_ARG_LOCATION.
7175 * gengtype.c (adjust_field_rtx_def): Likewise.
7176 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7178 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7179 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7182 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7185 PR tree-optimization/84452
7186 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7187 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7190 2018-02-19 Martin Liska <mliska@suse.cz>
7193 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7195 2018-02-19 Martin Liska <mliska@suse.cz>
7196 Richard Sandiford <richard.sandiford@linaro.org>
7198 PR tree-optimization/82491
7199 * gimple-fold.c (get_base_constructor): Make earlier bail out
7202 2018-02-19 Carl Love <cel@us.ibm.com>
7204 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7205 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7207 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7208 P8V_BUILTIN_VEC_NEG.
7210 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7212 * config/rl78/rl78.md (movdf): New define expand.
7214 2018-02-19 Martin Liska <mliska@suse.cz>
7217 * doc/invoke.texi: Fix typo.
7218 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7220 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7222 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7223 handle rs6000_single_float and rs6000_double_float specially for
7226 2018-02-16 Jeff Law <law@redhat.com>
7228 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7231 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7234 * ipa-inline.c (inline_small_functions): Fix a typo.
7236 2018-02-16 Nathan Sidwell <nathan@acm.org>
7238 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7240 2018-02-16 Carl Love <cel@us.ibm.com>
7242 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7243 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7244 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7245 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7246 expansion to P8V_BUILTIN_VEC_FLOAT2.
7248 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7250 PR rtl-optimization/70023
7251 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7252 src_regno into account.
7254 2018-02-16 Carl Love <cel@us.ibm.com>
7256 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7257 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7258 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7259 * config/rs6000/rs6000.c: Remove case statements for
7260 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7261 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7262 and P9V_BUILTIN_VEC_VINSERT4B.
7263 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7264 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7265 * config/rs6000/vsx.md:
7266 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7269 2018-02-16 Carl Love <cel@us.ibm.com>
7271 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7273 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7275 * config/rs6000/rs6000-c.c: Add the definitions for
7276 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7277 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7278 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7279 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7280 definition for insert4b and define insn *insert3b_internal.
7281 * doc/extend.texi: Add documentation for vec_extract4b.
7283 2018-02-16 Nathan Sidwell <nathan@acm.org>
7285 * doc/extend.texi (Backwards Compatibility): Mention friend
7286 injection. Note for-scope is deprecated.
7287 * doc/invoke.texi (-ffriend-injection): Deprecate.
7289 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7291 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7292 that moved to I2, also allow destinations that are a paradoxical
7293 subreg (instead of a normal reg).
7295 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7298 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7301 2018-02-16 Richard Biener <rguenther@suse.de>
7303 PR tree-optimization/84037
7304 PR tree-optimization/84016
7306 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7307 Adjust vec_construct for the fact we need additional higher latency
7308 128bit inserts for AVX256 and AVX512 vector builds.
7309 (ix86_add_stmt_cost): Scale vector construction cost for
7312 2018-02-16 Richard Biener <rguenther@suse.de>
7314 PR tree-optimization/84417
7315 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7316 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7317 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7319 2018-02-16 Martin Liska <mliska@suse.cz>
7322 * internal-fn.def (ASAN_CHECK): Set proper flags.
7323 (ASAN_MARK): Likewise.
7325 2018-02-16 Julia Koval <julia.koval@intel.com>
7327 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7328 from PTA_CANNONLAKE.
7330 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7333 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7334 Use ++iter rather than iter++ for std::list iterators.
7335 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7336 defer deleting them until all nodes in the forest are processed. Do
7337 free even leaf nodes. Change to_process into auto_vec.
7340 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7341 * vec.h (vec_default_construct): Use memset instead of placement new
7342 if BROKEN_VALUE_INITIALIZATION is defined.
7343 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7344 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7347 PR rtl-optimization/83723
7348 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7349 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7350 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7352 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7354 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7356 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7358 PR rtl-optimization/81443
7359 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7360 from inner REGs to paradoxical SUBREGs.
7362 2018-02-16 Richard Biener <rguenther@suse.de>
7364 PR tree-optimization/84399
7365 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7366 For operands we can analyze at their definition make sure we can
7367 analyze them at each use as well.
7369 2018-02-16 Richard Biener <rguenther@suse.de>
7371 PR tree-optimization/84190
7372 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7373 volatile accesses if the decl isn't volatile.
7375 2018-02-15 Jason Merrill <jason@redhat.com>
7377 PR c++/84314 - ICE with templates and fastcall attribute.
7378 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7379 TYPE_CANONICAL on an existing type.
7381 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7383 PR tree-optimization/84383
7384 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7385 dstoff nor call operand_equal_p if dstbase is NULL.
7387 PR tree-optimization/84334
7388 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7389 also a CONSTANT_CLASS_P, punt.
7391 2018-02-14 Jim Wilson <jimw@sifive.com>
7393 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7394 first SMALL_OPERAND check. New local min_second_step. Move assert
7395 to where locals are set. Add TARGET_RVC support.
7396 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7398 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7400 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7402 2018-02-14 Martin Sebor <msebor@redhat.com>
7404 PR tree-optimization/83698
7405 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7406 arrays constrain the offset range to their bounds.
7407 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7408 (builtin_access::overlap): Avoid setting the size of overlap if it's
7410 (maybe_diag_overlap): Also consider arrays when deciding what values
7411 of offsets to include in diagnostics.
7413 2018-02-14 Martin Sebor <msebor@redhat.com>
7416 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7417 that correspond to the kind of a declaration.
7419 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7422 * config/pa/pa.md: Load address of PIC label using the linkage table
7423 if the label is nonlocal.
7425 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7427 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7428 warning message if user requests -maltivec=be.
7429 * doc/invoke.texi: Document deprecation of -maltivec=be.
7431 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7434 * config/rs6000/rs6000-c.c: Update definitions for
7435 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7436 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7438 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7441 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7442 add _get_ssp intrinsics. Remove argument from
7443 __builtin_ia32_rdssp[d|q].
7444 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7445 * config/i386/i386-builtin.def: Remove argument from
7446 __builtin_ia32_rdssp[d|q].
7447 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7448 ix86_expand_special_args_builtin for _rdssp[d|q].
7449 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7450 Clear register before usage.
7451 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7452 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7454 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7456 PR tree-optimization/84357
7457 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7458 operand 1 of an ARRAY_REF too.
7460 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7463 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7464 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7466 (set_of_reg): New struct.
7467 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7468 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7469 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7471 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7472 Split into bitclr, bitset, bitinvert patterns if appropriate.
7473 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7474 use rx_fuse_in_memory_bitop.
7475 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7476 to named insn, correct maximum insn length.
7478 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7481 * machmode.def: Define a complex mode for PARTIAL_INT.
7482 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7484 * doc/rtl.texi: Document CSPImode.
7485 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7487 (msp430_hard_regno_nregs_with_padding): Likewise.
7489 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7492 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7494 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7496 PR rtl-optimization/84169
7497 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7498 we generated a parallel as new i3 and we split that to new i2 and i3
7499 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7500 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7501 those to i2, not i1. Partially rewrite this scan code.
7503 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7506 * stor-layout.c (place_field): For variable length fields, adjust
7507 offset_align afterwards not just based on the field's alignment,
7508 but also on the size.
7511 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7512 of exps and logs in the use_exp2 case.
7514 2018-02-13 Jeff Law <law@redhat.com>
7516 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7519 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7522 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7526 * common.opt (gas-loc-support, gas-locview-support): New.
7527 (ginline-points, ginternal-reset-location-views): New.
7528 * doc/invoke.texi: Document them. Use @itemx where intended.
7529 (gvariable-location-views): Adjust.
7530 * target.def (reset_location_view): New.
7531 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7532 (TARGET_RESET_LOCATION_VIEW): New.
7533 * doc/tm.texi: Rebuilt.
7534 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7535 (dwarf2out_default_as_locview_support): New.
7536 (output_asm_line_debug_info): Use option variables.
7537 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7538 (output_loc_list): Likewise.
7539 (add_high_low_attributes): Check option variables.
7540 Don't output entry view attribute in strict mode.
7541 (gen_inlined_subroutine_die): Check option variables.
7542 (dwarf2out_inline_entry): Likewise.
7543 (init_sections_and_labels): Likewise.
7544 (dwarf2out_early_finish): Likewise.
7545 (maybe_reset_location_view): New, from...
7546 (dwarf2out_var_location): ... here. Call it.
7547 * debug.h (dwarf2out_default_as_loc_support): Declare.
7548 (dwarf2out_default_as_locview_support): Declare.
7549 * hooks.c (hook_int_rtx_insn_0): New.
7550 * hooks.h (hook_int_rtx_insn_0): Declare.
7551 * toplev.c (process_options): Take -gas-loc-support and
7552 -gas-locview-support from dwarf2out. Enable
7553 -gvariable-location-views by default only with locview
7554 assembler support. Enable -ginternal-reset-location-views by
7555 default only if the target defines the corresponding hook.
7556 Enable -ginline-points by default if location views are
7557 enabled; force it disabled if statement frontiers are
7559 * tree-inline.c (expand_call_inline): Check option variables.
7560 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7562 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7564 PR tree-optimization/84321
7565 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7566 handling. Also check whether the anti-range contains any values
7567 that satisfy the mask; switch to a VR_RANGE if not.
7569 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7572 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7574 2018-02-13 Martin Jambor <mjambor@suse.cz>
7577 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7578 of call statements, also set location of a load to a temporary.
7580 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7582 * config/rl78/rl78.c (add_vector_labels): New function.
7583 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7584 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7585 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7586 which checks that no arguments are passed.
7587 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7588 * doc/extend.texi: Documentation for the new attribute.
7590 2018-02-13 Andreas Schwab <schwab@suse.de>
7592 * config/riscv/linux.h (CPP_SPEC): Define.
7594 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7597 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7598 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7599 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7600 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7601 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7602 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7603 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7604 temporarily for AES and PCLMUL builtins.
7606 PR tree-optimization/84339
7607 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7608 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7612 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7613 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7614 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7616 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7618 * omp-simd-clone.h: New file.
7619 * omp-simd-clone.c: Include omp-simd-clone.h.
7620 (expand_simd_clones): No longer static.
7621 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7622 cgraph.h and omp-simd-clone.h.
7623 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7624 (vect_recog_widen_shift_pattern): Formatting fix.
7625 (vect_pattern_recog_1): Don't check optab for calls.
7628 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7629 operands[2] into a REG before using gen_lowpart on it.
7631 2018-02-12 Jeff Law <law@redhat.com>
7634 * config/sh/sh.c (find_barrier): Consider a sibling call
7637 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7638 successfully back substituting a reg.
7640 2018-02-12 Richard Biener <rguenther@suse.de>
7642 PR tree-optimization/84037
7643 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7644 parameter, move visited init to caller.
7645 (vect_slp_analyze_operations): Separate cost from validity
7646 check, initialize visited once for all instances.
7647 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7649 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7651 (vect_model_promotion_demotion_cost): Likewise.
7652 (vectorizable_bswap): Guard cost modeling with !slp_node
7653 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7655 (vectorizable_call): Likewise.
7656 (vectorizable_conversion): Likewise.
7657 (vectorizable_assignment): Likewise.
7658 (vectorizable_shift): Likewise.
7659 (vectorizable_operation): Likewise.
7660 (vectorizable_store): Likewise.
7661 (vectorizable_load): Likewise.
7662 (vectorizable_condition): Likewise.
7663 (vectorizable_comparison): Likewise.
7665 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7668 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7669 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7670 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7672 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7675 * params.def (inline-min-speedup): Increase from 8 to 15.
7676 (max-inline-insns-auto): Decrease from 40 to 30.
7677 * ipa-split.c (consider_split): Add some buffer for function to
7678 be considered inlining candidate.
7679 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7682 2018-02-12 Richard Biener <rguenther@suse.de>
7684 PR tree-optimization/84037
7685 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7686 matched stmts if we cannot swap the non-matched ones.
7688 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7690 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7691 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7692 _mm_maskz_scalef_round_ss): New intrinsics.
7693 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7694 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7695 __builtin_ia32_scalefss_round): Remove.
7696 (__builtin_ia32_scalefsd_mask_round,
7697 __builtin_ia32_scalefss_mask_round): New intrinsics.
7698 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7699 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7700 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7701 "<round_constraint>")): Changed to ...
7702 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7703 "<round_scalar_constraint>")): ... this.
7704 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7705 %0, %1, %2<round_op3>}"): Changed to ...
7706 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7707 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7708 %2<round_scalar_mask_op3>}"): ... this.
7709 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7711 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7713 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7714 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7715 (_mm_maskz_sqrt_round_ss): New intrinsics.
7716 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7717 (__builtin_ia32_sqrtsd_mask_round)
7718 (__builtin_ia32_sqrtss_mask_round): New builtins.
7719 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7720 (__builtin_ia32_sqrtss_round): Remove.
7721 (__builtin_ia32_sqrtsd_mask_round)
7722 (__builtin_ia32_sqrtss_mask_round): New builtins.
7723 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7724 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7725 ((match_operand:VF_128 1 "vector_operand"
7726 "xBm,<round_constraint>")): Changed to ...
7727 ((match_operand:VF_128 1 "vector_operand"
7728 "xBm,<round_scalar_constraint>")): ... this.
7729 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7730 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7731 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7732 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7733 %<iptr>1<round_scalar_mask_op3>}): ... this.
7734 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7735 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7737 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7740 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7741 Cast vec_cmpeq result to correct type.
7742 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7743 Cast vec_cmpgt result to correct type.
7745 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7747 * final.c (final_scan_insn_1): Renamed from...
7748 (final_scan_insn): ... this. New wrapper, to recover
7749 seen from the outermost call in recursive ones.
7750 * config/sparc/sparc.c (output_return): Drop seen from call.
7751 (output_sibcall): Likewise.
7752 * config/visium/visium.c (output_branch): Likewise.
7754 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7756 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7759 2018-02-10 Alan Modra <amodra@gmail.com>
7762 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7763 Specify LR as an input.
7765 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7768 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7769 remove_member_access_dummy_vars): New functions.
7770 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7771 lower_omp_1, execute_lower_omp): Use them.
7773 PR rtl-optimization/84308
7774 * shrink-wrap.c (spread_components): Release todo vector.
7776 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7778 PR rtl-optimization/57193
7779 * ira-color.c (struct allocno_color_data): Add member
7780 conflict_allocno_hard_prefs.
7781 (update_conflict_allocno_hard_prefs): New.
7782 (bucket_allocno_compare_func): Add a preference based on
7783 conflict_allocno_hard_prefs.
7784 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7785 (color_allocnos): Remove a dead code. Initiate
7786 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7788 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7791 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7792 constraint from =wa to wa. Avoid a subreg on the output operand,
7793 instead use a pseudo and subreg it in a move.
7794 (p9_xxbrd_<mode>): Changed to ...
7795 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7796 (p9_xxbrd_v2df): New expander.
7797 (p9_xxbrw_<mode>): Changed to ...
7798 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7799 (p9_xxbrw_v4sf): New expander.
7801 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7803 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7805 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7808 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7809 multiply in 32-bit mode.
7810 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7811 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7814 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7816 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7817 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7818 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7819 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7821 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7824 * dwarf2out.c (is_trivial_indirect_ref): New function.
7825 (dwarf2out_late_global_decl): Do not generate a location
7826 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7827 and that are not defined in the current unit.
7829 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7831 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7832 instead of a libcall for UNORDERED.
7834 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7837 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7838 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7840 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7843 * config/s390/s390.c (s390_set_current_function): Invoke
7844 s390_indirect_branch_settings also if fndecl didn't change.
7846 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7848 * config/rs6000/rs6000.md (blockage): Set length to zero.
7850 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7852 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7854 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7857 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7858 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7862 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7863 PARALLEL incoming that failed vt_get_decl_and_offset check.
7866 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7867 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7869 (get_variable_section): For decls in named .bss* sections pass true as
7870 second argument to bss_initializer_p.
7872 2018-02-09 Marek Polacek <polacek@redhat.com>
7873 Jakub Jelinek <jakub@redhat.com>
7876 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7877 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7878 Sync some changes from cxx_fold_indirect_ref.
7880 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7882 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7884 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7885 (BLOCK_INLINE_ENTRY_LABEL): New.
7886 (dwarf2out_var_location): Disregard inline entry markers.
7887 (inline_entry_data): New struct.
7888 (inline_entry_data_hasher): New hashtable type.
7889 (inline_entry_data_hasher::hash): New.
7890 (inline_entry_data_hasher::equal): New.
7891 (inline_entry_data_table): New variable.
7892 (add_high_low_attributes): Add DW_AT_entry_pc and
7893 DW_AT_GNU_entry_view attributes if a pending entry is found
7894 in inline_entry_data_table. Add old entry_pc attribute only
7895 if debug nonbinding markers are disabled.
7896 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7897 markers are enabled.
7898 (block_within_block_p, dwarf2out_inline_entry): New.
7899 (dwarf2out_finish): Check that no entries remained in
7900 inline_entry_data_table.
7901 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7902 (final_scan_insn, notice_source_line): Likewise.
7903 (rest_of_clean_state): Skip inline entry markers.
7904 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7906 * gimple.c (gimple_build_debug_inline_entry): New.
7907 * gimple.h (enum gimple_debug_subcode): Add
7908 GIMPLE_DEBUG_INLINE_ENTRY.
7909 (gimple_build_debug_inline_entry): Declare.
7910 (gimple_debug_inline_entry_p): New.
7911 (gimple_debug_nonbind_marker_p): Adjust.
7912 * insn-notes.def (INLINE_ENTRY): New.
7913 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7914 inline entry marker notes.
7915 (print_insn): Likewise.
7916 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7917 (INSN_DEBUG_MARKER_KIND): Likewise.
7918 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7919 * tree-inline.c (expand_call_inline): Build and insert
7920 debug_inline_entry stmt.
7921 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7922 inline entry blocks early, if nonbind markers are enabled.
7923 (dump_scope_block): Dump fragment info.
7924 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7925 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7926 (gimple_build_debug_inline_entry): New.
7927 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7928 Enable/disable inline entry points too.
7929 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7930 (DEBUG_INSN): Describe inline entry markers.
7932 * common.opt (gvariable-location-views): New.
7933 (gvariable-location-views=incompat5): New.
7934 * config.in: Rebuilt.
7935 * configure: Rebuilt.
7936 * configure.ac: Test assembler for view support.
7937 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7938 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7939 * dwarf2out.c (var_loc_view): New typedef.
7940 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7941 (dwarf2out_locviews_in_attribute): New.
7942 (dwarf2out_locviews_in_loclist): New.
7943 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7944 (enum dw_line_info_opcode): Add LI_adv_address.
7945 (struct dw_line_info_table): Add view.
7946 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7947 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7948 (zero_view_p): New variable.
7949 (ZERO_VIEW_P): New macro.
7950 (output_asm_line_debug_info): New.
7951 (struct var_loc_node): Add view.
7952 (add_AT_view_list, AT_loc_list): New.
7953 (add_var_loc_to_decl): Add view param. Test it against last.
7954 (new_loc_list): Add view params. Record them.
7955 (AT_loc_list_ptr): Handle loc and view lists.
7956 (view_list_to_loc_list_val_node): New.
7957 (print_dw_val): Handle dw_val_class_view_list.
7958 (size_of_die): Likewise.
7959 (value_format): Likewise.
7960 (loc_list_has_views): New.
7961 (gen_llsym): Set vl_symbol too.
7962 (maybe_gen_llsym, skip_loc_list_entry): New.
7963 (dwarf2out_maybe_output_loclist_view_pair): New.
7964 (output_loc_list): Output view list or entries too.
7965 (output_view_list_offset): New.
7966 (output_die): Handle dw_val_class_view_list.
7967 (output_dwarf_version): New.
7968 (output_compilation_unit_header): Use it.
7969 (output_skeleton_debug_sections): Likewise.
7970 (output_rnglists, output_line_info): Likewise.
7971 (output_pubnames, output_aranges): Update version comments.
7972 (output_one_line_info_table): Output view numbers in asm comments.
7973 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7974 Call maybe_gen_llsym.
7975 (loc_list_from_tree_1): Adjust.
7976 (add_AT_location_description): Create view list attribute if
7977 needed, check it's absent otherwise.
7978 (convert_cfa_to_fb_loc_list): Adjust.
7979 (maybe_emit_file): Call output_asm_line_debug_info for test.
7980 (dwarf2out_var_location): Reset views as needed. Precompute
7981 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7982 attribute. Set view.
7983 (new_line_info_table): Reset next view.
7984 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7985 (dwarf2out_source_line): Likewise. Output view resets and labels to
7986 the assembler, or select appropriate line info opcodes.
7987 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7988 (optimize_string_length): Catch it. Adjust.
7989 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7990 dw_val_class_view_list, and remove it if no longer needed.
7991 (hash_loc_list): Hash view numbers.
7992 (loc_list_hasher::equal): Compare them.
7993 (optimize_location_lists): Check whether a view list symbol is
7994 needed, and whether the locview attribute is present, and
7995 whether they match. Remove the locview attribute if no longer
7997 (index_location_lists): Call skip_loc_list_entry for test.
7998 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7999 Use output_dwarf_version.
8000 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8001 (struct dw_val_node): Add val_view_list.
8002 * final.c (SEEN_NEXT_VIEW): New.
8003 (set_next_view_needed): New.
8004 (clear_next_view_needed): New.
8005 (maybe_output_next_view): New.
8006 (final_start_function): Rename to...
8007 (final_start_function_1): ... this. Take pointer to FIRST,
8008 add SEEN parameter. Emit param bindings in the initial view.
8009 (final_start_function): Reintroduce SEEN-less interface.
8010 (final): Rename to...
8011 (final_1): ... this. Take SEEN parameter. Output final pending
8012 next view at the end.
8013 (final): Reintroduce seen-less interface.
8014 (final_scan_insn): Output pending next view before switching
8015 sections or ending a block. Mark the next view as needed when
8016 outputting variable locations. Notify debug backend of section
8017 changes, and of location view changes.
8018 (rest_of_handle_final): Adjust.
8019 * toplev.c (process_options): Autodetect value for debug variable
8020 location views option. Warn on incompat5 without -gdwarf-5.
8021 * doc/invoke.texi (gvariable-location-views): New.
8022 (gvariable-location-views=incompat5): New.
8023 (gno-variable-location-views): New.
8025 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8027 PR tree-optimization/84136
8028 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8029 that the result of find_edge is non-NULL.
8031 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8034 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8035 storing integer register in SImode. Fix cost of 256 and 512
8036 byte aligned SSE register store.
8038 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8040 * config/i386/i386.c (ix86_multiplication_cost): Fix
8041 multiplication cost for TARGET_AVX512DQ.
8043 2018-02-08 Marek Polacek <polacek@redhat.com>
8045 PR tree-optimization/84238
8046 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8049 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8051 PR tree-optimization/84265
8052 * tree-vect-stmts.c (vectorizable_store): Don't treat
8053 VMAT_CONTIGUOUS accesses as grouped.
8054 (vectorizable_load): Likewise.
8056 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8058 PR tree-optimization/81635
8059 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8060 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8061 (test_round_for_mask): New functions.
8062 (wide_int_cc_tests): Call test_round_for_mask.
8063 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8064 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8065 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8066 range returned by get_range_info.
8068 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8071 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8072 * symtab.c: Include builtins.h
8073 (symtab_node::output_to_lto_symbol_table_p): Move here
8074 from lto-streamer-out.c:output_symbol_p.
8075 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8076 (output_symbol_p): Move all logic to symtab.c
8077 (produce_symtab): Update.
8079 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8081 * config/s390/s390-opts.h (enum indirect_branch): Define.
8082 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8083 (s390_indirect_branch_via_thunk)
8084 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8085 (enum s390_indirect_branch_type): Define.
8086 * config/s390/s390.c (struct s390_frame_layout, struct
8087 machine_function): Remove.
8088 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8089 (indirect_branch_table_label_no, indirect_branch_table_name):
8091 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8092 (enum s390_indirect_branch_option): Define.
8093 (s390_return_addr_from_memory): New function.
8094 (s390_handle_string_attribute): New function.
8095 (s390_attribute_table): Add new attribute handler.
8096 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8097 (s390_indirect_branch_via_thunk): New function.
8098 (s390_indirect_branch_via_inline_thunk): New function.
8099 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8100 sibling call optimization for non z10 compiles.
8101 (s390_emit_call): Force indirect branch target to be a single
8102 register. Add r1 clobber for non-z10 compiles.
8103 (s390_emit_epilogue): Emit return jump via return_use expander.
8104 (s390_reorg): Handle JUMP_INSNs as execute targets.
8105 (s390_option_override_internal): Perform validity checks for the
8106 new command line options.
8107 (s390_indirect_branch_attrvalue): New function.
8108 (s390_indirect_branch_settings): New function.
8109 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8110 (s390_output_indirect_thunk_function): New function.
8111 (s390_code_end): Implement target hook.
8112 (s390_case_values_threshold): Implement target hook.
8113 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8115 * config/s390/s390.h (struct s390_frame_layout)
8116 (struct machine_function): Move here from s390.c.
8117 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8118 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8119 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8120 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8121 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8122 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8123 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8124 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8125 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8126 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8127 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8128 (mnemonic attribute): Add values which aren't recognized
8130 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8131 pattern for branch conversion. Fix mnemonic attribute.
8132 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8133 indirect branch via thunk if requested.
8134 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8135 ("*indirect_jump"): Disable for branch conversion using out of
8137 ("indirect_jump_via_thunk<mode>_z10")
8138 ("indirect_jump_via_thunk<mode>")
8139 ("indirect_jump_via_inlinethunk<mode>_z10")
8140 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8141 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8142 ("casesi_jump_via_inlinethunk<mode>_z10")
8143 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8144 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8145 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8146 ("*indirect2_jump"): Disable for branch conversion.
8147 ("casesi_jump"): Turn into expander and expand patterns for branch
8149 ("return_use"): New expander.
8150 ("*return"): Emit return via thunk and rename it to ...
8151 ("*return<mode>"): ... this one.
8152 * config/s390/s390.opt: Add new options and and enum for the
8155 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8157 * lra-constraints.c (match_reload): Unconditionally use
8158 gen_lowpart_SUBREG, rather than selecting between that
8159 and equivalent gen_rtx_SUBREG code.
8161 2018-02-08 Richard Biener <rguenther@suse.de>
8163 PR tree-optimization/84233
8164 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8165 changed flag instead of boguously re-using phi_inserted.
8167 2018-02-08 Martin Jambor <mjambor@suse.cz>
8169 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8170 static local variables.
8172 2018-02-08 Richard Biener <rguenther@suse.de>
8174 PR tree-optimization/84278
8175 * tree-vect-stmts.c (vectorizable_store): When looking for
8176 smaller vector types to perform grouped strided loads/stores
8177 make sure the mode is supported by the target.
8178 (vectorizable_load): Likewise.
8180 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8182 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8183 Increase LDP/STP opportunities by adding adjacent callee-saves.
8185 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8187 PR rtl-optimization/84068
8188 PR rtl-optimization/83459
8189 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8191 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8193 PR tree-optimization/84224
8194 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8195 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8198 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8201 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8202 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8203 count, remove one USE.
8205 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8207 * doc/install.texi (Configuration): Document the
8208 --with-long-double-format={ibm,ieee} PowerPC configuration
8212 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8213 Convert from define_expand to be define_insn_and_split. Rework
8214 float/double/_Float128 conversions to QI/HI/SImode to work with
8215 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8216 conversions to QI/HImode types did a store and then a load to
8217 truncate the value. For conversions to VSX registers, don't split
8218 the insn, instead emit the code directly. Use the code iterator
8219 any_fix to combine signed and unsigned conversions.
8220 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8221 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8222 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8223 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8224 (fix_<mode>di2_hw): Likewise.
8225 (fixuns_<mode>di2_hw): Likewise.
8226 (fix_<mode>si2_hw): Likewise.
8227 (fixuns_<mode>si2_hw): Likewise.
8228 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8229 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8230 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8231 fix<uns>_trunc<SFDF:mode>si2_p8.
8232 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8234 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8235 (fix<uns>_<mode>_mem): Likewise.
8236 (fctiw<u>z_<mode>_mem): Likewise.
8237 (fix<uns>_<mode>_mem): Likewise.
8238 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8239 the register allocator from doing a direct move to the GPRs to do
8240 a store, and instead use the ISA 3.0 store byte/half-word from
8241 vector register instruction. For IEEE 128-bit floating point,
8242 also optimize stores of 32-bit ints.
8243 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8245 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8247 * genextract.c (push_pathstr_operand): New function to support
8249 (walk_rtx): Call push_pathstr_operand.
8250 (print_path): Support [a-zA-Z].
8252 2018-02-07 Richard Biener <rguenther@suse.de>
8254 PR tree-optimization/84037
8255 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8256 (cse_and_gimplify_to_preheader): Declare.
8257 (vect_get_place_in_interleaving_chain): Likewise.
8258 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8260 (_loop_vec_info::~_loop_vec_info): Delete it.
8261 (cse_and_gimplify_to_preheader): New function.
8262 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8263 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8264 (vectorizable_load): Likewise. For grouped stores always base
8265 the IV on the first element.
8266 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8267 condition before gimplifying.
8269 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8271 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8272 *DIV_EXPR and *MOD_EXPR.
8274 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8277 * config/i386/i386.c (ix86_option_override_internal): Mask out
8278 the CF_SET bit when checking -fcf-protection.
8280 2018-02-07 Tom de Vries <tom@codesourcery.com>
8283 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8286 2018-02-07 Richard Biener <rguenther@suse.de>
8288 PR tree-optimization/84204
8289 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8292 PR tree-optimization/84205
8293 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8294 special-case isl_ast_op_zdiv_r.
8296 PR tree-optimization/84223
8297 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8298 Only add conditions from within the region.
8299 (gather_bbs::after_dom_children): Adjust.
8301 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8304 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8305 * config/avr/avr.md: Only post-reload split REG-REG moves if
8306 either register is GENERAL_REG_P.
8308 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8310 PR tree-optimization/84235
8311 * tree-ssa-scopedtables.c
8312 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8313 if the subtraction is performed in floating point type where NaNs are
8314 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8315 build 1. Formatting fix.
8317 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8320 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8321 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8322 and skip it regardless of bb boundaries. Use CALL_P macro,
8323 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8324 unnecessarily, formatting fix.
8326 2018-02-06 Michael Collison <michael.collison@arm.com>
8328 * config/arm/thumb2.md:
8329 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8330 (*thumb_mov_notscc): Ditto.
8332 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8335 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8336 unsigned_fix, not "s".
8338 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8340 * configure.ac (gcc_fn_eh_frame_ro): New function.
8341 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8342 correct .eh_frame permissions.
8343 * configure: Regenerate.
8345 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8347 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8350 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8352 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8353 Display warning message for -mno-speculate-indirect-jumps.
8355 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8357 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8359 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8361 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8363 PR tree-optimization/84225
8364 * tree-eh.c (find_trapping_overflow): Only call
8365 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8367 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8370 * config/i386/i386.c: Reimplement the check of possible options
8371 -mibt/-mshstk conbination. Change error messages.
8372 * doc/invoke.texi: Fix a typo: remove extra '='.
8374 2018-02-06 Marek Polacek <polacek@redhat.com>
8376 PR tree-optimization/84228
8377 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8379 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8382 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8383 emitted arch directives.
8384 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8385 __ARM_FEATURE_COPROC before changing architectures.
8387 2018-02-06 Richard Biener <rguenther@suse.de>
8389 * config/i386/i386.c (print_reg): Fix typo.
8390 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8392 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8394 * configure: Regenerate.
8396 2018-02-05 Martin Sebor <msebor@redhat.com>
8398 PR tree-optimization/83369
8399 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8402 2018-02-05 Martin Liska <mliska@suse.cz>
8404 * doc/invoke.texi: Cherry-pick upstream r323995.
8406 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8408 * ira.c (ira_init_register_move_cost): Adjust comment.
8410 2018-02-05 Martin Liska <mliska@suse.cz>
8412 PR gcov-profile/84137
8413 * doc/gcov.texi: Fix typo in documentation.
8415 2018-02-05 Martin Liska <mliska@suse.cz>
8417 PR gcov-profile/83879
8418 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8419 using dlopen functionality.
8421 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8423 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8424 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8425 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8426 _mm_maskz_range_round_ss): New intrinsics.
8427 (__builtin_ia32_rangesd128_round)
8428 (__builtin_ia32_rangess128_round): Remove.
8429 (__builtin_ia32_rangesd128_mask_round,
8430 __builtin_ia32_rangess128_mask_round): New builtins.
8431 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8432 __builtin_ia32_rangess128_round): Remove.
8433 (__builtin_ia32_rangesd128_mask_round,
8434 __builtin_ia32_rangess128_mask_round): New builtins.
8435 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8436 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8437 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8438 "<round_saeonly_constraint>")): Changed to ...
8439 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8440 "<round_saeonly_scalar_constraint>")): ... this.
8441 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8442 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8443 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8444 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8445 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8447 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8449 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8451 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8452 Remove all values except native, 8540 and 8548.
8454 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8456 * config/i386/i386.c (ix86_output_function_return): Pass
8457 INVALID_REGNUM, instead of -1, as invalid register number to
8458 indirect_thunk_name and output_indirect_thunk.
8460 2018-02-02 Julia Koval <julia.koval@intel.com>
8462 * config.gcc: Add -march=icelake.
8463 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8464 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8465 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8466 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8467 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8468 (processor_target_table): Add icelake.
8469 (ix86_option_override_internal): Handle new PTAs.
8470 (get_builtin_code_for_version): Handle icelake.
8471 (M_INTEL_COREI7_ICELAKE): New.
8472 (fold_builtin_cpu): Handle icelake.
8473 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8474 * doc/invoke.texi: Add -march=icelake.
8476 2018-02-02 Julia Koval <julia.koval@intel.com>
8478 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8479 to wide_int_bitmask.
8480 * wide-int-bitmask.h: New.
8482 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8485 * config/i386/i386.md: Replace Pmode with word_mode in
8486 builtin_setjmp_setup and builtin_longjmp to support x32.
8488 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8492 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8494 (rs6000_supported_cpu_names): New static variable.
8495 (linux_cpu_translation_table): Likewise.
8496 (elf_platform) <cpu>: Define new static variable and use it.
8497 Translate kernel AT_PLATFORM name to canonical name if needed.
8498 Error if platform name is unknown.
8500 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8503 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8505 2018-02-01 Jeff Law <law@redhat.com>
8508 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8509 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8510 the scratch if RELEASE_VIA_POP is false.
8511 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8512 If we have to save a temporary register, decrement SIZE appropriately.
8513 Pass new arguments to release_scratch_register_on_entry.
8514 (ix86_adjust_stack_and_probe): Likewise.
8515 (ix86_emit_probe_stack_range): Pass new arguments to
8516 release_scratch_register_on_entry.
8518 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8520 PR rtl-optimization/84157
8521 * combine.c (change_zero_ext): Use REG_P predicate in
8522 front of HARD_REGISTER_P predicate.
8524 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8526 * config/avr/avr.c (avr_option_override): Move disabling of
8527 -fdelete-null-pointer-checks to...
8528 * common/config/avr/avr-common.c (avr_option_optimization_table):
8531 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8533 PR tree-optimization/81635
8534 * tree-data-ref.c (split_constant_offset_1): For types that
8535 wrap on overflow, try to use range info to prove that wrapping
8538 2018-02-01 Renlin Li <renlin.li@arm.com>
8541 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8543 (aarch64_register_move_cost): Likewise.
8544 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8546 (REG_CLASS_NAMES): Likewise.
8547 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8548 TAILCALL_ADDR_REGS. Remove IP registers.
8549 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8551 2018-02-01 Richard Biener <rguenther@suse.de>
8553 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8554 for specifying RPO order and allow NULL for that.
8555 * domwalk.c (dom_walker::dom_walker): Likewise.
8556 (dom_walker::walk): Handle NULL RPO order.
8557 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8559 (rewrite_update_dom_walker): Likewise.
8560 (mark_def_dom_walker): Likewise.
8562 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8564 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8565 (aarch64_maybe_expand_sve_subreg_move): Declare.
8566 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8567 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8569 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8570 that are semantically a reverse operation.
8571 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8572 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8573 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8575 (aarch64_can_change_mode_class): For big-endian, forbid changes
8576 between two SVE modes if they have different element sizes.
8578 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8580 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8581 the TImode handling for big-endian targets.
8583 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8585 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8586 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8588 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8589 Remove BSWAP handing for big-endian targets and use the form of
8590 LD1RQ appropariate for the mode.
8592 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8594 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8595 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8598 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8601 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8602 check for operands that need to go through aarch64_sve_reload_be.
8604 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8606 PR tree-optimization/81661
8607 PR tree-optimization/84117
8608 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8609 * tree-eh.c: Include gimplify.h.
8610 (find_trapping_overflow, replace_trapping_overflow,
8611 rewrite_to_non_trapping_overflow): New functions.
8612 * tree-vect-loop.c: Include tree-eh.h.
8613 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8614 * tree-data-ref.c: Include tree-eh.h.
8615 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8617 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8619 PR rtl-optimization/84123
8620 * combine.c (change_zero_ext): Check if hard register satisfies
8621 can_change_dest_mode before calling gen_lowpart_SUBREG.
8623 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8626 * ira.c (ira_init_register_move_cost): Remove assert.
8628 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8630 PR rtl-optimization/84071
8631 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8632 * doc/tm.texi: Regenerate.
8634 2018-01-31 Richard Biener <rguenther@suse.de>
8636 PR tree-optimization/84132
8637 * tree-data-ref.c (analyze_miv_subscript): Properly
8638 check whether evolution_function_is_affine_multivariate_p
8639 before calling gcd_of_steps_may_divide_p.
8641 2018-01-31 Julia Koval <julia.koval@intel.com>
8644 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8645 * config/i386/i386.md (rdpid_rex64) New.
8646 (rdpid): Make 32bit only.
8648 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8651 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8652 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8654 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8657 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8659 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8661 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8663 PR rtl-optimization/84071
8664 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8665 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8667 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8669 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8670 (arc_attribute_table): Add 'aux' attribute.
8671 (arc_in_small_data_p): Consider aux like variables.
8672 (arc_is_aux_reg_p): New function.
8673 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8674 (arc_get_aux_arg): New function.
8675 (prepare_move_operands): Handle aux-register access.
8676 (arc_handle_aux_attribute): New function.
8677 * doc/extend.texi (ARC Variable attributes): Add subsection.
8679 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8681 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8682 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8683 (arc_attribute_table): Add 'uncached' attribute.
8684 (arc_print_operand): Print '.di' flag for uncached memory
8686 (arc_in_small_data_p): Do not consider for small data the uncached
8688 (arc_is_uncached_mem_p): New function.
8689 * config/arc/predicates.md (compact_store_memory_operand): Check
8690 for uncached memory accesses.
8691 (nonvol_nonimm_operand): Likewise.
8692 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8694 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8697 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8698 falign-loops=): Add Optimization flag.
8700 2018-01-30 Jeff Law <law@redhat.com>
8703 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8704 INT_REGISTERS_SAVED. Check it prior to calling
8705 get_scratch_register_on_entry.
8706 (ix86_adjust_stack_and_probe): Similarly.
8707 (ix86_emit_probe_stack_range): Similarly.
8708 (ix86_expand_prologue): Corresponding changes.
8710 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8713 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8714 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8716 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8719 * lra-constraints.c (curr_insn_transform): Process AND in the
8722 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8724 PR rtl-optimization/83986
8725 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8726 dependence against last_pending_memory_flush in addition to
8729 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8731 PR tree-optimization/81611
8732 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8735 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8738 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8741 2018-01-30 Richard Biener <rguenther@suse.de>
8742 Jakub Jelinek <jakub@redhat.com>
8744 PR tree-optimization/84111
8745 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8746 inner loops added during recursion, as they don't have up-to-date
8749 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8752 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8753 (can_inline_edge_by_limits_p): ... here.
8754 (can_early_inline_edge_p, check_callers,
8755 update_caller_keys, update_callee_keys, recursive_inlining,
8756 add_new_edges_to_heap, speculation_useful_p,
8757 inline_small_functions,
8758 inline_small_functions, flatten_function,
8759 inline_to_all_callers_1): Update.
8761 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8763 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8766 2018-01-30 Richard Biener <rguenther@suse.de>
8768 PR tree-optimization/83008
8769 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8770 invariant and constant vector uses in stmts when they need
8773 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8776 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8777 * configure: Regenerate.
8779 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8781 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8783 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8784 Use gen_rtx_REG rather than gen_lowpart.
8786 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8788 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8789 rather than 0 when creating partial subregs.
8791 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8793 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8796 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8799 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8800 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8801 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8802 flags. This restores the settings used before the 2017-07-24.
8803 Turning off pre increment/decrement/modify allows IVOPTS to
8804 optimize DF/SF loops where the index is an int.
8806 2018-01-29 Richard Biener <rguenther@suse.de>
8807 Kelvin Nilsen <kelvin@gcc.gnu.org>
8810 * tree-vect-stmts.c (vectorizable_call): Don't call
8811 targetm.vectorize_builtin_md_vectorized_function if callee is
8814 2018-01-22 Carl Love <cel@us.ibm.com>
8816 * doc/extend.tex: Fix typo in second arg in
8817 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8819 2018-01-29 Richard Biener <rguenther@suse.de>
8821 PR tree-optimization/84086
8822 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8823 (flush_ssaname_freelist): When SSA names were released reset
8824 the SCEV hash table.
8826 2018-01-29 Richard Biener <rguenther@suse.de>
8828 PR tree-optimization/84057
8829 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8830 removed paths when removing edges.
8832 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8834 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8835 -mfunction-return=@var{choice}.
8837 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8840 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8842 (layout::print_source_line): Likewise.
8843 (test_get_line_width_without_trailing_whitespace): Add test cases.
8845 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8848 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8851 2018-01-26 Jim Wilson <jimw@sifive.com>
8853 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8855 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8858 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8860 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8861 and CMP + SUB-immediate -> SUBS.
8863 2018-01-26 Martin Sebor <msebor@redhat.com>
8865 PR tree-optimization/83896
8866 * tree-ssa-strlen.c (get_string_len): Rename...
8867 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8868 Avoid assuming length is constant.
8869 (handle_char_store): Use HOST_WIDE_INT for string length.
8871 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8874 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8875 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8877 2018-01-26 Richard Biener <rguenther@suse.de>
8879 PR rtl-optimization/84003
8880 * dse.c (record_store): Only record redundant stores when
8881 the earlier store aliases at least all accesses the later one does.
8883 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8885 PR rtl-optimization/83985
8886 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8887 REG_CFA_RESTORE insns.
8888 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8889 REG_CFA_RESTORE insns here.
8892 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8893 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8895 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8897 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8898 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8899 (arc_init): Likewise.
8900 (arc_override_options): Likewise.
8901 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8903 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8905 * config/arc/arc.h (TARGET_DBNZ): Define.
8906 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8907 properly set the tune attribute.
8908 (dbnz): Use TARGET_DBNZ guard.
8909 * config/arc/arc.opt (mtune): Add core3 option.
8911 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8913 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8914 recognize new pic like addresses.
8915 (arc_delegitimize_address): Clean up.
8917 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8919 * config/arc/arc-arches.def: Option mrf16 valid for all
8921 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8922 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8923 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8924 * config/arc/arc-tables.opt: Regenerate.
8925 * config/arc/arc.c (arc_conditional_register_usage): Handle
8926 reduced register file case.
8927 (arc_file_start): Set must have build attributes.
8928 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8930 * config/arc/arc.opt (mrf16): Add new option.
8931 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8932 * config/arc/genmultilib.awk: Handle new mrf16 option.
8933 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8934 * config/arc/t-multilib: Regenerate.
8935 * doc/invoke.texi (ARC Options): Document mrf16 option.
8937 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8939 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8940 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8941 (arc_attribute_table): Add 'secure_call' attribute.
8942 (arc_print_operand): Print secure call operand.
8943 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8945 (arc_is_secure_call_p): New function. * config/arc/arc.md
8946 (call_i): Add support for sjli instruction.
8947 (call_value_i): Likewise.
8948 * config/arc/constraints.md (Csc): New constraint.
8950 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8951 John Eric Martin <John.Martin@emmicro-us.com>
8953 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8954 * config/arc/arc.c (_arc_jli_section): New struct.
8955 (arc_jli_section): New type.
8956 (rc_jli_sections): New static variable.
8957 (arc_handle_jli_attribute): New function.
8958 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8959 (arc_file_end): New function.
8960 (TARGET_ASM_FILE_END): Define.
8961 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8962 (arc_add_jli_section): New function.
8963 (jli_call_scan): Likewise.
8964 (arc_reorg): Call jli_call_scan.
8965 (arc_output_addsi): Remove 'S' from printing asm operand.
8966 (arc_is_jli_call_p): New function.
8967 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8969 (movhi_insn): Likewise.
8970 (movsi_insn): Likewise.
8971 (movsi_set_cc_insn): Likewise.
8972 (loadqi_update): Likewise.
8973 (load_zeroextendqisi_update): Likewise.
8974 (load_signextendqisi_update): Likewise.
8975 (loadhi_update): Likewise.
8976 (load_zeroextendhisi_update): Likewise.
8977 (load_signextendhisi_update): Likewise.
8978 (loadsi_update): Likewise.
8979 (loadsf_update): Likewise.
8980 (movsicc_insn): Likewise.
8981 (bset_insn): Likewise.
8982 (bxor_insn): Likewise.
8983 (bclr_insn): Likewise.
8984 (bmsk_insn): Likewise.
8985 (bicsi3_insn): Likewise.
8986 (cmpsi_cc_c_insn): Likewise.
8987 (movsi_ne): Likewise.
8988 (movsi_cond_exec): Likewise.
8989 (clrsbsi2): Likewise.
9000 (call_i): Remove 'S' asm letter, add jli instruction.
9001 (call_value_i): Likewise.
9002 * config/arc/arc.op (mjli-always): New option.
9003 * config/arc/constraints.md (Cji): New constraint.
9004 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9006 (subsf3_fpx): Likewise.
9007 (mulsf3_fpx): Likewise.
9008 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9010 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9012 * doc/invoke.texi (ARC): Document mjli-always option.
9014 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9016 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9017 avoid addition with 0 and use incw and decw where possible.
9019 2018-01-26 Richard Biener <rguenther@suse.de>
9021 PR tree-optimization/81082
9022 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9023 association if it requires casting to unsigned.
9024 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9025 from fold_plusminus_mult_expr to catch important cases late when
9026 range info is available.
9028 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9030 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9031 * configure.ac (hidden_linkonce): New test.
9032 * configure: Regenerate.
9033 * config.in: Regenerate.
9035 2018-01-26 Julia Koval <julia.koval@intel.com>
9037 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9038 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9039 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9040 _mm_mask_bitshuffle_epi64_mask): Fix type.
9041 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9042 USI_FTYPE_V4DI_V4DI_USI): Remove.
9043 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9044 __builtin_ia32_vpshufbitqmb256_mask,
9045 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9046 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9047 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9049 2018-01-26 Alan Modra <amodra@gmail.com>
9052 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9053 UNSPEC_VBPERMQ. Sort other unspecs.
9055 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9057 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9059 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9062 * predict.c (drop_profile): Do not push/pop cfun; update also
9064 (handle_missing_profiles): Fix logic looking for zero profiles.
9066 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9069 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9070 on functions with #pragma omp declare simd or functions with simd
9072 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9073 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9074 Remove trailing \n from warning_at calls.
9076 2018-01-25 Tom de Vries <tom@codesourcery.com>
9079 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9080 for neutered workers.
9082 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9085 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9086 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9088 2018-01-24 Jeff Law <law@redhat.com>
9091 * i386.c (get_probe_interval): Move to earlier point.
9092 (ix86_compute_frame_layout): If -fstack-clash-protection and
9093 the frame is larger than the probe interval, then use pushes
9094 to save registers rather than reg->mem moves.
9095 (ix86_expand_prologue): Remove conditional for int_registers_saved
9098 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9101 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9102 min/max for never referenced object.
9104 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9107 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9109 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9110 attributes from DECL_ATTRIBUTES (decl) without affecting
9111 DECL_ATTRIBUTES (current_function_decl).
9112 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9113 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9115 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9117 PR tree-optimization/83979
9118 * fold-const.c (fold_comparison): Use constant_boolean_node
9119 instead of boolean_{true,false}_node.
9121 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9123 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9126 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9128 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9129 Simplify the clause that sets the length attribute.
9130 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9131 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9132 clause that sets the length attribute.
9133 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9135 2018-01-24 Tom de Vries <tom@codesourcery.com>
9138 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9139 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9140 Add strict parameter.
9141 (prevent_branch_around_nothing): Insert dummy insn between branch to
9142 label and label with no ptx insn inbetween.
9143 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9145 2018-01-24 Tom de Vries <tom@codesourcery.com>
9148 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9149 for neutered threads in warp.
9150 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9152 2018-01-24 Richard Biener <rguenther@suse.de>
9154 PR tree-optimization/83176
9155 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9158 2018-01-24 Richard Biener <rguenther@suse.de>
9160 PR tree-optimization/82819
9161 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9162 code generating pluses that are no-ops in the target precision.
9164 2018-01-24 Richard Biener <rguenther@suse.de>
9167 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9169 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9171 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9172 to merge probabilities.
9173 * predict.c (probably_never_executed): Also mark as cold functions
9174 with global 0 profile and guessed local profile.
9175 * profile-count.c (profile_probability::combine_with_count): New
9177 * profile-count.h (profile_probability::operator*,
9178 profile_probability::operator*=, profile_probability::operator/,
9179 profile_probability::operator/=): Reduce precision to adjusted
9180 and set value to guessed on contradictory divisions.
9181 (profile_probability::combine_with_freq): Remove.
9182 (profile_probability::combine_wiht_count): Declare.
9183 (profile_count::force_nonzero):: Set to adjusted.
9184 (profile_count::probability_in):: Set quality to adjusted.
9185 * tree-ssa-tail-merge.c (replace_block_by): Use
9188 2018-01-23 Andrew Waterman <andrew@sifive.com>
9189 Jim Wilson <jimw@sifive.com>
9191 * config/riscv/riscv.c (riscv_stack_boundary): New.
9192 (riscv_option_override): Set riscv_stack_boundary. Handle
9193 riscv_preferred_stack_boundary_arg.
9194 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9195 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9196 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9197 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9198 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9199 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9201 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9204 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9205 of struct ix86_frame.
9206 (ix86_expand_epilogue): Likewise. Add a local variable for
9207 the reg_save_offset field in struct ix86_frame.
9209 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9211 PR tree-optimization/82604
9212 * tree-loop-distribution.c (enum partition_kind): New enum item
9213 PKIND_PARTIAL_MEMSET.
9214 (partition_builtin_p): Support above new enum item.
9215 (generate_code_for_partition): Ditto.
9216 (compute_access_range): Differentiate cases that equality can be
9217 proven at all loops, the innermost loops or no loops.
9218 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9219 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9220 (finalize_partitions, distribute_loop): Don't fuse partition of
9221 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9222 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9225 2018-01-23 Martin Liska <mliska@suse.cz>
9227 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9228 order to ignore the predictor.
9229 (PRED_POLYMORPHIC_CALL): Likewise.
9230 (PRED_RECURSIVE_CALL): Likewise.
9232 2018-01-23 Martin Liska <mliska@suse.cz>
9234 * tree-profile.c (tree_profiling): Print function header to
9235 aware reader which function we are working on.
9236 * value-prof.c (gimple_find_values_to_profile): Do not print
9237 not interesting value histograms.
9239 2018-01-23 Martin Liska <mliska@suse.cz>
9241 * profile-count.h (enum profile_quality): Add
9242 profile_uninitialized as the first value. Do not number values
9243 as they are zero based.
9244 (profile_count::verify): Update sanity check.
9245 (profile_probability::verify): Likewise.
9247 2018-01-23 Nathan Sidwell <nathan@acm.org>
9249 * doc/invoke.texi (ffor-scope): Deprecate.
9251 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9253 PR tree-optimization/83510
9254 * domwalk.c (set_all_edges_as_executable): New function.
9255 (dom_walker::dom_walker): Convert bool param
9256 "skip_unreachable_blocks" to enum reachability. Move setup of
9257 edge flags to set_all_edges_as_executable and only do it when
9258 reachability is REACHABLE_BLOCKS.
9259 * domwalk.h (enum dom_walker::reachability): New enum.
9260 (dom_walker::dom_walker): Convert bool param
9261 "skip_unreachable_blocks" to enum reachability.
9262 (set_all_edges_as_executable): New decl.
9263 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9264 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9266 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9267 but converting true to REACHABLE_BLOCKS.
9268 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9270 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9271 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9272 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9274 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9275 if check_all_array_refs will be called.
9277 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9279 * tree.c (selftest::test_location_wrappers): Add more test
9282 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9284 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9285 (selftest::test_bit_in_range): Likewise.
9287 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9290 * doc/sourcebuild.texi (vect_float): Say that the selector
9291 only describes the situation when -funsafe-math-optimizations is on.
9292 (vect_float_strict): Document.
9294 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9296 PR tree-optimization/83965
9297 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9298 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9299 instead of checking only for a reduction.
9300 (vect_recog_widen_sum_pattern): Likewise.
9302 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9304 * predict.c (probably_never_executed): Only use precise profile info.
9305 (compute_function_frequency): Skip after inlining hack since we now
9306 have quality checking.
9308 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9310 * profile-count.h (profile_probability::very_unlikely,
9311 profile_probability::unlikely, profile_probability::even): Set
9312 precision to guessed.
9314 2018-01-23 Richard Biener <rguenther@suse.de>
9316 PR tree-optimization/83963
9317 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9318 Properly terminate dominator walk when crossing the exit edge not
9319 when visiting its source block.
9321 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9324 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9325 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9327 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9329 PR tree-optimization/83957
9330 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9331 semicolon after for body surrounded by braces.
9333 PR tree-optimization/83081
9334 * profile-count.h (profile_probability::split): New method.
9335 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9336 Use profile_probability::split.
9337 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9338 when splitting a single conditional jump into 2.
9340 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9342 PR tree-optimization/69452
9343 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9346 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9348 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9349 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9350 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9352 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9354 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9355 * config/rl78/rl78.md (movdi): New define_expand.
9356 * config/rl78/rl78.c (rl78_split_movdi): New function.
9358 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9361 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9363 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9364 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9365 128-bit to produce an UNSPEC move to get the double word with the
9366 signbit and then a shift directly to do signbit.
9367 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9368 implementation with a new version that just does either a direct
9369 move or a regular move. Move memory interface to separate insns.
9370 Move insns so they are next to the expander.
9371 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9372 with signbit move. Split big and little endian case.
9373 (signbit<mode>2_dm_mem_le): Likewise.
9374 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9375 (signbit<mode>2_dm2): Likewise.
9377 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9379 * config/rl78/rl78.md (anddi3): New define_expand.
9381 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9383 * config/rl78/rl78.md (umindi3): New define_expand.
9385 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9387 * config/rl78/rl78.md (smindi3): New define_expand.
9389 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9391 * config/rl78/rl78.md (smaxdi3): New define_expand.
9393 2018-01-22 Carl Love <cel@us.ibm.com>
9395 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9396 LVX_V1TI): Add macro expansion.
9397 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9398 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9399 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9400 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9401 Change check to determine if the instruction is a byte reversing
9402 entry. Fix typo in comment.
9403 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9404 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9405 Add def_builtin calls for new builtins.
9406 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9407 Add define_insn expansion.
9409 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9411 * config/rl78/rl78.md (umaxdi3): New define_expand.
9413 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9415 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9416 for non-QImode registers.
9418 2018-01-22 Richard Biener <rguenther@suse.de>
9420 PR tree-optimization/83963
9421 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9422 including the loop exit block.
9423 (scop_detection::merge_sese): Likewise.
9424 (scop_detection::add_scop): Do it here instead.
9426 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9428 * doc/sourcebuild.texi (arm_softfloat): Document.
9430 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9433 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9434 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9435 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9437 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9438 David Edelsohn <dje.gcc@gmail.com>
9441 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9442 Change "crset eq" to "crset 2".
9443 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9444 (*call_indirect_aix<mode>_nospec): Likewise.
9445 (*call_value_indirect_aix<mode>_nospec): Likewise.
9446 (*call_indirect_elfv2<mode>_nospec): Likewise.
9447 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9448 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9449 change assembly output from . to $.
9450 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9451 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9452 (*tablejump<mode>_internal1_nospec): Likewise.
9454 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9457 * config/sh/sh_optimize_sett_clrt.cc:
9458 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9460 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9462 PR tree-optimization/83940
9463 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9464 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9465 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9466 use it to pass back the definition type.
9467 (vect_check_store_rhs): Likewise rhs_dt_out.
9468 (vect_build_gather_load_calls): Add a mask_dt argument and use
9469 it instead of a call to vect_is_simple_use.
9470 (vectorizable_store): Update calls to vect_check_load_store_mask
9471 and vect_check_store_rhs. Use the dt returned by the latter instead
9472 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9473 instead of calls to vect_is_simple_use. Pass the scalar rather
9474 than the vector operand to vect_is_simple_use when handling
9475 second and subsequent copies of an rhs value.
9476 (vectorizable_load): Update calls to vect_check_load_store_mask
9477 and vect_build_gather_load_calls. Use the cached mask_dt and
9478 gs_info.offset_dt instead of calls to vect_is_simple_use.
9480 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9483 * tree-emutls.c: Include gimplify.h.
9484 (lower_emutls_2): New function.
9485 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9486 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9487 it before further processing.
9490 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9491 UINTVAL (trueop1) instead of INTVAL (op1).
9493 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9497 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9498 INCOMING_FRAME_SP_OFFSET if not defined.
9499 (scan_trace): Add ENTRY argument. If true and
9500 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9501 emit a note to adjust the CFA offset.
9502 (create_cfi_notes): Adjust scan_trace callers.
9503 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9504 INCOMING_FRAME_SP_OFFSET in the CIE.
9505 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9506 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9508 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9509 * doc/tm.texi: Regenerated.
9511 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9513 PR rtl-optimization/83147
9514 * lra-constraints.c (remove_inheritance_pseudos): Use
9515 lra_substitute_pseudo_within_insn.
9517 2018-01-19 Tom de Vries <tom@codesourcery.com>
9518 Cesar Philippidis <cesar@codesourcery.com>
9521 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9523 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9526 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9527 spaces for function labels.
9529 2018-01-19 Martin Liska <mliska@suse.cz>
9531 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9532 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9533 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9534 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9535 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9536 (PRED_CONST_RETURN): Change from 69 to 65.
9537 (PRED_NULL_RETURN): Change from 91 to 71.
9538 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9539 (PRED_LOOP_GUARD): Change from 66 to 73.
9541 2018-01-19 Martin Liska <mliska@suse.cz>
9543 * predict.c (predict_insn_def): Add new assert.
9544 (struct branch_predictor): Change type to signed integer.
9545 (test_prediction_value_range): Amend test to cover
9547 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9548 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9549 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9550 (PRED_LOOP_IV_COMPARE): Likewise.
9551 * predict.h (PROB_UNINITIALIZED): Define new constant.
9553 2018-01-19 Martin Liska <mliska@suse.cz>
9555 * predict.c (dump_prediction): Add new format for
9556 analyze_brprob.py script which is enabled with -details
9558 * profile-count.h (precise_p): New function.
9560 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9562 PR tree-optimization/83922
9563 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9564 there are no statements that need masking.
9565 (vect_active_double_reduction_p): New function.
9566 (vect_analyze_loop_operations): Use it when handling phis that
9567 are not in the loop header.
9569 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9571 PR tree-optimization/83914
9572 * tree-vect-loop.c (vectorizable_induction): Don't convert
9573 init_expr or apply the peeling adjustment for inductions
9574 that are nested within the vectorized loop.
9576 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9578 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9581 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9585 * function.h (gimplify_parameters): Add gimple_seq * argument.
9586 * function.c: Include gimple.h and options.h.
9587 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9588 for the added local temporaries if needed.
9589 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9590 if there are any parameter cleanups, wrap whole body into a
9591 try/finally with the cleanups.
9593 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9596 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9597 Use GET_MODE_CLASS for scalar floating point.
9599 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9603 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9604 Fix call of call_cgraph_insertion_hooks.
9606 2018-01-18 Martin Sebor <msebor@redhat.com>
9608 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9610 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9613 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9616 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9619 * common.opt: (-ffile-prefix-map): New option.
9620 * opts.c (common_handle_option): Defer it.
9621 * opts-global.c (handle_common_deferred_options): Handle it.
9622 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9623 * file-prefix-map.h: New file.
9624 (remap_debug_filename, add_debug_prefix_map): ...here.
9625 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9626 * final.c (debug_prefix_map, add_debug_prefix_map
9627 remap_debug_filename): Move to...
9628 * file-prefix-map.c: New file.
9629 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9630 generalize, get rid of alloca(), use strrchr() instead of strchr().
9631 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9632 Implement in terms of add_prefix_map().
9633 (remap_macro_filename, remap_debug_filename): Implement in term of
9635 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9636 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9637 * dbxout.c: Include file-prefix-map.h.
9638 * varasm.c: Likewise.
9639 * vmsdbgout.c: Likewise.
9640 * xcoffout.c: Likewise.
9641 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9642 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9643 * doc/invoke.texi (-ffile-prefix-map): Document.
9644 (-fdebug-prefix-map): Update description.
9646 2018-01-18 Martin Liska <mliska@suse.cz>
9648 * config/i386/i386.c (indirect_thunk_name): Document that also
9650 (output_indirect_thunk): Document why both instructions
9651 (pause and lfence) are generated.
9653 2018-01-18 Richard Biener <rguenther@suse.de>
9655 PR tree-optimization/83887
9656 * graphite-scop-detection.c
9657 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9658 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9659 (scop_detection::merge_sese): Re-implement with a flood-fill
9660 algorithm that properly finds a SESE region if it exists.
9662 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9665 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9666 pointer_diff optimizations use view_convert instead of convert.
9668 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9670 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9671 Generate different code for -mno-speculate-indirect-jumps.
9672 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9673 (*call_indirect_aix<mode>): Disable for
9674 -mno-speculate-indirect-jumps.
9675 (*call_indirect_aix<mode>_nospec): New define_insn.
9676 (*call_value_indirect_aix<mode>): Disable for
9677 -mno-speculate-indirect-jumps.
9678 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9679 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9680 -mno-speculate-indirect-jumps.
9681 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9683 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9685 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9686 long double type, set the flags for noting the default long double
9687 type, even if we don't pass or return a long double type.
9689 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9692 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9695 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9697 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9698 support for merge[hl].
9699 (fold_mergehl_helper): New helper function.
9700 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9701 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9702 (altivec_vmrglw_direct): Add xxmrglw insn.
9704 2018-01-17 Andrew Waterman <andrew@sifive.com>
9706 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9707 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9709 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9712 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9713 call the lto_location_cache before reading the
9714 DECL_SOURCE_LOCATION of the types.
9716 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9717 Richard Sandiford <richard.sandiford@linaro.org>
9719 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9720 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9721 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9722 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9723 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9725 * config/aarch64/constraints.md (aarch64_movti_operand):
9727 * config/aarch64/predicates.md (Uti): Add new constraint.
9729 2018-01-17 Carl Love <cel@us.ibm.com>
9731 * config/rs6000/vsx.md (define_expand xl_len_r,
9732 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9733 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9735 (define_expand, define_insn): Move the shift left from the
9736 define_insn to the define_expand for lxvl and stxvl instructions.
9737 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9738 and XL_LEN_R definitions to PURE.
9740 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9742 * config/i386/i386.c (indirect_thunk_name): Declare regno
9743 as unsigned int. Compare regno with INVALID_REGNUM.
9744 (output_indirect_thunk): Ditto.
9745 (output_indirect_thunk_function): Ditto.
9746 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9747 in the call to output_indirect_thunk_function.
9749 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9752 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9753 rather than the size of inner_type to determine the stack slot size
9754 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9756 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9759 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9762 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9764 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9765 endian Linux systems to optionally enable multilibs for selecting
9766 the long double type if the user configured an explicit type.
9767 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9768 have no long double multilibs if not defined.
9769 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9770 warn if the user used -mabi={ieee,ibm}longdouble and we built
9771 multilibs for long double.
9772 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9773 appropriate multilib option.
9774 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9776 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9777 for building long double multilibs.
9778 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9780 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9782 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9785 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9787 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9790 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9793 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9796 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9798 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9799 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9801 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9803 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9804 different rtl trees depending on TARGET_64BIT.
9805 (rs6000_gen_lvx): Likewise.
9807 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9809 * config/visium/visium.md (nop): Tweak comment.
9810 (hazard_nop): Likewise.
9812 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9814 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9815 -mspeculate-indirect-jumps.
9816 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9817 for -mno-speculate-indirect-jumps.
9818 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9819 (*call_value_indirect_elfv2<mode>): Disable for
9820 -mno-speculate-indirect-jumps.
9821 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9822 (indirect_jump): Emit different RTL for
9823 -mno-speculate-indirect-jumps.
9824 (*indirect_jump<mode>): Disable for
9825 -mno-speculate-indirect-jumps.
9826 (*indirect_jump<mode>_nospec): New define_insn.
9827 (tablejump): Emit different RTL for
9828 -mno-speculate-indirect-jumps.
9829 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9830 (tablejumpsi_nospec): New define_expand.
9831 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9832 (tablejumpdi_nospec): New define_expand.
9833 (*tablejump<mode>_internal1): Disable for
9834 -mno-speculate-indirect-jumps.
9835 (*tablejump<mode>_internal1_nospec): New define_insn.
9836 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9839 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9841 * caller-save.c (insert_save): Drop unnecessary parameter. All
9844 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9845 Richard Biener <rguenth@suse.de>
9848 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9849 return early, inline manually is_gimple_sizepos. Make sure if we
9850 call gimplify_expr we don't end up with a gimple constant.
9851 * tree.c (variably_modified_type_p): Don't return true for
9852 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9853 * gimplify.h (is_gimple_sizepos): Remove.
9855 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9857 PR tree-optimization/83857
9858 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9859 vectorizable_live_operation for pure SLP statements.
9860 (vectorizable_live_operation): Handle PHIs.
9862 2018-01-16 Richard Biener <rguenther@suse.de>
9864 PR tree-optimization/83867
9865 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9866 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9868 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9871 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9872 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9873 If off is not INTEGER_CST, issue a may not be aligned warning
9874 rather than isn't aligned. Use isn%'t rather than isn't.
9875 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9877 <case MULT_EXPR>: Improve the case when bottom and one of the
9878 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9879 operand, in that case check if the other operand is multiple of
9880 bottom divided by the INTEGER_CST operand.
9882 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9885 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9886 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9887 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9888 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9889 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9890 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9891 (pa_function_arg_size): New function.
9893 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9895 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9896 in a separate statement.
9898 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9900 PR tree-optimization/83847
9901 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9902 group gathers and scatters.
9904 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9906 PR rtl-optimization/86620
9907 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9909 PR rtl-optimization/83213
9910 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9911 to last if both are JUMP_INSNs.
9913 PR tree-optimization/83843
9914 * gimple-ssa-store-merging.c
9915 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9916 store_immediate_info for bswap/nop orig_stores.
9918 2018-01-15 Andrew Waterman <andrew@sifive.com>
9920 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9922 <UDIV>: Increase cost if !TARGET_DIV.
9924 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9926 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9927 (define_attr "cr_logical_3op"): New.
9928 (cceq_ior_compare): Adjust.
9929 (cceq_ior_compare_complement): Adjust.
9930 (*cceq_rev_compare): Adjust.
9931 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9932 (is_cracked_insn): Adjust.
9933 (insn_must_be_first_in_group): Adjust.
9934 * config/rs6000/40x.md: Adjust.
9935 * config/rs6000/440.md: Adjust.
9936 * config/rs6000/476.md: Adjust.
9937 * config/rs6000/601.md: Adjust.
9938 * config/rs6000/603.md: Adjust.
9939 * config/rs6000/6xx.md: Adjust.
9940 * config/rs6000/7450.md: Adjust.
9941 * config/rs6000/7xx.md: Adjust.
9942 * config/rs6000/8540.md: Adjust.
9943 * config/rs6000/cell.md: Adjust.
9944 * config/rs6000/e300c2c3.md: Adjust.
9945 * config/rs6000/e500mc.md: Adjust.
9946 * config/rs6000/e500mc64.md: Adjust.
9947 * config/rs6000/e5500.md: Adjust.
9948 * config/rs6000/e6500.md: Adjust.
9949 * config/rs6000/mpc.md: Adjust.
9950 * config/rs6000/power4.md: Adjust.
9951 * config/rs6000/power5.md: Adjust.
9952 * config/rs6000/power6.md: Adjust.
9953 * config/rs6000/power7.md: Adjust.
9954 * config/rs6000/power8.md: Adjust.
9955 * config/rs6000/power9.md: Adjust.
9956 * config/rs6000/rs64.md: Adjust.
9957 * config/rs6000/titan.md: Adjust.
9959 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9961 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9962 ix86_indirect_branch_register logic.
9964 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9966 * config/i386/constraints.md (Bs): Update
9967 ix86_indirect_branch_register check. Don't check
9968 ix86_indirect_branch_register with GOT_memory_operand.
9970 * config/i386/predicates.md (GOT_memory_operand): Don't check
9971 ix86_indirect_branch_register here.
9972 (GOT32_symbol_operand): Likewise.
9974 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9976 * config/i386/predicates.md (constant_call_address_operand):
9977 Rewrite ix86_indirect_branch_register logic.
9978 (sibcall_insn_operand): Likewise.
9980 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9982 * config/i386/constraints.md (Bs): Replace
9983 ix86_indirect_branch_thunk_register with
9984 ix86_indirect_branch_register.
9986 * config/i386/i386.md (indirect_jump): Likewise.
9987 (tablejump): Likewise.
9988 (*sibcall_memory): Likewise.
9989 (*sibcall_value_memory): Likewise.
9990 Peepholes of indirect call and jump via memory: Likewise.
9991 * config/i386/i386.opt: Likewise.
9992 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9993 (GOT_memory_operand): Likewise.
9994 (call_insn_operand): Likewise.
9995 (sibcall_insn_operand): Likewise.
9996 (GOT32_symbol_operand): Likewise.
9998 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10000 PR middle-end/83837
10001 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10002 type rather than type addr's type points to.
10003 (expand_omp_atomic_mutex): Likewise.
10004 (expand_omp_atomic): Likewise.
10006 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10009 * config/i386/i386.c (output_indirect_thunk_function): Use
10010 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10011 for __x86_return_thunk.
10013 2018-01-15 Richard Biener <rguenther@suse.de>
10015 PR middle-end/83850
10016 * expmed.c (extract_bit_field_1): Fix typo.
10018 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10021 * config/arm/iterators.md (VF): New mode iterator.
10022 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10023 Remove integer-related logic from pattern.
10024 (neon_vabd<mode>_3): Likewise.
10026 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10028 PR middle-end/82694
10029 * common.opt (fstrict-overflow): No longer an alias.
10030 (fwrapv-pointer): New option.
10031 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10032 also for pointer types based on flag_wrapv_pointer.
10033 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10034 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10035 opts->x_flag_wrapv got set.
10036 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10037 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10038 POINTER_TYPE_OVERFLOW_UNDEFINED.
10039 * match.pd: Likewise in address comparison pattern.
10040 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10042 2018-01-15 Richard Biener <rguenther@suse.de>
10045 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10046 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10047 Reset type names to their identifier if their TYPE_DECL doesn't
10048 have linkage (and thus is used for ODR and devirt).
10049 (save_debug_info_for_decl): Remove.
10050 (save_debug_info_for_type): Likewise.
10051 (add_tree_to_fld_list): Adjust.
10052 * tree-pretty-print.c (dump_generic_node): Make dumping of
10053 type names more robust.
10055 2018-01-15 Richard Biener <rguenther@suse.de>
10057 * BASE-VER: Bump to 8.0.1.
10059 2018-01-14 Martin Sebor <msebor@redhat.com>
10062 * builtins.c (check_access): Avoid warning when the no-warning bit
10065 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10067 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10068 * ira-color (allocno_hard_regs_compare): Likewise.
10070 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10073 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10074 Use .pushsection/.popsection.
10076 2018-01-14 Martin Sebor <msebor@redhat.com>
10079 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10081 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10083 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10084 entry from extra_headers.
10085 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10086 extra_headers, make the list bitwise identical to the i?86-*-* one.
10088 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10090 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10091 -mcmodel=large with -mindirect-branch=thunk,
10092 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10093 -mfunction-return=thunk-extern.
10094 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10095 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10096 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10098 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10100 * config/i386/i386.c (print_reg): Print the name of the full
10101 integer register without '%'.
10102 (ix86_print_operand): Handle 'V'.
10103 * doc/extend.texi: Document 'V' modifier.
10105 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10107 * config/i386/constraints.md (Bs): Disallow memory operand for
10108 -mindirect-branch-register.
10110 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10111 (GOT_memory_operand): Likewise.
10112 (call_insn_operand): Likewise.
10113 (sibcall_insn_operand): Likewise.
10114 (GOT32_symbol_operand): Likewise.
10115 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10116 for -mindirect-branch-register.
10117 (tablejump): Likewise.
10118 (*sibcall_memory): Likewise.
10119 (*sibcall_value_memory): Likewise.
10120 Disallow peepholes of indirect call and jump via memory for
10121 -mindirect-branch-register.
10122 (*call_pop): Replace m with Bw.
10123 (*call_value_pop): Likewise.
10124 (*sibcall_pop_memory): Replace m with Bs.
10125 * config/i386/i386.opt (mindirect-branch-register): New option.
10126 * doc/invoke.texi: Document -mindirect-branch-register option.
10128 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10130 * config/i386/i386-protos.h (ix86_output_function_return): New.
10131 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10132 set function_return_type.
10133 (indirect_thunk_name): Add ret_p to indicate thunk for function
10135 (output_indirect_thunk_function): Pass false to
10136 indirect_thunk_name.
10137 (ix86_output_indirect_branch_via_reg): Likewise.
10138 (ix86_output_indirect_branch_via_push): Likewise.
10139 (output_indirect_thunk_function): Create alias for function
10140 return thunk if regno < 0.
10141 (ix86_output_function_return): New function.
10142 (ix86_handle_fndecl_attribute): Handle function_return.
10143 (ix86_attribute_table): Add function_return.
10144 * config/i386/i386.h (machine_function): Add
10145 function_return_type.
10146 * config/i386/i386.md (simple_return_internal): Use
10147 ix86_output_function_return.
10148 (simple_return_internal_long): Likewise.
10149 * config/i386/i386.opt (mfunction-return=): New option.
10150 (indirect_branch): Mention -mfunction-return=.
10151 * doc/extend.texi: Document function_return function attribute.
10152 * doc/invoke.texi: Document -mfunction-return= option.
10154 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10156 * config/i386/i386-opts.h (indirect_branch): New.
10157 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10158 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10159 with local indirect jump when converting indirect call and jump.
10160 (ix86_set_indirect_branch_type): New.
10161 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10162 (indirectlabelno): New.
10163 (indirect_thunk_needed): Likewise.
10164 (indirect_thunk_bnd_needed): Likewise.
10165 (indirect_thunks_used): Likewise.
10166 (indirect_thunks_bnd_used): Likewise.
10167 (INDIRECT_LABEL): Likewise.
10168 (indirect_thunk_name): Likewise.
10169 (output_indirect_thunk): Likewise.
10170 (output_indirect_thunk_function): Likewise.
10171 (ix86_output_indirect_branch_via_reg): Likewise.
10172 (ix86_output_indirect_branch_via_push): Likewise.
10173 (ix86_output_indirect_branch): Likewise.
10174 (ix86_output_indirect_jmp): Likewise.
10175 (ix86_code_end): Call output_indirect_thunk_function if needed.
10176 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10178 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10179 (ix86_attribute_table): Add indirect_branch.
10180 * config/i386/i386.h (machine_function): Add indirect_branch_type
10181 and has_local_indirect_jump.
10182 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10184 (tablejump): Likewise.
10185 (*indirect_jump): Use ix86_output_indirect_jmp.
10186 (*tablejump_1): Likewise.
10187 (simple_return_indirect_internal): Likewise.
10188 * config/i386/i386.opt (mindirect-branch=): New option.
10189 (indirect_branch): New.
10192 (thunk-inline): Likewise.
10193 (thunk-extern): Likewise.
10194 * doc/extend.texi: Document indirect_branch function attribute.
10195 * doc/invoke.texi: Document -mindirect-branch= option.
10197 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10200 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10202 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10204 * ipa-inline.c (want_inline_small_function_p): Return false if
10205 inlining has already failed with CIF_FINAL_ERROR.
10206 (update_caller_keys): Call want_inline_small_function_p before
10208 (update_callee_keys): Likewise.
10210 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10212 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10214 (rs6000_quadword_masked_address_p): Likewise.
10215 (quad_aligned_load_p): Likewise.
10216 (quad_aligned_store_p): Likewise.
10217 (const_load_sequence_p): Add comment to describe the outer-most loop.
10218 (mimic_memory_attributes_and_flags): New function.
10219 (rs6000_gen_stvx): Likewise.
10220 (replace_swapped_aligned_store): Likewise.
10221 (rs6000_gen_lvx): Likewise.
10222 (replace_swapped_aligned_load): Likewise.
10223 (replace_swapped_load_constant): Capitalize argument name in
10224 comment describing this function.
10225 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10226 and stores that access quad-word aligned addresses and replace
10227 with stvx or lvx instructions when appropriate.
10228 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10229 New function prototype.
10230 (rs6000_quadword_masked_address_p): Likewise.
10231 (rs6000_gen_lvx): Likewise.
10232 (rs6000_gen_stvx): Likewise.
10233 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10234 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10235 when memory address is aligned.
10236 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10237 this split to select lvx instruction when memory address is aligned.
10238 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10239 instruction when memory address is aligned.
10240 (*vsx_le_perm_load_v16qi): Likewise.
10241 (four unnamed splitters): Modify to select the stvx instruction
10242 when memory is aligned.
10244 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10246 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10247 which appears in the queue multiple times.
10249 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10250 Alan Hayward <alan.hayward@arm.com>
10251 David Sherwood <david.sherwood@arm.com>
10253 * tree-vectorizer.h (vec_lower_bound): New structure.
10254 (_loop_vec_info): Add check_nonzero and lower_bounds.
10255 (LOOP_VINFO_CHECK_NONZERO): New macro.
10256 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10257 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10258 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10259 fields. Make seg_len the distance travelled, not including the
10261 (dr_direction_indicator): Declare.
10262 (dr_zero_step_indicator): Likewise.
10263 (dr_known_forward_stride_p): Likewise.
10264 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10266 (runtime_alias_check_p): Allow runtime alias checks with
10268 (operator ==): Compare access_size and align.
10269 (prune_runtime_alias_test_list): Rework for new distinction between
10270 the access_size and seg_len.
10271 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10273 (get_segment_min_max): New function.
10274 (create_intersect_range_checks): Use it.
10275 (dr_step_indicator): New function.
10276 (dr_direction_indicator): Likewise.
10277 (dr_zero_step_indicator): Likewise.
10278 (dr_known_forward_stride_p): Likewise.
10279 * tree-loop-distribution.c (data_ref_segment_size): Return
10280 DR_STEP * (niters - 1).
10281 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10283 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10284 (vect_preserves_scalar_order_p): New function, split out from...
10285 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10286 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10287 (vect_vfa_access_size): New function.
10288 (vect_vfa_align): Likewise.
10289 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10290 (dump_lower_bound): New function.
10291 (vect_check_lower_bound): Likewise.
10292 (vect_small_gap_p): Likewise.
10293 (vectorizable_with_step_bound_p): Likewise.
10294 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10295 depencies if the vectorization factor is 1. Convert the checks
10296 for nonzero steps into checks on the bounds of DR_STEP. Try using
10297 a bunds check for variable steps if the minimum required step is
10298 relatively small. Update calls to the dr_with_seg_len
10299 constructor and to vect_compile_time_alias.
10300 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10302 (vect_loop_versioning): Call it.
10303 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10305 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10307 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10308 Alan Hayward <alan.hayward@arm.com>
10309 David Sherwood <david.sherwood@arm.com>
10311 * doc/sourcebuild.texi (vect_scatter_store): Document.
10312 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10314 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10316 * genopinit.c (main): Add supports_vec_scatter_store and
10317 supports_vec_scatter_store_cached to target_optabs.
10318 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10319 IFN_MASK_SCATTER_STORE.
10320 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10322 * internal-fn.h (internal_store_fn_p): Declare.
10323 (internal_fn_stored_value_index): Likewise.
10324 * internal-fn.c (scatter_store_direct): New macro.
10325 (expand_scatter_store_optab_fn): New function.
10326 (direct_scatter_store_optab_supported_p): New macro.
10327 (internal_store_fn_p): New function.
10328 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10329 IFN_MASK_SCATTER_STORE.
10330 (internal_fn_mask_index): Likewise.
10331 (internal_fn_stored_value_index): New function.
10332 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10333 for scatter stores.
10334 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10335 * optabs-query.c (supports_vec_scatter_store_p): New function.
10336 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10337 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10338 true for scatter stores.
10339 (vect_gather_scatter_fn_p): Handle scatter stores too.
10340 (vect_check_gather_scatter): Consider using scatter stores if
10341 supports_vec_scatter_store_p.
10342 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10343 scatter stores too.
10344 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10345 internal_fn_stored_value_index.
10346 (check_load_store_masking): Handle scatter stores too.
10347 (vect_get_store_rhs): Make public.
10348 (vectorizable_call): Use internal_store_fn_p.
10349 (vectorizable_store): Handle scatter store internal functions.
10350 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10351 when deciding whether the end of the group has been reached.
10352 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10353 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10354 (mask_scatter_store<mode>): New insns.
10356 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10357 Alan Hayward <alan.hayward@arm.com>
10358 David Sherwood <david.sherwood@arm.com>
10360 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10361 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10362 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10364 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10365 Use vect_truncate_gather_scatter_offset if we can't treat the
10366 operation as a normal gather load or scatter store.
10367 (get_group_load_store_type): Take the gather_scatter_info
10368 as argument. Try using a gather load or scatter store for
10369 single-element groups.
10370 (get_load_store_type): Update calls to get_group_load_store_type
10371 and vect_use_strided_gather_scatters_p.
10373 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10374 Alan Hayward <alan.hayward@arm.com>
10375 David Sherwood <david.sherwood@arm.com>
10377 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10378 optional tree argument.
10379 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10381 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10382 but continue to use the current value as a fallback.
10383 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10384 to compare the updates.
10385 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10386 (get_load_store_type): Use it when handling a strided access.
10387 (vect_get_strided_load_store_ops): New function.
10388 (vect_get_data_ptr_increment): Likewise.
10389 (vectorizable_load): Handle strided gather loads. Always pass
10390 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10392 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10393 Alan Hayward <alan.hayward@arm.com>
10394 David Sherwood <david.sherwood@arm.com>
10396 * doc/md.texi (gather_load@var{m}): Document.
10397 (mask_gather_load@var{m}): Likewise.
10398 * genopinit.c (main): Add supports_vec_gather_load and
10399 supports_vec_gather_load_cached to target_optabs.
10400 * optabs-tree.c (init_tree_optimization_optabs): Use
10401 ggc_cleared_alloc to allocate target_optabs.
10402 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10403 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10405 * internal-fn.h (internal_load_fn_p): Declare.
10406 (internal_gather_scatter_fn_p): Likewise.
10407 (internal_fn_mask_index): Likewise.
10408 (internal_gather_scatter_fn_supported_p): Likewise.
10409 * internal-fn.c (gather_load_direct): New macro.
10410 (expand_gather_load_optab_fn): New function.
10411 (direct_gather_load_optab_supported_p): New macro.
10412 (direct_internal_fn_optab): New function.
10413 (internal_load_fn_p): Likewise.
10414 (internal_gather_scatter_fn_p): Likewise.
10415 (internal_fn_mask_index): Likewise.
10416 (internal_gather_scatter_fn_supported_p): Likewise.
10417 * optabs-query.c (supports_at_least_one_mode_p): New function.
10418 (supports_vec_gather_load_p): Likewise.
10419 * optabs-query.h (supports_vec_gather_load_p): Declare.
10420 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10421 and memory_type field.
10422 (NUM_PATTERNS): Bump to 15.
10423 * tree-vect-data-refs.c: Include internal-fn.h.
10424 (vect_gather_scatter_fn_p): New function.
10425 (vect_describe_gather_scatter_call): Likewise.
10426 (vect_check_gather_scatter): Try using internal functions for
10427 gather loads. Recognize existing calls to a gather load function.
10428 (vect_analyze_data_refs): Consider using gather loads if
10429 supports_vec_gather_load_p.
10430 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10431 (vect_get_gather_scatter_offset_type): Likewise.
10432 (vect_convert_mask_for_vectype): Likewise.
10433 (vect_add_conversion_to_patterm): Likewise.
10434 (vect_try_gather_scatter_pattern): Likewise.
10435 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10436 (vect_vect_recog_func_ptrs): Add it.
10437 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10438 internal_fn_mask_index and internal_gather_scatter_fn_p.
10439 (check_load_store_masking): Take the gather_scatter_info as an
10440 argument and handle gather loads.
10441 (vect_get_gather_scatter_ops): New function.
10442 (vectorizable_call): Check internal_load_fn_p.
10443 (vectorizable_load): Likewise. Handle gather load internal
10445 (vectorizable_store): Update call to check_load_store_masking.
10446 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10447 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10448 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10449 (aarch64_gather_scale_operand_d): New predicates.
10450 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10451 (mask_gather_load<mode>): New insns.
10453 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10454 Alan Hayward <alan.hayward@arm.com>
10455 David Sherwood <david.sherwood@arm.com>
10457 * optabs.def (fold_left_plus_optab): New optab.
10458 * doc/md.texi (fold_left_plus_@var{m}): Document.
10459 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10460 * internal-fn.c (fold_left_direct): Define.
10461 (expand_fold_left_optab_fn): Likewise.
10462 (direct_fold_left_optab_supported_p): Likewise.
10463 * fold-const-call.c (fold_const_fold_left): New function.
10464 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10465 * tree-parloops.c (valid_reduction_p): New function.
10466 (gather_scalar_reductions): Use it.
10467 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10468 (vect_finish_replace_stmt): Declare.
10469 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10470 (needs_fold_left_reduction_p): New function, split out from...
10471 (vect_is_simple_reduction): ...here. Accept reductions that
10472 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10473 (vect_force_simple_reduction): Also store the reduction type in
10474 the assignment's STMT_VINFO_REDUC_TYPE.
10475 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10476 (merge_with_identity): New function.
10477 (vect_expand_fold_left): Likewise.
10478 (vectorize_fold_left_reduction): Likewise.
10479 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10480 scalar phi in place for it. Check for target support and reject
10481 cases that would reassociate the operation. Defer the transform
10482 phase to vectorize_fold_left_reduction.
10483 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10484 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10485 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10487 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10489 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10490 call to ifc_temp_var.
10492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10493 Alan Hayward <alan.hayward@arm.com>
10494 David Sherwood <david.sherwood@arm.com>
10496 * target.def (legitimize_address_displacement): Take the original
10497 offset as a poly_int.
10498 * targhooks.h (default_legitimize_address_displacement): Update
10500 * targhooks.c (default_legitimize_address_displacement): Likewise.
10501 * doc/tm.texi: Regenerate.
10502 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10503 as an argument, moving assert of ad->disp == ad->disp_term to...
10504 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10505 Try calling targetm.legitimize_address_displacement before expanding
10506 the address rather than afterwards, and adjust for the new interface.
10507 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10508 Match the new hook interface. Handle SVE addresses.
10509 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10510 new hook interface.
10512 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10514 * Makefile.in (OBJS): Add early-remat.o.
10515 * target.def (select_early_remat_modes): New hook.
10516 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10517 * doc/tm.texi: Regenerate.
10518 * targhooks.h (default_select_early_remat_modes): Declare.
10519 * targhooks.c (default_select_early_remat_modes): New function.
10520 * timevar.def (TV_EARLY_REMAT): New timevar.
10521 * passes.def (pass_early_remat): New pass.
10522 * tree-pass.h (make_pass_early_remat): Declare.
10523 * early-remat.c: New file.
10524 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10526 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10528 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10529 Alan Hayward <alan.hayward@arm.com>
10530 David Sherwood <david.sherwood@arm.com>
10532 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10533 vfm1 with a bound_epilog parameter.
10534 (vect_do_peeling): Update calls accordingly, and move the prologue
10535 call earlier in the function. Treat the base bound_epilog as 0 for
10536 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10537 this base when peeling for gaps.
10538 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10539 with fully-masked loops.
10540 (vect_estimate_min_profitable_iters): Handle the single peeled
10541 iteration in that case.
10543 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10544 Alan Hayward <alan.hayward@arm.com>
10545 David Sherwood <david.sherwood@arm.com>
10547 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10548 single-element interleaving even if the size is not a power of 2.
10549 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10550 accesses for single-element interleaving if the group size is
10553 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10554 Alan Hayward <alan.hayward@arm.com>
10555 David Sherwood <david.sherwood@arm.com>
10557 * doc/md.texi (fold_extract_last_@var{m}): Document.
10558 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10559 * optabs.def (fold_extract_last_optab): New optab.
10560 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10561 * internal-fn.c (fold_extract_direct): New macro.
10562 (expand_fold_extract_optab_fn): Likewise.
10563 (direct_fold_extract_optab_supported_p): Likewise.
10564 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10565 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10566 EXTRACT_LAST_REDUCTION.
10567 (get_initial_def_for_reduction): Do not create an initial vector
10568 for EXTRACT_LAST_REDUCTION reductions.
10569 (vectorizable_reduction): Leave the scalar phi in place for
10570 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10571 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10572 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10573 transform phase to vectorizable_condition.
10574 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10576 (vect_finish_stmt_generation): ...here.
10577 (vect_finish_replace_stmt): New function.
10578 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10579 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10581 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10583 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10584 Alan Hayward <alan.hayward@arm.com>
10585 David Sherwood <david.sherwood@arm.com>
10587 * doc/md.texi (extract_last_@var{m}): Document.
10588 * optabs.def (extract_last_optab): New optab.
10589 * internal-fn.def (EXTRACT_LAST): New internal function.
10590 * internal-fn.c (cond_unary_direct): New macro.
10591 (expand_cond_unary_optab_fn): Likewise.
10592 (direct_cond_unary_optab_supported_p): Likewise.
10593 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10594 loops using EXTRACT_LAST.
10595 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10596 (extract_last_<mode>): ...this optab.
10597 (vec_extract<mode><Vel>): Update accordingly.
10599 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10600 Alan Hayward <alan.hayward@arm.com>
10601 David Sherwood <david.sherwood@arm.com>
10603 * target.def (empty_mask_is_expensive): New hook.
10604 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10605 * doc/tm.texi: Regenerate.
10606 * targhooks.h (default_empty_mask_is_expensive): Declare.
10607 * targhooks.c (default_empty_mask_is_expensive): New function.
10608 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10609 if the target says that empty masks are expensive.
10610 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10612 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10615 Alan Hayward <alan.hayward@arm.com>
10616 David Sherwood <david.sherwood@arm.com>
10618 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10619 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10620 (vect_use_loop_mask_for_alignment_p): New function.
10621 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10622 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10623 niters_skip argument. Make sure that the first niters_skip elements
10624 of the first iteration are inactive.
10625 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10626 Update call to vect_set_loop_masks_directly.
10627 (get_misalign_in_elems): New function, split out from...
10628 (vect_gen_prolog_loop_niters): ...here.
10629 (vect_update_init_of_dr): Take a code argument that specifies whether
10630 the adjustment should be added or subtracted.
10631 (vect_update_init_of_drs): Likewise.
10632 (vect_prepare_for_masked_peels): New function.
10633 (vect_do_peeling): Skip prologue peeling if we're using a mask
10634 instead. Update call to vect_update_inits_of_drs.
10635 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10637 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10638 alignment. Do not include the number of peeled iterations in
10639 the minimum threshold in that case.
10640 (vectorizable_induction): Adjust the start value down by
10641 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10642 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10643 Take the number of skipped iterations into account when calculating
10645 * tree-vect-stmts.c (vect_gen_while_not): New function.
10647 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10648 Alan Hayward <alan.hayward@arm.com>
10649 David Sherwood <david.sherwood@arm.com>
10651 * doc/sourcebuild.texi (vect_fully_masked): Document.
10652 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10653 default value to 0.
10654 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10656 (vect_analyze_loop_2): ...here. Don't check the vectorization
10657 factor against the number of loop iterations if the loop is
10660 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10661 Alan Hayward <alan.hayward@arm.com>
10662 David Sherwood <david.sherwood@arm.com>
10664 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10665 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10666 (dump_groups): Update accordingly.
10667 (iv_use::mem_type): New member variable.
10668 (address_p): New function.
10669 (record_use): Add a mem_type argument and initialize the new
10671 (record_group_use): Add a mem_type argument. Use address_p.
10672 Remove obsolete null checks of base_object. Update call to record_use.
10673 (find_interesting_uses_op): Update call to record_group_use.
10674 (find_interesting_uses_cond): Likewise.
10675 (find_interesting_uses_address): Likewise.
10676 (get_mem_type_for_internal_fn): New function.
10677 (find_address_like_use): Likewise.
10678 (find_interesting_uses_stmt): Try find_address_like_use before
10679 calling find_interesting_uses_op.
10680 (addr_offset_valid_p): Use the iv mem_type field as the type
10681 of the addressed memory.
10682 (add_autoinc_candidates): Likewise.
10683 (get_address_cost): Likewise.
10684 (split_small_address_groups_p): Use address_p.
10685 (split_address_groups): Likewise.
10686 (add_iv_candidate_for_use): Likewise.
10687 (autoinc_possible_for_pair): Likewise.
10688 (rewrite_groups): Likewise.
10689 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10690 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10691 (get_alias_ptr_type_for_ptr_address): New function.
10692 (rewrite_use_address): Rewrite address uses in calls that were
10693 identified by find_address_like_use.
10695 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10696 Alan Hayward <alan.hayward@arm.com>
10697 David Sherwood <david.sherwood@arm.com>
10699 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10701 * gimple-expr.h (is_gimple_addressable: Likewise.
10702 * gimple-expr.c (is_gimple_address): Likewise.
10703 * internal-fn.c (expand_call_mem_ref): New function.
10704 (expand_mask_load_optab_fn): Use it.
10705 (expand_mask_store_optab_fn): Likewise.
10707 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10708 Alan Hayward <alan.hayward@arm.com>
10709 David Sherwood <david.sherwood@arm.com>
10711 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10712 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10713 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10714 (cond_umax@var{mode}): Document.
10715 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10716 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10717 (cond_umin_optab, cond_umax_optab): New optabs.
10718 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10719 (COND_IOR, COND_XOR): New internal functions.
10720 * internal-fn.h (get_conditional_internal_fn): Declare.
10721 * internal-fn.c (cond_binary_direct): New macro.
10722 (expand_cond_binary_optab_fn): Likewise.
10723 (direct_cond_binary_optab_supported_p): Likewise.
10724 (get_conditional_internal_fn): New function.
10725 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10726 Cope with reduction statements that are vectorized as calls rather
10728 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10729 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10730 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10731 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10732 (UNSPEC_COND_EOR): New unspecs.
10733 (optab): Add mappings for them.
10734 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10735 (sve_int_op, sve_fp_op): New int attributes.
10737 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10738 Alan Hayward <alan.hayward@arm.com>
10739 David Sherwood <david.sherwood@arm.com>
10741 * optabs.def (while_ult_optab): New optab.
10742 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10743 * internal-fn.def (WHILE_ULT): New internal function.
10744 * internal-fn.h (direct_internal_fn_supported_p): New override
10745 that takes two types as argument.
10746 * internal-fn.c (while_direct): New macro.
10747 (expand_while_optab_fn): New function.
10748 (convert_optab_supported_p): Likewise.
10749 (direct_while_optab_supported_p): New macro.
10750 * wide-int.h (wi::udiv_ceil): New function.
10751 * tree-vectorizer.h (rgroup_masks): New structure.
10752 (vec_loop_masks): New typedef.
10753 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10754 and fully_masked_p.
10755 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10756 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10757 (vect_max_vf): New function.
10758 (slpeel_make_loop_iterate_ntimes): Delete.
10759 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10760 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10761 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10762 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10763 internal-fn.h, stor-layout.h and optabs-query.h.
10764 (vect_set_loop_mask): New function.
10765 (add_preheader_seq): Likewise.
10766 (add_header_seq): Likewise.
10767 (interleave_supported_p): Likewise.
10768 (vect_maybe_permute_loop_masks): Likewise.
10769 (vect_set_loop_masks_directly): Likewise.
10770 (vect_set_loop_condition_masked): Likewise.
10771 (vect_set_loop_condition_unmasked): New function, split out from
10772 slpeel_make_loop_iterate_ntimes.
10773 (slpeel_make_loop_iterate_ntimes): Rename to..
10774 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10775 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10776 (vect_do_peeling): Update call accordingly.
10777 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10779 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10780 mask_compare_type, can_fully_mask_p and fully_masked_p.
10781 (release_vec_loop_masks): New function.
10782 (_loop_vec_info): Use it to free the loop masks.
10783 (can_produce_all_loop_masks_p): New function.
10784 (vect_get_max_nscalars_per_iter): Likewise.
10785 (vect_verify_full_masking): Likewise.
10786 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10787 retries, and free the mask rgroups before retrying. Check loop-wide
10788 reasons for disallowing fully-masked loops. Make the final decision
10789 about whether use a fully-masked loop or not.
10790 (vect_estimate_min_profitable_iters): Do not assume that peeling
10791 for the number of iterations will be needed for fully-masked loops.
10792 (vectorizable_reduction): Disable fully-masked loops.
10793 (vectorizable_live_operation): Likewise.
10794 (vect_halve_mask_nunits): New function.
10795 (vect_double_mask_nunits): Likewise.
10796 (vect_record_loop_mask): Likewise.
10797 (vect_get_loop_mask): Likewise.
10798 (vect_transform_loop): Handle the case in which the final loop
10799 iteration might handle a partial vector. Call vect_set_loop_condition
10800 instead of slpeel_make_loop_iterate_ntimes.
10801 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10802 (check_load_store_masking): New function.
10803 (prepare_load_store_mask): Likewise.
10804 (vectorizable_store): Handle fully-masked loops.
10805 (vectorizable_load): Likewise.
10806 (supportable_widening_operation): Use vect_halve_mask_nunits for
10808 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10809 (vect_gen_while): New function.
10810 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10811 (aarch64_uqdec<mode>): New insn.
10813 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10814 Alan Hayward <alan.hayward@arm.com>
10815 David Sherwood <david.sherwood@arm.com>
10817 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10818 (reduc_xor_scal_optab): New optabs.
10819 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10820 (reduc_xor_scal_@var{m}): Document.
10821 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10822 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10823 internal functions.
10824 * fold-const-call.c (fold_const_call): Handle them.
10825 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10826 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10827 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10828 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10829 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10830 (UNSPEC_XORV): New unspecs.
10831 (optab): Add entries for them.
10832 (BITWISEV): New int iterator.
10833 (bit_reduc_op): New int attributes.
10835 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10836 Alan Hayward <alan.hayward@arm.com>
10837 David Sherwood <david.sherwood@arm.com>
10839 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10840 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10841 * optabs.def (vec_shl_insert_optab): New optab.
10842 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10843 (duplicate_and_interleave): Likewise.
10844 * tree-vect-loop.c: Include internal-fn.h.
10845 (neutral_op_for_slp_reduction): New function, split out from
10846 get_initial_defs_for_reduction.
10847 (get_initial_def_for_reduction): Handle option 2 for variable-length
10848 vectors by loading the neutral value into a vector and then shifting
10849 the initial value into element 0.
10850 (get_initial_defs_for_reduction): Replace the code argument with
10851 the neutral value calculated by neutral_op_for_slp_reduction.
10852 Use gimple_build_vector for constant-length vectors.
10853 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10854 but the first group_size elements have a neutral value.
10855 Use duplicate_and_interleave otherwise.
10856 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10857 Update call to get_initial_defs_for_reduction. Handle SLP
10858 reductions for variable-length vectors by creating one vector
10859 result for each scalar result, with the elements associated
10860 with other scalar results stubbed out with the neutral value.
10861 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10862 Require IFN_VEC_SHL_INSERT for double reductions on
10863 variable-length vectors, or SLP reductions that have
10864 a neutral value. Require can_duplicate_and_interleave_p
10865 support for variable-length unchained SLP reductions if there
10866 is no neutral value, such as for MIN/MAX reductions. Also require
10867 the number of vector elements to be a multiple of the number of
10868 SLP statements when doing variable-length unchained SLP reductions.
10869 Update call to vect_create_epilog_for_reduction.
10870 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10871 and remove initial values.
10872 (duplicate_and_interleave): Make public.
10873 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10874 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10876 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10877 Alan Hayward <alan.hayward@arm.com>
10878 David Sherwood <david.sherwood@arm.com>
10880 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10881 (can_duplicate_and_interleave_p): New function.
10882 (vect_get_and_check_slp_defs): Take the vector of statements
10883 rather than just the current one. Remove excess parentheses.
10884 Restriction rejectinon of vect_constant_def and vect_external_def
10885 for variable-length vectors to boolean types, or types for which
10886 can_duplicate_and_interleave_p is false.
10887 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10888 (duplicate_and_interleave): New function.
10889 (vect_get_constant_vectors): Use gimple_build_vector for
10890 constant-length vectors and suitable variable-length constant
10891 vectors. Use duplicate_and_interleave for other variable-length
10892 vectors. Don't defer the update when inserting new statements.
10894 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10895 Alan Hayward <alan.hayward@arm.com>
10896 David Sherwood <david.sherwood@arm.com>
10898 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10899 min_profitable_iters doesn't go negative.
10901 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10902 Alan Hayward <alan.hayward@arm.com>
10903 David Sherwood <david.sherwood@arm.com>
10905 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10906 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10907 * optabs.def (vec_mask_load_lanes_optab): New optab.
10908 (vec_mask_store_lanes_optab): Likewise.
10909 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10910 (MASK_STORE_LANES): Likewise.
10911 * internal-fn.c (mask_load_lanes_direct): New macro.
10912 (mask_store_lanes_direct): Likewise.
10913 (expand_mask_load_optab_fn): Handle masked operations.
10914 (expand_mask_load_lanes_optab_fn): New macro.
10915 (expand_mask_store_optab_fn): Handle masked operations.
10916 (expand_mask_store_lanes_optab_fn): New macro.
10917 (direct_mask_load_lanes_optab_supported_p): Likewise.
10918 (direct_mask_store_lanes_optab_supported_p): Likewise.
10919 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10921 (vect_load_lanes_supported): Likewise.
10922 * tree-vect-data-refs.c (strip_conversion): New function.
10923 (can_group_stmts_p): Likewise.
10924 (vect_analyze_data_ref_accesses): Use it instead of checking
10925 for a pair of assignments.
10926 (vect_store_lanes_supported): Take a masked_p parameter.
10927 (vect_load_lanes_supported): Likewise.
10928 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10929 vect_store_lanes_supported and vect_load_lanes_supported.
10930 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10931 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10932 parameter. Don't allow gaps for masked accesses.
10933 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10934 and vect_load_lanes_supported.
10935 (get_load_store_type): Take a masked_p parameter and update
10936 call to get_group_load_store_type.
10937 (vectorizable_store): Update call to get_load_store_type.
10938 Handle IFN_MASK_STORE_LANES.
10939 (vectorizable_load): Update call to get_load_store_type.
10940 Handle IFN_MASK_LOAD_LANES.
10942 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10943 Alan Hayward <alan.hayward@arm.com>
10944 David Sherwood <david.sherwood@arm.com>
10946 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10948 * config/aarch64/aarch64-protos.h
10949 (aarch64_sve_struct_memory_operand_p): Declare.
10950 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10951 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10952 (VPRED, vpred): Handle SVE structure modes.
10953 * config/aarch64/constraints.md (Utx): New constraint.
10954 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10955 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10956 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10957 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10958 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10959 structure modes. Split into pieces after RA.
10960 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10961 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10963 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10964 SVE structure modes.
10965 (aarch64_classify_address): Likewise.
10966 (sizetochar): Move earlier in file.
10967 (aarch64_print_operand): Handle SVE register lists.
10968 (aarch64_array_mode): New function.
10969 (aarch64_sve_struct_memory_operand_p): Likewise.
10970 (TARGET_ARRAY_MODE): Redefine.
10972 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10973 Alan Hayward <alan.hayward@arm.com>
10974 David Sherwood <david.sherwood@arm.com>
10976 * target.def (array_mode): New target hook.
10977 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10978 * doc/tm.texi: Regenerate.
10979 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10980 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10981 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10982 targetm.array_mode.
10983 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10987 Alan Hayward <alan.hayward@arm.com>
10988 David Sherwood <david.sherwood@arm.com>
10990 * fold-const.c (fold_binary_loc): Check the argument types
10991 rather than the result type when testing for a vector operation.
10993 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10995 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10996 * doc/tm.texi: Regenerate.
10998 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10999 Alan Hayward <alan.hayward@arm.com>
11000 David Sherwood <david.sherwood@arm.com>
11002 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11003 (sve): Document new AArch64 extension.
11004 * doc/md.texi (w): Extend the description of the AArch64
11005 constraint to include SVE vectors.
11006 (Upl, Upa): Document new AArch64 predicate constraints.
11007 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11009 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11010 (msve-vector-bits=): New option.
11011 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11012 SVE when these are disabled.
11013 (sve): New extension.
11014 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11015 modes. Adjust their number of units based on aarch64_sve_vg.
11016 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11017 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11018 aarch64_addr_query_type.
11019 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11020 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11021 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11022 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11023 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11024 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11025 (aarch64_simd_imm_zero_p): Delete.
11026 (aarch64_check_zero_based_sve_index_immediate): Declare.
11027 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11028 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11029 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11030 (aarch64_sve_float_mul_immediate_p): Likewise.
11031 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11032 rather than an rtx.
11033 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11034 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11035 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11036 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11037 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11038 (aarch64_regmode_natural_size): Likewise.
11039 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11040 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11042 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11043 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11044 for VG and the SVE predicate registers.
11045 (V_ALIASES): Add a "z"-prefixed alias.
11046 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11047 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11048 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11049 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11050 (REG_CLASS_NAMES): Add entries for them.
11051 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11052 and the predicate registers.
11053 (aarch64_sve_vg): Declare.
11054 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11055 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11056 (REGMODE_NATURAL_SIZE): Define.
11057 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11059 * config/aarch64/aarch64.c: Include cfgrtl.h.
11060 (simd_immediate_info): Add a constructor for series vectors,
11061 and an associated step field.
11062 (aarch64_sve_vg): New variable.
11063 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11064 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11065 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11066 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11067 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11068 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11069 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11070 (aarch64_get_mask_mode): New functions.
11071 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11072 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11073 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11074 predicate modes and predicate registers. Explicitly restrict
11075 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11076 to store a vector mode if it is recognized by
11077 aarch64_classify_vector_mode.
11078 (aarch64_regmode_natural_size): New function.
11079 (aarch64_hard_regno_caller_save_mode): Return the original mode
11081 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11082 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11083 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11084 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11086 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11087 does not overlap dest if the function is frame-related. Handle
11089 (aarch64_split_add_offset): New function.
11090 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11091 them aarch64_add_offset.
11092 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11093 and update call to aarch64_sub_sp.
11094 (aarch64_add_cfa_expression): New function.
11095 (aarch64_expand_prologue): Pass extra temporary registers to the
11096 functions above. Handle the case in which we need to emit new
11097 DW_CFA_expressions for registers that were originally saved
11098 relative to the stack pointer, but now have to be expressed
11099 relative to the frame pointer.
11100 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11102 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11103 IP0 and IP1 values for SVE frames.
11104 (aarch64_expand_vec_series): New function.
11105 (aarch64_expand_sve_widened_duplicate): Likewise.
11106 (aarch64_expand_sve_const_vector): Likewise.
11107 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11108 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11109 into the register, rather than emitting a SET directly.
11110 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11111 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11112 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11113 (offset_9bit_signed_scaled_p): New functions.
11114 (aarch64_replicate_bitmask_imm): New function.
11115 (aarch64_bitmask_imm): Use it.
11116 (aarch64_cannot_force_const_mem): Reject expressions involving
11117 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11118 (aarch64_classify_index): Handle SVE indices, by requiring
11119 a plain register index with a scale that matches the element size.
11120 (aarch64_classify_address): Handle SVE addresses. Assert that
11121 the mode of the address is VOIDmode or an integer mode.
11122 Update call to aarch64_classify_symbol.
11123 (aarch64_classify_symbolic_expression): Update call to
11124 aarch64_classify_symbol.
11125 (aarch64_const_vec_all_in_range_p): New function.
11126 (aarch64_print_vector_float_operand): Likewise.
11127 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11128 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11129 and the FP immediates 1.0 and 0.5.
11130 (aarch64_print_address_internal): Handle SVE addresses.
11131 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11132 (aarch64_regno_regclass): Handle predicate registers.
11133 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11135 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11136 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11137 (aarch64_convert_sve_vector_bits): New function.
11138 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11139 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11140 rather than an rtx.
11141 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11142 Handle SVE vector and predicate modes. Accept VL-based constants
11143 that need only one temporary register, and VL offsets that require
11144 no temporary registers.
11145 (aarch64_conditional_register_usage): Mark the predicate registers
11146 as fixed if SVE isn't available.
11147 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11148 Return true for SVE vector and predicate modes.
11149 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11150 rather than an unsigned int. Handle SVE modes.
11151 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11153 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11155 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11156 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11157 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11158 (aarch64_sve_float_mul_immediate_p): New functions.
11159 (aarch64_sve_valid_immediate): New function.
11160 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11161 Explicitly reject structure modes. Check for INDEX constants.
11162 Handle PTRUE and PFALSE constants.
11163 (aarch64_check_zero_based_sve_index_immediate): New function.
11164 (aarch64_simd_imm_zero_p): Delete.
11165 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11166 vector modes. Accept constants in the range of CNT[BHWD].
11167 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11168 ask for an Advanced SIMD mode.
11169 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11170 (aarch64_simd_vector_alignment): Handle SVE predicates.
11171 (aarch64_vectorize_preferred_vector_alignment): New function.
11172 (aarch64_simd_vector_alignment_reachable): Use it instead of
11174 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11175 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11177 (MAX_VECT_LEN): Delete.
11178 (expand_vec_perm_d): Add a vec_flags field.
11179 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11180 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11181 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11183 (aarch64_evpc_rev): Rename to...
11184 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11185 (aarch64_evpc_rev_global): New function.
11186 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11187 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11189 (aarch64_evpc_sve_tbl): New function.
11190 (aarch64_expand_vec_perm_const_1): Update after rename of
11191 aarch64_evpc_rev. Handle SVE permutes too, trying
11192 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11193 than aarch64_evpc_tbl.
11194 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11195 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11196 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11197 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11198 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11199 (aarch64_expand_sve_vcond): New functions.
11200 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11201 of aarch64_vector_mode_p.
11202 (aarch64_dwarf_poly_indeterminate_value): New function.
11203 (aarch64_compute_pressure_classes): Likewise.
11204 (aarch64_can_change_mode_class): Likewise.
11205 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11206 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11207 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11208 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11209 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11210 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11211 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11212 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11214 (Dn, Dl, Dr): Accept const as well as const_vector.
11215 (Dz): Likewise. Compare against CONST0_RTX.
11216 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11217 of "vector" where appropriate.
11218 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11219 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11220 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11221 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11222 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11223 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11224 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11225 (v_int_equiv): Extend to SVE modes.
11226 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11228 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11229 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11230 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11231 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11232 (SVE_COND_FP_CMP): New int iterators.
11233 (perm_hilo): Handle the new unpack unspecs.
11234 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11236 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11237 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11238 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11239 (aarch64_equality_operator, aarch64_constant_vector_operand)
11240 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11241 (aarch64_sve_nonimmediate_operand): Likewise.
11242 (aarch64_sve_general_operand): Likewise.
11243 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11244 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11245 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11246 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11247 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11248 (aarch64_sve_float_arith_immediate): Likewise.
11249 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11250 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11251 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11252 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11253 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11254 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11255 (aarch64_sve_float_arith_operand): Likewise.
11256 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11257 (aarch64_sve_float_mul_operand): Likewise.
11258 (aarch64_sve_vec_perm_operand): Likewise.
11259 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11260 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11261 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11262 as well as const_vector.
11263 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11264 in file. Use CONST0_RTX and CONSTM1_RTX.
11265 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11266 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11267 Use aarch64_simd_imm_zero.
11268 * config/aarch64/aarch64-sve.md: New file.
11269 * config/aarch64/aarch64.md: Include it.
11270 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11271 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11272 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11273 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11274 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11275 (sve): New attribute.
11276 (enabled): Disable instructions with the sve attribute unless
11278 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11279 aarch64_expand_mov_immediate.
11280 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11281 CNT[BHSD] immediates.
11282 (movti): Split CONST_POLY_INT moves into two halves.
11283 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11284 Split additions that need a temporary here if the destination
11285 is the stack pointer.
11286 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11287 (*add<mode>3_poly_1): New instruction.
11288 (set_clobber_cc): New expander.
11290 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11292 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11293 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11294 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11295 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11296 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11297 Change innermode from fixed_mode_size to machine_mode.
11298 (simplify_subreg): Update call accordingly. Handle a constant-sized
11299 subreg of a variable-length CONST_VECTOR.
11301 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11302 Alan Hayward <alan.hayward@arm.com>
11303 David Sherwood <david.sherwood@arm.com>
11305 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11306 (add_offset_to_base): New function, split out from...
11307 (create_mem_ref): ...here. When handling a scale other than 1,
11308 check first whether the address is valid without the offset.
11309 Add it into the base if so, leaving the index and scale as-is.
11311 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11314 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11315 fold_for_warn before checking if arg2 is INTEGER_CST.
11317 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11319 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11320 (store_multiple_operation): Delete.
11321 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11322 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11323 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11324 guarded by TARGET_STRING.
11325 (rs6000_output_load_multiple): Delete.
11326 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11327 OPTION_MASK_STRING / TARGET_STRING handling.
11328 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11329 (const rs6000_opt_masks) <"string">: Change mask to 0.
11330 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11331 (MASK_STRING): Delete.
11332 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11334 (load_multiple): Delete.
11341 (store_multiple): Delete.
11348 (movmemsi_8reg): Delete.
11349 (corresponding unnamed define_insn): Delete.
11350 (movmemsi_6reg): Delete.
11351 (corresponding unnamed define_insn): Delete.
11352 (movmemsi_4reg): Delete.
11353 (corresponding unnamed define_insn): Delete.
11354 (movmemsi_2reg): Delete.
11355 (corresponding unnamed define_insn): Delete.
11356 (movmemsi_1reg): Delete.
11357 (corresponding unnamed define_insn): Delete.
11358 * config/rs6000/rs6000.opt (mno-string): New.
11359 (mstring): Replace by deprecation warning stub.
11360 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11362 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11364 * regrename.c (regrename_do_replace): If replacing the same
11365 reg multiple times, try to reuse last created gen_raw_REG.
11368 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11369 main to workaround a bug in GDB.
11371 2018-01-12 Tom de Vries <tom@codesourcery.com>
11374 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11376 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11378 PR rtl-optimization/80481
11379 * ira-color.c (get_cap_member): New function.
11380 (allocnos_conflict_by_live_ranges_p): Use it.
11381 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11382 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11384 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11387 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11388 (*saddl_se_1): Ditto.
11389 (*ssubsi_1): Ditto.
11390 (*ssubl_se_1): Ditto.
11392 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11394 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11395 rather than wi::to_widest for DR_INITs.
11396 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11397 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11398 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11400 (vect_analyze_group_access_1): Note that here.
11402 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11404 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11405 polynomial type sizes.
11407 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11409 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11410 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11411 (gimple_add_tmp_var): Likewise.
11413 2018-01-12 Martin Liska <mliska@suse.cz>
11415 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11416 (gimple_alloc_sizes): Likewise.
11417 (dump_gimple_statistics): Use PRIu64 in printf format.
11418 * gimple.h: Change uint64_t to int.
11420 2018-01-12 Martin Liska <mliska@suse.cz>
11422 * tree-core.h: Use uint64_t instead of int.
11423 * tree.c (tree_node_counts): Likewise.
11424 (tree_node_sizes): Likewise.
11425 (dump_tree_statistics): Use PRIu64 in printf format.
11427 2018-01-12 Martin Liska <mliska@suse.cz>
11429 * Makefile.in: As qsort_chk is implemented in vec.c, add
11430 vec.o to linkage of gencfn-macros.
11431 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11432 passing the info to record_node_allocation_statistics.
11433 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11435 * ggc-common.c (struct ggc_usage): Add operator== and use
11436 it in operator< and compare function.
11437 * mem-stats.h (struct mem_usage): Likewise.
11438 * vec.c (struct vec_usage): Remove operator< and compare
11439 function. Can be simply inherited.
11441 2018-01-12 Martin Jambor <mjambor@suse.cz>
11444 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11445 * tree-ssa-math-opts.c: Include domwalk.h.
11446 (convert_mult_to_fma_1): New function.
11447 (fma_transformation_info): New type.
11448 (fma_deferring_state): Likewise.
11449 (cancel_fma_deferring): New function.
11450 (result_of_phi): Likewise.
11451 (last_fma_candidate_feeds_initial_phi): Likewise.
11452 (convert_mult_to_fma): Added deferring logic, split actual
11453 transformation to convert_mult_to_fma_1.
11454 (math_opts_dom_walker): New type.
11455 (math_opts_dom_walker::after_dom_children): New method, body moved
11456 here from pass_optimize_widening_mul::execute, added deferring logic
11458 (pass_optimize_widening_mul::execute): Moved most of code to
11459 math_opts_dom_walker::after_dom_children.
11460 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11461 * config/i386/i386.c (ix86_option_override_internal): Added
11462 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11464 2018-01-12 Richard Biener <rguenther@suse.de>
11467 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11468 inline instance vars.
11470 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11473 * config/rx/rx.c (rx_is_restricted_memory_address):
11474 Handle SUBREG case.
11476 2018-01-12 Richard Biener <rguenther@suse.de>
11478 PR tree-optimization/80846
11479 * target.def (split_reduction): New target hook.
11480 * targhooks.c (default_split_reduction): New function.
11481 * targhooks.h (default_split_reduction): Declare.
11482 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11483 target requests first reduce vectors by combining low and high
11485 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11486 (get_vectype_for_scalar_type_and_size): Export.
11487 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11488 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11489 * doc/tm.texi: Regenerate.
11490 * config/i386/i386.c (ix86_split_reduction): Implement
11491 TARGET_VECTORIZE_SPLIT_REDUCTION.
11493 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11496 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11497 in PIC mode except for TARGET_VXWORKS_RTP.
11498 * config/sparc/sparc.c: Include cfgrtl.h.
11499 (TARGET_INIT_PIC_REG): Define.
11500 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11501 (sparc_pic_register_p): New predicate.
11502 (sparc_legitimate_address_p): Use it.
11503 (sparc_legitimize_pic_address): Likewise.
11504 (sparc_delegitimize_address): Likewise.
11505 (sparc_mode_dependent_address_p): Likewise.
11506 (gen_load_pcrel_sym): Remove 4th parameter.
11507 (load_got_register): Adjust call to above. Remove obsolete stuff.
11508 (sparc_expand_prologue): Do not call load_got_register here.
11509 (sparc_flat_expand_prologue): Likewise.
11510 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11511 (sparc_use_pseudo_pic_reg): New function.
11512 (sparc_init_pic_reg): Likewise.
11513 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11514 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11516 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11518 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11519 Add item for branch_cost.
11521 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11523 PR rtl-optimization/83565
11524 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11525 not extend the result to a larger mode for rotate operations.
11526 (num_sign_bit_copies1): Likewise.
11528 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11531 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11533 Use values-Xc.o for -pedantic.
11534 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11536 2018-01-12 Martin Liska <mliska@suse.cz>
11539 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11541 (possible_polymorphic_call_targets): Use it.
11542 (ipa_devirt): Likewise.
11544 2018-01-12 Martin Liska <mliska@suse.cz>
11546 * profile-count.h (enum profile_quality): Use 0 as invalid
11547 enum value of profile_quality.
11549 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11551 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11552 -mext-string options.
11554 2018-01-12 Richard Biener <rguenther@suse.de>
11556 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11557 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11558 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11560 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11562 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11564 * configure.ac (--with-long-double-format): Add support for the
11565 configuration option to change the default long double format on
11567 * config.gcc (powerpc*-linux*-*): Likewise.
11568 * configure: Regenerate.
11569 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11570 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11571 used without modification.
11573 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11575 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11576 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11577 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11578 MISC_BUILTIN_SPEC_BARRIER.
11579 (rs6000_init_builtins): Likewise.
11580 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11582 (speculation_barrier): New define_insn.
11583 * doc/extend.texi: Document __builtin_speculation_barrier.
11585 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11588 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11589 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11590 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11592 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11593 integral modes instead of "ss" and "sd".
11594 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11595 vectors with 32-bit and 64-bit elements.
11596 (vecdupssescalarmodesuffix): New mode attribute.
11597 (vec_dup<mode>): Use it.
11599 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11602 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11603 frame if argument is passed on stack.
11605 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11608 * ree.c (combine_reaching_defs): Optimize also
11609 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11610 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11612 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11614 PR middle-end/83189
11615 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11617 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11619 PR middle-end/83718
11620 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11621 after they are computed.
11623 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11625 PR tree-optimization/83695
11626 * gimple-loop-linterchange.cc
11627 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11628 reset cached scev information after interchange.
11629 (pass_linterchange::execute): Remove call to scev_reset_htab.
11631 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11633 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11634 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11635 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11636 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11637 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11638 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11639 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11640 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11641 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11642 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11643 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11644 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11645 (V_lane_reg): Likewise.
11646 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11648 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11649 (vfmal_lane_low<mode>_intrinsic,
11650 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11651 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11652 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11653 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11654 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11655 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11657 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11659 * config/arm/arm-cpus.in (fp16fml): New feature.
11660 (ALL_SIMD): Add fp16fml.
11661 (armv8.2-a): Add fp16fml as an option.
11662 (armv8.3-a): Likewise.
11663 (armv8.4-a): Add fp16fml as part of fp16.
11664 * config/arm/arm.h (TARGET_FP16FML): Define.
11665 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11667 * config/arm/arm-modes.def (V2HF): Define.
11668 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11669 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11670 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11671 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11672 vfmsl_low, vfmsl_high): New set of builtins.
11673 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11674 (vfml_op): New code attribute.
11675 (VFMLHALVES): New int iterator.
11676 (VFML, VFMLSEL): New mode attributes.
11677 (V_reg): Define mapping for V2HF.
11678 (V_hi, V_lo): New mode attributes.
11679 (VF_constraint): Likewise.
11680 (vfml_half, vfml_half_selector): New int attributes.
11681 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11683 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11684 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11686 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11687 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11688 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11689 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11691 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11692 Document new effective target and option set.
11694 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11696 * config/arm/arm-cpus.in (armv8_4): New feature.
11697 (ARMv8_4a): New fgroup.
11698 (armv8.4-a): New arch.
11699 * config/arm/arm-tables.opt: Regenerate.
11700 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11701 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11702 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11703 Add matching rules for -march=armv8.4-a and extensions.
11704 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11706 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11709 * config/rx/rx.md (BW): New mode attribute.
11710 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11712 2018-01-11 Richard Biener <rguenther@suse.de>
11714 PR tree-optimization/83435
11715 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11716 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11717 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11719 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11720 Alan Hayward <alan.hayward@arm.com>
11721 David Sherwood <david.sherwood@arm.com>
11723 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11725 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11726 (aarch64_print_address_internal): Use it to check for a zero offset.
11728 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11729 Alan Hayward <alan.hayward@arm.com>
11730 David Sherwood <david.sherwood@arm.com>
11732 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11733 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11734 Return a poly_int64 rather than a HOST_WIDE_INT.
11735 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11736 rather than a HOST_WIDE_INT.
11737 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11738 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11739 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11740 final_offset from HOST_WIDE_INT to poly_int64.
11741 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11742 to_constant when getting the number of units in an Advanced SIMD
11744 (aarch64_builtin_vectorized_function): Check for a constant number
11746 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11748 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11749 attribute instead of GET_MODE_NUNITS.
11750 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11751 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11752 GET_MODE_SIZE for fixed-size registers.
11753 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11754 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11755 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11756 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11757 (aarch64_print_operand, aarch64_print_address_internal)
11758 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11759 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11760 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11761 Handle polynomial GET_MODE_SIZE.
11762 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11763 wider than SImode without modification.
11764 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11765 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11766 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11767 passing and returning SVE modes.
11768 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11769 rather than GEN_INT.
11770 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11771 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11772 (aarch64_allocate_and_probe_stack_space): Likewise.
11773 (aarch64_layout_frame): Cope with polynomial offsets.
11774 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11775 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11776 polynomial offsets.
11777 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11778 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11779 poly_int64 rather than a HOST_WIDE_INT.
11780 (aarch64_get_separate_components, aarch64_process_components)
11781 (aarch64_expand_prologue, aarch64_expand_epilogue)
11782 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11783 (aarch64_anchor_offset): New function, split out from...
11784 (aarch64_legitimize_address): ...here.
11785 (aarch64_builtin_vectorization_cost): Handle polynomial
11786 TYPE_VECTOR_SUBPARTS.
11787 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11789 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11790 number of elements from the PARALLEL rather than the mode.
11791 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11792 rather than GET_MODE_BITSIZE.
11793 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11794 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11795 (aarch64_expand_vec_perm_const_1): Handle polynomial
11796 d->perm.length () and d->perm elements.
11797 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11798 Apply to_constant to d->perm elements.
11799 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11800 polynomial CONST_VECTOR_NUNITS.
11801 (aarch64_move_pointer): Take amount as a poly_int64 rather
11803 (aarch64_progress_pointer): Avoid temporary variable.
11804 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11805 the mode attribute instead of GET_MODE.
11807 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11808 Alan Hayward <alan.hayward@arm.com>
11809 David Sherwood <david.sherwood@arm.com>
11811 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11812 x exists before using it.
11813 (aarch64_add_constant_internal): Rename to...
11814 (aarch64_add_offset_1): ...this. Replace regnum with separate
11815 src and dest rtxes. Handle the case in which they're different,
11816 including when the offset is zero. Replace scratchreg with an rtx.
11817 Use 2 additions if there is no spare register into which we can
11818 move a 16-bit constant.
11819 (aarch64_add_constant): Delete.
11820 (aarch64_add_offset): Replace reg with separate src and dest
11821 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11822 Use aarch64_add_offset_1.
11823 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11824 an rtx rather than an int. Take the delta as a poly_int64
11825 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11826 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11827 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11828 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11829 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11830 and aarch64_add_sp.
11831 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11832 aarch64_add_constant.
11834 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11836 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11837 Use scalar_float_mode.
11839 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11841 * config/aarch64/aarch64-simd.md
11842 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11843 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11844 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11845 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11846 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11847 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11848 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11849 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11850 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11851 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11853 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11856 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11857 targ_options->x_arm_arch_string is non NULL.
11859 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11861 * config/aarch64/aarch64.h
11862 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11864 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11867 * expmed.c (emit_store_flag_force): Swap if const op0
11868 and change VOIDmode to mode of op0.
11870 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11872 PR rtl-optimization/83761
11873 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11874 than bytes to mode_for_size.
11876 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11878 PR middle-end/83189
11879 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11880 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11883 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11885 PR middle-end/83575
11886 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11887 when in layout mode.
11888 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11889 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11892 2018-01-10 Michael Collison <michael.collison@arm.com>
11894 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11895 * config/aarch64/aarch64-option-extension.def: Add
11896 AARCH64_OPT_EXTENSION of 'fp16fml'.
11897 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11898 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11899 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11900 * config/aarch64/constraints.md (Ui7): New constraint.
11901 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11902 (VFMLA_SEL_W): Ditto.
11905 (VFMLA16_LOW): New int iterator.
11906 (VFMLA16_HIGH): Ditto.
11907 (UNSPEC_FMLAL): New unspec.
11908 (UNSPEC_FMLSL): Ditto.
11909 (UNSPEC_FMLAL2): Ditto.
11910 (UNSPEC_FMLSL2): Ditto.
11911 (f16mac): New code attribute.
11912 * config/aarch64/aarch64-simd-builtins.def
11913 (aarch64_fmlal_lowv2sf): Ditto.
11914 (aarch64_fmlsl_lowv2sf): Ditto.
11915 (aarch64_fmlalq_lowv4sf): Ditto.
11916 (aarch64_fmlslq_lowv4sf): Ditto.
11917 (aarch64_fmlal_highv2sf): Ditto.
11918 (aarch64_fmlsl_highv2sf): Ditto.
11919 (aarch64_fmlalq_highv4sf): Ditto.
11920 (aarch64_fmlslq_highv4sf): Ditto.
11921 (aarch64_fmlal_lane_lowv2sf): Ditto.
11922 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11923 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11924 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11925 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11926 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11927 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11928 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11929 (aarch64_fmlal_lane_highv2sf): Ditto.
11930 (aarch64_fmlsl_lane_highv2sf): Ditto.
11931 (aarch64_fmlal_laneq_highv2sf): Ditto.
11932 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11933 (aarch64_fmlalq_lane_highv4sf): Ditto.
11934 (aarch64_fmlsl_lane_highv4sf): Ditto.
11935 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11936 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11937 * config/aarch64/aarch64-simd.md:
11938 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11939 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11940 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11941 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11942 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11943 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11944 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11945 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11946 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11947 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11948 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11949 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11950 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11951 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11952 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11953 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11954 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11955 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11956 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11957 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11958 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11959 (vfmlsl_low_u32): Ditto.
11960 (vfmlalq_low_u32): Ditto.
11961 (vfmlslq_low_u32): Ditto.
11962 (vfmlal_high_u32): Ditto.
11963 (vfmlsl_high_u32): Ditto.
11964 (vfmlalq_high_u32): Ditto.
11965 (vfmlslq_high_u32): Ditto.
11966 (vfmlal_lane_low_u32): Ditto.
11967 (vfmlsl_lane_low_u32): Ditto.
11968 (vfmlal_laneq_low_u32): Ditto.
11969 (vfmlsl_laneq_low_u32): Ditto.
11970 (vfmlalq_lane_low_u32): Ditto.
11971 (vfmlslq_lane_low_u32): Ditto.
11972 (vfmlalq_laneq_low_u32): Ditto.
11973 (vfmlslq_laneq_low_u32): Ditto.
11974 (vfmlal_lane_high_u32): Ditto.
11975 (vfmlsl_lane_high_u32): Ditto.
11976 (vfmlal_laneq_high_u32): Ditto.
11977 (vfmlsl_laneq_high_u32): Ditto.
11978 (vfmlalq_lane_high_u32): Ditto.
11979 (vfmlslq_lane_high_u32): Ditto.
11980 (vfmlalq_laneq_high_u32): Ditto.
11981 (vfmlslq_laneq_high_u32): Ditto.
11982 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11983 (AARCH64_FL_FOR_ARCH8_4): New.
11984 (AARCH64_ISA_F16FML): New ISA flag.
11985 (TARGET_F16FML): New feature flag for fp16fml.
11986 (doc/invoke.texi): Document new fp16fml option.
11988 2018-01-10 Michael Collison <michael.collison@arm.com>
11990 * config/aarch64/aarch64-builtins.c:
11991 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11992 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11993 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11994 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11995 (AARCH64_ISA_SHA3): New ISA flag.
11996 (TARGET_SHA3): New feature flag for sha3.
11997 * config/aarch64/iterators.md (sha512_op): New int attribute.
11998 (CRYPTO_SHA512): New int iterator.
11999 (UNSPEC_SHA512H): New unspec.
12000 (UNSPEC_SHA512H2): Ditto.
12001 (UNSPEC_SHA512SU0): Ditto.
12002 (UNSPEC_SHA512SU1): Ditto.
12003 * config/aarch64/aarch64-simd-builtins.def
12004 (aarch64_crypto_sha512hqv2di): New builtin.
12005 (aarch64_crypto_sha512h2qv2di): Ditto.
12006 (aarch64_crypto_sha512su0qv2di): Ditto.
12007 (aarch64_crypto_sha512su1qv2di): Ditto.
12008 (aarch64_eor3qv8hi): Ditto.
12009 (aarch64_rax1qv2di): Ditto.
12010 (aarch64_xarqv2di): Ditto.
12011 (aarch64_bcaxqv8hi): Ditto.
12012 * config/aarch64/aarch64-simd.md:
12013 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12014 (aarch64_crypto_sha512su0qv2di): Ditto.
12015 (aarch64_crypto_sha512su1qv2di): Ditto.
12016 (aarch64_eor3qv8hi): Ditto.
12017 (aarch64_rax1qv2di): Ditto.
12018 (aarch64_xarqv2di): Ditto.
12019 (aarch64_bcaxqv8hi): Ditto.
12020 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12021 (vsha512h2q_u64): Ditto.
12022 (vsha512su0q_u64): Ditto.
12023 (vsha512su1q_u64): Ditto.
12024 (veor3q_u16): Ditto.
12025 (vrax1q_u64): Ditto.
12026 (vxarq_u64): Ditto.
12027 (vbcaxq_u16): Ditto.
12028 * config/arm/types.md (crypto_sha512): New type attribute.
12029 (crypto_sha3): Ditto.
12030 (doc/invoke.texi): Document new sha3 option.
12032 2018-01-10 Michael Collison <michael.collison@arm.com>
12034 * config/aarch64/aarch64-builtins.c:
12035 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12036 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12037 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12038 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12039 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12040 (AARCH64_ISA_SM4): New ISA flag.
12041 (TARGET_SM4): New feature flag for sm4.
12042 * config/aarch64/aarch64-simd-builtins.def
12043 (aarch64_sm3ss1qv4si): Ditto.
12044 (aarch64_sm3tt1aq4si): Ditto.
12045 (aarch64_sm3tt1bq4si): Ditto.
12046 (aarch64_sm3tt2aq4si): Ditto.
12047 (aarch64_sm3tt2bq4si): Ditto.
12048 (aarch64_sm3partw1qv4si): Ditto.
12049 (aarch64_sm3partw2qv4si): Ditto.
12050 (aarch64_sm4eqv4si): Ditto.
12051 (aarch64_sm4ekeyqv4si): Ditto.
12052 * config/aarch64/aarch64-simd.md:
12053 (aarch64_sm3ss1qv4si): Ditto.
12054 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12055 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12056 (aarch64_sm4eqv4si): Ditto.
12057 (aarch64_sm4ekeyqv4si): Ditto.
12058 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12059 (sm3part_op): Ditto.
12060 (CRYPTO_SM3TT): Ditto.
12061 (CRYPTO_SM3PART): Ditto.
12062 (UNSPEC_SM3SS1): New unspec.
12063 (UNSPEC_SM3TT1A): Ditto.
12064 (UNSPEC_SM3TT1B): Ditto.
12065 (UNSPEC_SM3TT2A): Ditto.
12066 (UNSPEC_SM3TT2B): Ditto.
12067 (UNSPEC_SM3PARTW1): Ditto.
12068 (UNSPEC_SM3PARTW2): Ditto.
12069 (UNSPEC_SM4E): Ditto.
12070 (UNSPEC_SM4EKEY): Ditto.
12071 * config/aarch64/constraints.md (Ui2): New constraint.
12072 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12073 * config/arm/types.md (crypto_sm3): New type attribute.
12074 (crypto_sm4): Ditto.
12075 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12076 (vsm3tt1aq_u32): Ditto.
12077 (vsm3tt1bq_u32): Ditto.
12078 (vsm3tt2aq_u32): Ditto.
12079 (vsm3tt2bq_u32): Ditto.
12080 (vsm3partw1q_u32): Ditto.
12081 (vsm3partw2q_u32): Ditto.
12082 (vsm4eq_u32): Ditto.
12083 (vsm4ekeyq_u32): Ditto.
12084 (doc/invoke.texi): Document new sm4 option.
12086 2018-01-10 Michael Collison <michael.collison@arm.com>
12088 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12089 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12090 (AARCH64_FL_FOR_ARCH8_4): New.
12091 (AARCH64_FL_V8_4): New flag.
12092 (doc/invoke.texi): Document new armv8.4-a option.
12094 2018-01-10 Michael Collison <michael.collison@arm.com>
12096 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12097 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12098 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12099 * config/aarch64/aarch64-option-extension.def: Add
12100 AARCH64_OPT_EXTENSION of 'sha2'.
12101 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12102 (crypto): Disable sha2 and aes if crypto disabled.
12103 (crypto): Enable aes and sha2 if enabled.
12104 (simd): Disable sha2 and aes if simd disabled.
12105 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12107 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12108 (TARGET_SHA2): New feature flag for sha2.
12109 (TARGET_AES): New feature flag for aes.
12110 * config/aarch64/aarch64-simd.md:
12111 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12112 conditional on TARGET_AES.
12113 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12114 (aarch64_crypto_sha1hsi): Make pattern conditional
12116 (aarch64_crypto_sha1hv4si): Ditto.
12117 (aarch64_be_crypto_sha1hv4si): Ditto.
12118 (aarch64_crypto_sha1su1v4si): Ditto.
12119 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12120 (aarch64_crypto_sha1su0v4si): Ditto.
12121 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12122 (aarch64_crypto_sha256su0v4si): Ditto.
12123 (aarch64_crypto_sha256su1v4si): Ditto.
12124 (doc/invoke.texi): Document new aes and sha2 options.
12126 2018-01-10 Martin Sebor <msebor@redhat.com>
12128 PR tree-optimization/83781
12129 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12132 2018-01-11 Martin Sebor <msebor@gmail.com>
12133 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12135 PR tree-optimization/83501
12136 PR tree-optimization/81703
12138 * tree-ssa-strlen.c (get_string_cst): Rename...
12139 (get_string_len): ...to this. Handle global constants.
12140 (handle_char_store): Adjust.
12142 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12143 Jim Wilson <jimw@sifive.com>
12145 * config/riscv/riscv-protos.h (riscv_output_return): New.
12146 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12147 (riscv_attribute_table, riscv_output_return),
12148 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12149 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12150 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12152 (riscv_expand_prologue): Add early return for naked function.
12153 (riscv_expand_epilogue): Likewise.
12154 (riscv_function_ok_for_sibcall): Return false for naked function.
12155 (riscv_set_current_function): New.
12156 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12157 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12158 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12159 * doc/extend.texi (RISC-V Function Attributes): New.
12161 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12163 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12164 check for 128-bit long double before checking TCmode.
12165 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12166 128-bit long doubles before checking TFmode or TCmode.
12167 (FLOAT128_IBM_P): Likewise.
12169 2018-01-10 Martin Sebor <msebor@redhat.com>
12171 PR tree-optimization/83671
12172 * builtins.c (c_strlen): Unconditionally return zero for the empty
12174 Use -Warray-bounds for warnings.
12175 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12176 for non-constant array indices with COMPONENT_REF, arrays of
12177 arrays, and pointers to arrays.
12178 (gimple_fold_builtin_strlen): Determine and set length range for
12179 non-constant character arrays.
12181 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12183 PR middle-end/81897
12184 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12187 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12189 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12191 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12194 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12195 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12196 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12197 indexed_or_indirect_operand predicate.
12198 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12199 (*vsx_le_perm_load_v8hi): Likewise.
12200 (*vsx_le_perm_load_v16qi): Likewise.
12201 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12202 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12203 (*vsx_le_perm_store_v8hi): Likewise.
12204 (*vsx_le_perm_store_v16qi): Likewise.
12205 (eight unnamed splitters): Likewise.
12207 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12209 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12210 * config/rs6000/emmintrin.h: Likewise.
12211 * config/rs6000/mmintrin.h: Likewise.
12212 * config/rs6000/xmmintrin.h: Likewise.
12214 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12217 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12219 * tree.c (tree_nop_conversion): Return true for location wrapper
12221 (maybe_wrap_with_location): New function.
12222 (selftest::check_strip_nops): New function.
12223 (selftest::test_location_wrappers): New function.
12224 (selftest::tree_c_tests): Call it.
12225 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12226 (maybe_wrap_with_location): New decl.
12227 (EXPR_LOCATION_WRAPPER_P): New macro.
12228 (location_wrapper_p): New inline function.
12229 (tree_strip_any_location_wrapper): New inline function.
12231 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12234 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12235 stack_realign_offset for the largest alignment of stack slot
12237 (ix86_find_max_used_stack_alignment): New function.
12238 (ix86_finalize_stack_frame_flags): Use it. Set
12239 max_used_stack_alignment if we don't realign stack.
12240 * config/i386/i386.h (machine_function): Add
12241 max_used_stack_alignment.
12243 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12245 * config/arm/arm.opt (-mbranch-cost): New option.
12246 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12249 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12252 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12253 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12255 2018-01-10 Richard Biener <rguenther@suse.de>
12258 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12259 early out so it also covers the case where we have a non-NULL
12262 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12264 PR tree-optimization/83753
12265 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12266 for non-strided grouped accesses if the number of elements is 1.
12268 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12271 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12272 * i386.h (TARGET_USE_GATHER): Define.
12273 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12275 2018-01-10 Martin Liska <mliska@suse.cz>
12278 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12279 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12281 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12282 CLEANUP_NO_PARTITIONING is not set.
12284 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12286 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12287 for vectors, as a partial revert of r254296.
12288 * rtl.h (const_vec_p): Delete.
12289 (const_vec_duplicate_p): Don't test for vector CONSTs.
12290 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12291 * expmed.c (make_tree): Likewise.
12294 * common.md (E, F): Use CONSTANT_P instead of checking for
12296 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12297 checking for CONST_VECTOR.
12299 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12301 PR middle-end/83575
12302 * predict.c (force_edge_cold): Handle in more sane way edges
12303 with no prediction.
12305 2018-01-09 Carl Love <cel@us.ibm.com>
12307 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12309 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12310 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12311 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12312 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12313 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12314 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12315 * config/rs6000/rs6000-protos.h: Add extern defition for
12316 rs6000_generate_float2_double_code.
12317 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12319 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12320 (float2_v2df): Add define_expand.
12322 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12325 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12326 op_mode in the force_to_mode call.
12328 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12330 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12331 instead of checking each element individually.
12332 (aarch64_evpc_uzp): Likewise.
12333 (aarch64_evpc_zip): Likewise.
12334 (aarch64_evpc_ext): Likewise.
12335 (aarch64_evpc_rev): Likewise.
12336 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12337 instead of checking each element individually. Return true without
12339 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12340 whether all selected elements come from the same input, instead of
12341 checking each element individually. Remove calls to gen_rtx_REG,
12342 start_sequence and end_sequence and instead assert that no rtl is
12345 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12347 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12348 order of HIGH and CONST checks.
12350 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12352 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12353 if the destination isn't an SSA_NAME.
12355 2018-01-09 Richard Biener <rguenther@suse.de>
12357 PR tree-optimization/83668
12358 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12360 (canonicalize_loop_form): ... here, renamed from ...
12361 (canonicalize_loop_closed_ssa_form): ... this and amended to
12362 swap successor edges for loop exit blocks to make us use
12363 the RPO order we need for initial schedule generation.
12365 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12367 PR tree-optimization/64811
12368 * match.pd: When optimizing comparisons with Inf, avoid
12369 introducing or losing exceptions from comparisons with NaN.
12371 2018-01-09 Martin Liska <mliska@suse.cz>
12374 * asan.c (shadow_mem_size): Add gcc_assert.
12376 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12378 Don't save registers in main().
12381 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12382 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12383 * config/avr/avr.c (avr_set_current_function): Don't error if
12384 naked, OS_task or OS_main are specified at the same time.
12385 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12387 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12389 * common/config/avr/avr-common.c (avr_option_optimization_table):
12390 Switch on -mmain-is-OS_task for optimizing compilations.
12392 2018-01-09 Richard Biener <rguenther@suse.de>
12394 PR tree-optimization/83572
12395 * graphite.c: Include cfganal.h.
12396 (graphite_transform_loops): Connect infinite loops to exit
12397 and remove fake edges at the end.
12399 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12401 * ipa-inline.c (edge_badness): Revert accidental checkin.
12403 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12406 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12407 symbols; not inline clones.
12409 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12412 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12413 hard registers. Formatting fixes.
12415 PR preprocessor/83722
12416 * gcc.c (try_generate_repro): Pass
12417 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12418 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12421 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12422 Kito Cheng <kito.cheng@gmail.com>
12424 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12425 (riscv_leaf_function_p): Delete.
12426 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12428 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12430 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12432 (do_ifelse): New function.
12433 (do_isel): New function.
12434 (do_sub3): New function.
12435 (do_add3): New function.
12436 (do_load_mask_compare): New function.
12437 (do_overlap_load_compare): New function.
12438 (expand_compare_loop): New function.
12439 (expand_block_compare): Call expand_compare_loop() when appropriate.
12440 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12441 option description.
12442 (-mblock-compare-inline-loop-limit): New option.
12444 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12447 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12448 Reverse order of second and third operands in first alternative.
12449 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12450 of first and second elements in UNSPEC_VPERMR vector.
12451 (altivec_expand_vec_perm_le): Likewise.
12453 2018-01-08 Jeff Law <law@redhat.com>
12455 PR rtl-optimizatin/81308
12456 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12457 (process_switch): If group_case_labels makes a change, then set
12459 (pass_convert_switch::execute): If a switch is converted, then
12460 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12462 PR rtl-optimization/81308
12463 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12466 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12468 PR target/83663 - Revert r255946
12469 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12470 generation for cases where splatting a value is not useful.
12471 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12472 across a vec_duplicate and a paradoxical subreg forming a vector
12473 mode to a vec_concat.
12475 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12477 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12478 -march=armv8.3-a variants.
12479 * config/arm/t-multilib: Likewise.
12480 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12482 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12484 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12486 (cceq_ior_compare_complement): Give it a name so I can use it, and
12487 change boolean_or_operator predicate to boolean_operator so it can
12488 be used to generate a crand.
12489 (eqne): New code iterator.
12490 (bd/bd_neg): New code_attrs.
12491 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12492 a single define_insn.
12493 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12494 decrement (bdnzt/bdnzf/bdzt/bdzf).
12495 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12496 with the new names of the branch decrement patterns, and added the
12497 names of the branch decrement conditional patterns.
12499 2018-01-08 Richard Biener <rguenther@suse.de>
12501 PR tree-optimization/83563
12502 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12505 2018-01-08 Richard Biener <rguenther@suse.de>
12507 PR middle-end/83713
12508 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12510 2018-01-08 Richard Biener <rguenther@suse.de>
12512 PR tree-optimization/83685
12513 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12514 references to abnormals.
12516 2018-01-08 Richard Biener <rguenther@suse.de>
12519 * dwarf2out.c (output_indirect_strings): Handle empty
12520 skeleton_debug_str_hash.
12521 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12523 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12525 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12526 (emit_store_direct): Likewise.
12527 (arc_trampoline_adjust_address): Likewise.
12528 (arc_asm_trampoline_template): New function.
12529 (arc_initialize_trampoline): Use asm_trampoline_template.
12530 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12531 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12532 * config/arc/arc.md (flush_icache): Delete pattern.
12534 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12536 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12537 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12540 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12543 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12544 by not USED_FOR_TARGET.
12545 (make_pass_resolve_sw_modes): Likewise.
12547 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12549 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12552 2018-01-08 Richard Biener <rguenther@suse.de>
12554 PR middle-end/83580
12555 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12557 2018-01-08 Richard Biener <rguenther@suse.de>
12559 PR middle-end/83517
12560 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12562 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12564 PR middle-end/81897
12565 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12566 basic blocks with a small number of successors.
12567 (convert_control_dep_chain_into_preds): Improve handling of
12569 (dump_predicates): Split apart into...
12570 (dump_pred_chain): ...here...
12571 (dump_pred_info): ...and here.
12572 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12573 (can_chain_union_be_invalidated_p): Improve check for invalidation
12575 (uninit_uses_cannot_happen): Avoid unnecessary if
12576 convert_control_dep_chain_into_preds yielded nothing.
12578 2018-01-06 Martin Sebor <msebor@redhat.com>
12580 PR tree-optimization/83640
12581 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12582 subtracting negative offset from size.
12583 (builtin_access::overlap): Adjust offset bounds of the access to fall
12584 within the size of the object if possible.
12586 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12588 PR rtl-optimization/83699
12589 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12590 extract_bit_field_as_subreg to cases in which the extracted
12591 value is also a vector.
12593 * lra-constraints.c (process_alt_operands): Test for the equivalence
12594 substitutions when detecting a possible reload cycle.
12596 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12599 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12600 by default if flag_selective_schedling{,2}. Formatting fixes.
12602 PR rtl-optimization/83682
12603 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12604 if it has non-VECTOR_MODE element mode.
12605 (vec_duplicate_p): Likewise.
12607 PR middle-end/83694
12608 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12609 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12611 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12614 * config/i386/i386-builtin.def
12615 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12616 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12617 Require also OPTION_MASK_ISA_AVX512F in addition to
12618 OPTION_MASK_ISA_GFNI.
12619 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12620 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12621 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12622 to OPTION_MASK_ISA_GFNI.
12623 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12624 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12625 OPTION_MASK_ISA_AVX512BW.
12626 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12627 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12628 addition to OPTION_MASK_ISA_GFNI.
12629 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12630 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12631 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12632 to OPTION_MASK_ISA_GFNI.
12633 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12634 a requirement for all ISAs rather than any of them with a few
12636 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12638 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12639 bitmasks to be enabled with 3 exceptions, instead of requiring any
12640 enabled ISA with lots of exceptions.
12641 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12642 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12643 Change avx512bw in isa attribute to avx512f.
12644 * config/i386/sgxintrin.h: Add license boilerplate.
12645 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12646 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12647 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12648 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12650 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12651 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12652 temporarily sse2 rather than sse if not enabled already.
12655 * config/i386/sse.md (VI248_VLBW): Rename to ...
12656 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12657 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12658 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12659 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12660 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12661 mode iterator instead of VI248_VLBW.
12663 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12665 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12666 (record_modified): Skip clobbers; add debug output.
12667 (param_change_prob): Use sreal frequencies.
12669 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12671 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12672 punt for user-aligned variables.
12674 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12676 * tree-chrec.c (chrec_contains_symbols): Return true for
12679 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12682 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12683 of (x|y) == x for BICS pattern.
12685 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12687 PR tree-optimization/83605
12688 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12689 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12692 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12694 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12695 * config/epiphany/rtems.h: New file.
12697 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12698 Uros Bizjak <ubizjak@gmail.com>
12701 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12702 QIreg_operand instead of register_operand predicate.
12703 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12704 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12705 comments instead of -fmitigate[-_]rop.
12707 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12710 * cgraphunit.c (symbol_table::compile): Switch to text_section
12711 before calling assembly_start debug hook.
12712 * run-rtl-passes.c (run_rtl_passes): Likewise.
12715 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12717 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12718 range_int_cst_p rather than !symbolic_range_p before calling
12719 extract_range_from_multiplicative_op_1.
12721 2018-01-04 Jeff Law <law@redhat.com>
12723 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12724 redundant test in assertion.
12726 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12728 * doc/rtl.texi: Document machine_mode wrapper classes.
12730 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12732 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12733 using tree_to_uhwi.
12735 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12737 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12738 the VEC_PERM_EXPR fold to fail.
12740 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12743 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12744 to switched_sections.
12746 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12749 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12750 test for d.testing.
12752 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12755 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12756 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12758 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12761 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12762 is BLKmode and bitpos not zero or mode change is needed.
12764 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12767 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12770 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12773 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12774 instead of MULT rtx. Update all corresponding splitters.
12775 (*saddl_se): Ditto.
12776 (*ssub<modesuffix>): Ditto.
12777 (*ssubl_se): Ditto.
12778 (*cmp_sadd_di): Update split patterns.
12779 (*cmp_sadd_si): Ditto.
12780 (*cmp_sadd_sidi): Ditto.
12781 (*cmp_ssub_di): Ditto.
12782 (*cmp_ssub_si): Ditto.
12783 (*cmp_ssub_sidi): Ditto.
12784 * config/alpha/predicates.md (const23_operand): New predicate.
12785 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12786 Look for ASHIFT, not MULT inner operand.
12787 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12789 2018-01-04 Martin Liska <mliska@suse.cz>
12791 PR gcov-profile/83669
12792 * gcov.c (output_intermediate_file): Add version to intermediate
12794 * doc/gcov.texi: Document new field 'version' in intermediate
12795 file format. Fix location of '-k' option of gcov command.
12797 2018-01-04 Martin Liska <mliska@suse.cz>
12800 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12802 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12804 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12806 2018-01-03 Martin Sebor <msebor@redhat.com>
12808 PR tree-optimization/83655
12809 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12810 checking calls with invalid arguments.
12812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12814 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12815 (vectorizable_mask_load_store): Delete.
12816 (vectorizable_call): Return false for masked loads and stores.
12817 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12818 instead of gimple_assign_rhs1.
12819 (vectorizable_load): Handle IFN_MASK_LOAD.
12820 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12824 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12826 (vectorizable_mask_load_store): ...here.
12827 (vectorizable_load): ...and here.
12829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12831 * tree-vect-stmts.c (vect_build_all_ones_mask)
12832 (vect_build_zero_merge_argument): New functions, split out from...
12833 (vectorizable_load): ...here.
12835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12837 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12839 (vectorizable_mask_load_store): ...here.
12840 (vectorizable_store): ...and here.
12842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12844 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12846 (vectorizable_mask_load_store): ...here.
12848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12850 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12851 (vect_model_store_cost): Take a vec_load_store_type instead of a
12853 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12854 (vect_model_store_cost): Take a vec_load_store_type instead of a
12856 (vectorizable_mask_load_store): Update accordingly.
12857 (vectorizable_store): Likewise.
12858 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12862 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12863 IFN_MASK_LOAD calls here rather than...
12864 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12867 Alan Hayward <alan.hayward@arm.com>
12868 David Sherwood <david.sherwood@arm.com>
12870 * expmed.c (extract_bit_field_1): For vector extracts,
12871 fall back to extract_bit_field_as_subreg if vec_extract
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 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12879 they are variable or constant sized.
12880 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12881 slots for constant-sized data.
12883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12884 Alan Hayward <alan.hayward@arm.com>
12885 David Sherwood <david.sherwood@arm.com>
12887 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12888 handling COND_EXPRs with boolean comparisons, try to find a better
12889 basis for the mask type than the boolean itself.
12891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12893 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12894 is calculated and how it can be overridden.
12895 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12896 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12898 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12902 Alan Hayward <alan.hayward@arm.com>
12903 David Sherwood <david.sherwood@arm.com>
12905 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12906 Remove the mode argument.
12907 (aarch64_simd_valid_immediate): Remove the mode and inverse
12909 * config/aarch64/iterators.md (bitsize): New iterator.
12910 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12911 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12912 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12913 aarch64_simd_valid_immediate.
12914 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12915 (aarch64_reg_or_bic_imm): Likewise.
12916 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12917 with an insn_type enum and msl with a modifier_type enum.
12918 Replace element_width with a scalar_mode. Change the shift
12919 to unsigned int. Add constructors for scalar_float_mode and
12920 scalar_int_mode elements.
12921 (aarch64_vect_float_const_representable_p): Delete.
12922 (aarch64_can_const_movi_rtx_p)
12923 (aarch64_simd_scalar_immediate_valid_for_move)
12924 (aarch64_simd_make_constant): Update call to
12925 aarch64_simd_valid_immediate.
12926 (aarch64_advsimd_valid_immediate_hs): New function.
12927 (aarch64_advsimd_valid_immediate): Likewise.
12928 (aarch64_simd_valid_immediate): Remove mode and inverse
12929 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12930 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12931 and aarch64_float_const_representable_p on the result.
12932 (aarch64_output_simd_mov_immediate): Remove mode argument.
12933 Update call to aarch64_simd_valid_immediate and use of
12934 simd_immediate_info.
12935 (aarch64_output_scalar_simd_mov_immediate): Update call
12938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12939 Alan Hayward <alan.hayward@arm.com>
12940 David Sherwood <david.sherwood@arm.com>
12942 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12943 (mode_nunits): Likewise CONST_MODE_NUNITS.
12944 * machmode.def (ADJUST_NUNITS): Document.
12945 * genmodes.c (mode_data::need_nunits_adj): New field.
12946 (blank_mode): Update accordingly.
12947 (adj_nunits): New variable.
12948 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12950 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12951 listed in adj_nunits.
12952 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12953 listed in adj_nunits. Don't emit case statements for such modes.
12954 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12955 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12956 nothing if adj_nunits is nonnull.
12957 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12958 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12959 (emit_mode_fbit): Update use of print_maybe_const_decl.
12960 (emit_move_size): Likewise. Treat the array as non-const
12962 (emit_mode_adjustments): Handle adj_nunits.
12964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12966 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12967 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12968 (VECTOR_MODES): Use it.
12969 (make_vector_modes): Take the prefix as an argument.
12971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12972 Alan Hayward <alan.hayward@arm.com>
12973 David Sherwood <david.sherwood@arm.com>
12975 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12976 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12977 for MODE_VECTOR_BOOL.
12978 * machmode.def (VECTOR_BOOL_MODE): Document.
12979 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12980 (make_vector_bool_mode): New function.
12981 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12983 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12984 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12986 * stor-layout.c (int_mode_for_mode): Likewise.
12987 * tree.c (build_vector_type_for_mode): Likewise.
12988 * varasm.c (output_constant_pool_2): Likewise.
12989 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12990 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12991 for MODE_VECTOR_BOOL.
12992 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12993 of mode class checks.
12994 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12995 instead of a list of mode class checks.
12996 (expand_vector_scalar_condition): Likewise.
12997 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13000 Alan Hayward <alan.hayward@arm.com>
13001 David Sherwood <david.sherwood@arm.com>
13003 * machmode.h (mode_size): Change from unsigned short to
13005 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13006 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13007 or if measurement_type is not polynomial.
13008 (fixed_size_mode::includes_p): Check for constant-sized modes.
13009 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13010 return a poly_uint16 rather than an unsigned short.
13011 (emit_mode_size): Change the type of mode_size from unsigned short
13012 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13013 (emit_mode_adjustments): Cope with polynomial vector sizes.
13014 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13016 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13018 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13019 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13020 * caller-save.c (setup_save_areas): Likewise.
13021 (replace_reg_with_saved_mem): Likewise.
13022 * calls.c (emit_library_call_value_1): Likewise.
13023 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13024 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13025 (gen_lowpart_for_combine): Likewise.
13026 * convert.c (convert_to_integer_1): Likewise.
13027 * cse.c (equiv_constant, cse_insn): Likewise.
13028 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13029 (cselib_subst_to_values): Likewise.
13030 * dce.c (word_dce_process_block): Likewise.
13031 * df-problems.c (df_word_lr_mark_ref): Likewise.
13032 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13033 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13034 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13035 (rtl_for_decl_location): Likewise.
13036 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13037 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13038 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13039 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13040 (expand_expr_real_1): Likewise.
13041 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13042 (pad_below): Likewise.
13043 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13044 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13045 * ira.c (get_subreg_tracking_sizes): Likewise.
13046 * ira-build.c (ira_create_allocno_objects): Likewise.
13047 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13048 (ira_sort_regnos_for_alter_reg): Likewise.
13049 * ira-costs.c (record_operand_costs): Likewise.
13050 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13051 (resolve_simple_move): Likewise.
13052 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13053 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13054 (lra_constraints): Likewise.
13055 (CONST_POOL_OK_P): Reject variable-sized modes.
13056 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13057 (add_pseudo_to_slot, lra_spill): Likewise.
13058 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13059 * optabs-query.c (get_best_extraction_insn): Likewise.
13060 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13061 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13062 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13063 * recog.c (offsettable_address_addr_space_p): Likewise.
13064 * regcprop.c (maybe_mode_change): Likewise.
13065 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13066 * regrename.c (build_def_use): Likewise.
13067 * regstat.c (dump_reg_info): Likewise.
13068 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13069 (find_reloads, find_reloads_subreg_address): Likewise.
13070 * reload1.c (eliminate_regs_1): Likewise.
13071 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13072 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13073 (simplify_binary_operation_1, simplify_subreg): Likewise.
13074 * targhooks.c (default_function_arg_padding): Likewise.
13075 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13076 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13077 (verify_gimple_assign_ternary): Likewise.
13078 * tree-inline.c (estimate_move_cost): Likewise.
13079 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13080 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13081 (get_address_cost_ainc): Likewise.
13082 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13083 (vect_supportable_dr_alignment): Likewise.
13084 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13085 (vectorizable_reduction): Likewise.
13086 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13087 (vectorizable_operation, vectorizable_load): Likewise.
13088 * tree.c (build_same_sized_truth_vector_type): Likewise.
13089 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13090 * var-tracking.c (emit_note_insn_var_location): Likewise.
13091 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13092 (ADDR_VEC_ALIGN): Likewise.
13094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13095 Alan Hayward <alan.hayward@arm.com>
13096 David Sherwood <david.sherwood@arm.com>
13098 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13100 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13101 or if measurement_type is polynomial.
13102 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13103 * combine.c (make_extraction): Likewise.
13104 * dse.c (find_shift_sequence): Likewise.
13105 * dwarf2out.c (mem_loc_descriptor): Likewise.
13106 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13107 (extract_bit_field, extract_low_bits): Likewise.
13108 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13109 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13110 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13111 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13112 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13113 * reload.c (find_reloads): Likewise.
13114 * reload1.c (alter_reg): Likewise.
13115 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13116 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13117 * tree-if-conv.c (predicate_mem_writes): Likewise.
13118 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13119 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13120 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13121 * valtrack.c (dead_debug_insert_temp): Likewise.
13122 * varasm.c (mergeable_constant_section): Likewise.
13123 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13125 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13126 Alan Hayward <alan.hayward@arm.com>
13127 David Sherwood <david.sherwood@arm.com>
13129 * expr.c (expand_assignment): Cope with polynomial mode sizes
13130 when assigning to a CONCAT.
13132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13133 Alan Hayward <alan.hayward@arm.com>
13134 David Sherwood <david.sherwood@arm.com>
13136 * machmode.h (mode_precision): Change from unsigned short to
13138 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13140 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13141 or if measurement_type is not polynomial.
13142 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13143 in which the mode is already known to be a scalar_int_mode.
13144 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13145 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13147 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13148 for GET_MODE_PRECISION.
13149 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13150 for GET_MODE_PRECISION.
13151 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13153 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13154 (expand_field_assignment, make_extraction): Likewise.
13155 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13156 (get_last_value): Likewise.
13157 * convert.c (convert_to_integer_1): Likewise.
13158 * cse.c (cse_insn): Likewise.
13159 * expr.c (expand_expr_real_1): Likewise.
13160 * lra-constraints.c (simplify_operand_subreg): Likewise.
13161 * optabs-query.c (can_atomic_load_p): Likewise.
13162 * optabs.c (expand_atomic_load): Likewise.
13163 (expand_atomic_store): Likewise.
13164 * ree.c (combine_reaching_defs): Likewise.
13165 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13166 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13167 * tree.h (type_has_mode_precision_p): Likewise.
13168 * ubsan.c (instrument_si_overflow): Likewise.
13170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13171 Alan Hayward <alan.hayward@arm.com>
13172 David Sherwood <david.sherwood@arm.com>
13174 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13175 polynomial numbers of units.
13176 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13177 (valid_vector_subparts_p): New function.
13178 (build_vector_type): Remove temporary shim and take the number
13179 of units as a poly_uint64 rather than an int.
13180 (build_opaque_vector_type): Take the number of units as a
13181 poly_uint64 rather than an int.
13182 * tree.c (build_vector_from_ctor): Handle polynomial
13183 TYPE_VECTOR_SUBPARTS.
13184 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13185 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13186 (build_vector_from_val): If the number of units is variable,
13187 use build_vec_duplicate_cst for constant operands and
13188 VEC_DUPLICATE_EXPR otherwise.
13189 (make_vector_type): Remove temporary is_constant ().
13190 (build_vector_type, build_opaque_vector_type): Take the number of
13191 units as a poly_uint64 rather than an int.
13192 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13194 * cfgexpand.c (expand_debug_expr): Likewise.
13195 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13196 (store_constructor, expand_expr_real_1): Likewise.
13197 (const_scalar_mask_from_tree): Likewise.
13198 * fold-const-call.c (fold_const_reduction): Likewise.
13199 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13200 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13201 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13202 (fold_relational_const): Likewise.
13203 (native_interpret_vector): Likewise. Change the size from an
13204 int to an unsigned int.
13205 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13206 TYPE_VECTOR_SUBPARTS.
13207 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13208 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13209 duplicating a non-constant operand into a variable-length vector.
13210 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13211 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13212 * ipa-icf.c (sem_variable::equals): Likewise.
13213 * match.pd: Likewise.
13214 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13215 * print-tree.c (print_node): Likewise.
13216 * stor-layout.c (layout_type): Likewise.
13217 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13218 * tree-cfg.c (verify_gimple_comparison): Likewise.
13219 (verify_gimple_assign_binary): Likewise.
13220 (verify_gimple_assign_ternary): Likewise.
13221 (verify_gimple_assign_single): Likewise.
13222 * tree-pretty-print.c (dump_generic_node): Likewise.
13223 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13224 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13225 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13226 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13227 (vect_shift_permute_load_chain): Likewise.
13228 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13229 (expand_vector_condition, optimize_vector_constructor): Likewise.
13230 (lower_vec_perm, get_compute_type): Likewise.
13231 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13232 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13233 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13234 (vect_recog_mask_conversion_pattern): Likewise.
13235 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13236 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13237 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13238 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13239 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13240 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13241 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13242 (supportable_widening_operation): Likewise.
13243 (supportable_narrowing_operation): Likewise.
13244 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13246 * varasm.c (output_constant): Likewise.
13248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13249 Alan Hayward <alan.hayward@arm.com>
13250 David Sherwood <david.sherwood@arm.com>
13252 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13253 so that both the length == 3 and length != 3 cases set up their
13254 own permute vectors. Add comments explaining why we know the
13255 number of elements is constant.
13256 (vect_permute_load_chain): Likewise.
13258 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13259 Alan Hayward <alan.hayward@arm.com>
13260 David Sherwood <david.sherwood@arm.com>
13262 * machmode.h (mode_nunits): Change from unsigned char to
13264 (ONLY_FIXED_SIZE_MODES): New macro.
13265 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13266 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13267 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13269 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13270 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13271 or if measurement_type is not polynomial.
13272 * genmodes.c (ZERO_COEFFS): New macro.
13273 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13275 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13276 Use ZERO_COEFFS when emitting initializers.
13277 * data-streamer.h (bp_pack_poly_value): New function.
13278 (bp_unpack_poly_value): Likewise.
13279 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13280 for GET_MODE_NUNITS.
13281 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13282 for GET_MODE_NUNITS.
13283 * tree.c (make_vector_type): Remove temporary shim and make
13284 the real function take the number of units as a poly_uint64
13285 rather than an int.
13286 (build_vector_type_for_mode): Handle polynomial nunits.
13287 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13288 * emit-rtl.c (const_vec_series_p_1): Likewise.
13289 (gen_rtx_CONST_VECTOR): Likewise.
13290 * fold-const.c (test_vec_duplicate_folding): Likewise.
13291 * genrecog.c (validate_pattern): Likewise.
13292 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13293 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13294 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13295 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13296 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13297 * rtlanal.c (subreg_get_info): Likewise.
13298 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13299 (vect_grouped_load_supported): Likewise.
13300 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13301 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13302 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13303 (simplify_const_unary_operation, simplify_binary_operation_1)
13304 (simplify_const_binary_operation, simplify_ternary_operation)
13305 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13306 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13307 instead of CONST_VECTOR_NUNITS.
13308 * varasm.c (output_constant_pool_2): Likewise.
13309 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13310 explicit-encoded elements in the XVEC for variable-length vectors.
13312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13314 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13316 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13317 Alan Hayward <alan.hayward@arm.com>
13318 David Sherwood <david.sherwood@arm.com>
13320 * coretypes.h (fixed_size_mode): Declare.
13321 (fixed_size_mode_pod): New typedef.
13322 * builtins.h (target_builtins::x_apply_args_mode)
13323 (target_builtins::x_apply_result_mode): Change type to
13324 fixed_size_mode_pod.
13325 * builtins.c (apply_args_size, apply_result_size, result_vector)
13326 (expand_builtin_apply_args_1, expand_builtin_apply)
13327 (expand_builtin_return): Update accordingly.
13329 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13331 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13332 * cselib.c (cselib_hash_rtx): Likewise.
13333 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13334 CONST_VECTOR encoding.
13336 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13337 Jeff Law <law@redhat.com>
13340 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13341 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13342 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13343 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13346 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13347 explicitly probe *sp in a noreturn function if there were any callee
13348 register saves or frame pointer is needed.
13350 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13353 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13354 BLKmode for ternary, binary or unary expressions.
13357 * var-tracking.c (delete_vta_debug_insn): New inline function.
13358 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13359 insns from get_insns () to NULL instead of each bb separately.
13360 Use delete_vta_debug_insn. No longer static.
13361 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13362 delete_vta_debug_insns callers.
13363 * rtl.h (delete_vta_debug_insns): Declare.
13364 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13365 instead of variable_tracking_main.
13367 2018-01-03 Martin Sebor <msebor@redhat.com>
13369 PR tree-optimization/83603
13370 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13371 arguments past the endof the argument list in functions declared
13372 without a prototype.
13373 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13374 Avoid checking when arguments are null.
13376 2018-01-03 Martin Sebor <msebor@redhat.com>
13379 * doc/extend.texi (attribute const): Fix a typo.
13380 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13381 issuing -Wsuggest-attribute for void functions.
13383 2018-01-03 Martin Sebor <msebor@redhat.com>
13385 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13386 offset_int::from instead of wide_int::to_shwi.
13387 (maybe_diag_overlap): Remove assertion.
13388 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13389 * gimple-ssa-sprintf.c (format_directive): Same.
13390 (parse_directive): Same.
13391 (sprintf_dom_walker::compute_format_length): Same.
13392 (try_substitute_return_value): Same.
13394 2018-01-03 Jeff Law <law@redhat.com>
13396 PR middle-end/83654
13397 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13398 non-constant residual for zero at runtime and avoid probing in
13399 that case. Reorganize code for trailing problem to mirror handling
13402 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13404 PR tree-optimization/83501
13405 * tree-ssa-strlen.c (get_string_cst): New.
13406 (handle_char_store): Call get_string_cst.
13408 2018-01-03 Martin Liska <mliska@suse.cz>
13410 PR tree-optimization/83593
13411 * tree-ssa-strlen.c: Include tree-cfg.h.
13412 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13413 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13414 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13416 (strlen_dom_walker::before_dom_children): Call
13417 gimple_purge_dead_eh_edges. Dump tranformation with details
13419 (strlen_dom_walker::before_dom_children): Update call by adding
13420 new argument cleanup_eh.
13421 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13423 2018-01-03 Martin Liska <mliska@suse.cz>
13426 * cif-code.def (VARIADIC_THUNK): New enum value.
13427 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13430 2018-01-03 Jan Beulich <jbeulich@suse.com>
13432 * sse.md (mov<mode>_internal): Tighten condition for when to use
13433 vmovdqu<ssescalarsize> for TI and OI modes.
13435 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13437 Update copyright years.
13439 2018-01-03 Martin Liska <mliska@suse.cz>
13442 * ipa-visibility.c (function_and_variable_visibility): Skip
13443 functions with noipa attribure.
13445 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13447 * gcc.c (process_command): Update copyright notice dates.
13448 * gcov-dump.c (print_version): Ditto.
13449 * gcov.c (print_version): Ditto.
13450 * gcov-tool.c (print_version): Ditto.
13451 * gengtype.c (create_file): Ditto.
13452 * doc/cpp.texi: Bump @copying's copyright year.
13453 * doc/cppinternals.texi: Ditto.
13454 * doc/gcc.texi: Ditto.
13455 * doc/gccint.texi: Ditto.
13456 * doc/gcov.texi: Ditto.
13457 * doc/install.texi: Ditto.
13458 * doc/invoke.texi: Ditto.
13460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13462 * vector-builder.h (vector_builder::m_full_nelts): Change from
13463 unsigned int to poly_uint64.
13464 (vector_builder::full_nelts): Update prototype accordingly.
13465 (vector_builder::new_vector): Likewise.
13466 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13467 (vector_builder::operator ==): Likewise.
13468 (vector_builder::finalize): Likewise.
13469 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13470 Take the number of elements as a poly_uint64 rather than an
13472 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13473 from unsigned int to poly_uint64.
13474 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13475 (vec_perm_indices::new_vector): Likewise.
13476 (vec_perm_indices::length): Likewise.
13477 (vec_perm_indices::nelts_per_input): Likewise.
13478 (vec_perm_indices::input_nelts): Likewise.
13479 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13480 number of elements per input as a poly_uint64 rather than an
13481 unsigned int. Use the original encoding for variable-length
13482 vectors, rather than clamping each individual element.
13483 For the second and subsequent elements in each pattern,
13484 clamp the step and base before clamping their sum.
13485 (vec_perm_indices::series_p): Handle polynomial element counts.
13486 (vec_perm_indices::all_in_range_p): Likewise.
13487 (vec_perm_indices_to_tree): Likewise.
13488 (vec_perm_indices_to_rtx): Likewise.
13489 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13490 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13491 (tree_vector_builder::new_binary_operation): Handle polynomial
13492 element counts. Return false if we need to know the number
13493 of elements at compile time.
13494 * fold-const.c (fold_vec_perm): Punt if the number of elements
13495 isn't known at compile time.
13497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13499 * vec-perm-indices.h (vec_perm_builder): Change element type
13500 from HOST_WIDE_INT to poly_int64.
13501 (vec_perm_indices::element_type): Update accordingly.
13502 (vec_perm_indices::clamp): Handle polynomial element_types.
13503 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13504 (vec_perm_indices::all_in_range_p): Likewise.
13505 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13507 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13508 polynomial vec_perm_indices element types.
13509 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13510 * fold-const.c (fold_vec_perm): Likewise.
13511 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13512 * tree-vect-generic.c (lower_vec_perm): Likewise.
13513 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13514 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13515 element type to HOST_WIDE_INT.
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 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13522 rather than an int. Use plus_constant.
13523 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13524 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13527 Alan Hayward <alan.hayward@arm.com>
13528 David Sherwood <david.sherwood@arm.com>
13530 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13531 a HOST_WIDE_INT to a poly_int64.
13533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13534 Alan Hayward <alan.hayward@arm.com>
13535 David Sherwood <david.sherwood@arm.com>
13537 * calls.c (load_register_parameters): Cope with polynomial
13538 mode sizes. Require a constant size for BLKmode parameters
13539 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13540 forces a parameter to be padded at the lsb end in order to
13541 fill a complete number of words, require the parameter size
13542 to be ordered wrt UNITS_PER_WORD.
13544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13545 Alan Hayward <alan.hayward@arm.com>
13546 David Sherwood <david.sherwood@arm.com>
13548 * reload1.c (spill_stack_slot_width): Change element type
13549 from unsigned int to poly_uint64_pod.
13550 (alter_reg): Treat mode sizes as polynomial.
13552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13553 Alan Hayward <alan.hayward@arm.com>
13554 David Sherwood <david.sherwood@arm.com>
13556 * reload.c (complex_word_subreg_p): New function.
13557 (reload_inner_reg_of_subreg, push_reload): Use it.
13559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13560 Alan Hayward <alan.hayward@arm.com>
13561 David Sherwood <david.sherwood@arm.com>
13563 * lra-constraints.c (process_alt_operands): Reject matched
13564 operands whose sizes aren't ordered.
13565 (match_reload): Refer to this check here.
13567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13568 Alan Hayward <alan.hayward@arm.com>
13569 David Sherwood <david.sherwood@arm.com>
13571 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13572 that the mode size is in the set {1, 2, 4, 8, 16}.
13574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13575 Alan Hayward <alan.hayward@arm.com>
13576 David Sherwood <david.sherwood@arm.com>
13578 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13579 Use plus_constant instead of gen_rtx_PLUS.
13581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13582 Alan Hayward <alan.hayward@arm.com>
13583 David Sherwood <david.sherwood@arm.com>
13585 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13586 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13587 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13588 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13589 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13590 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13591 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13592 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13593 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13594 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13596 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13597 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13598 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13599 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13600 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13601 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13602 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13603 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13604 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13605 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13607 * expr.c (emit_move_resolve_push): Treat the input and result
13608 of PUSH_ROUNDING as a poly_int64.
13609 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13610 (emit_push_insn): Likewise.
13611 * lra-eliminations.c (mark_not_eliminable): Likewise.
13612 * recog.c (push_operand): Likewise.
13613 * reload1.c (elimination_effects): Likewise.
13614 * rtlanal.c (nonzero_bits1): Likewise.
13615 * calls.c (store_one_arg): Likewise. Require the padding to be
13616 known at compile time.
13618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13619 Alan Hayward <alan.hayward@arm.com>
13620 David Sherwood <david.sherwood@arm.com>
13622 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13623 Use plus_constant instead of gen_rtx_PLUS.
13625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13626 Alan Hayward <alan.hayward@arm.com>
13627 David Sherwood <david.sherwood@arm.com>
13629 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13630 rather than an int.
13632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13633 Alan Hayward <alan.hayward@arm.com>
13634 David Sherwood <david.sherwood@arm.com>
13636 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13637 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13638 via stack temporaries. Treat the mode size as polynomial too.
13640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13641 Alan Hayward <alan.hayward@arm.com>
13642 David Sherwood <david.sherwood@arm.com>
13644 * expr.c (expand_expr_real_2): When handling conversions involving
13645 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13646 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13647 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
13660 REGMODE_NATURAL_SIZE.
13661 * expmed.c (store_bit_field_1): Likewise.
13662 * expr.c (store_constructor): Likewise.
13663 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13664 and polynomial REGMODE_NATURAL_SIZE.
13665 (gen_lowpart_common): Likewise.
13666 * reginfo.c (record_subregs_of_mode): Likewise.
13667 * rtlanal.c (read_modify_subreg_p): Likewise.
13669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13670 Alan Hayward <alan.hayward@arm.com>
13671 David Sherwood <david.sherwood@arm.com>
13673 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13674 numbers of elements.
13676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13677 Alan Hayward <alan.hayward@arm.com>
13678 David Sherwood <david.sherwood@arm.com>
13680 * match.pd: Cope with polynomial numbers of vector elements.
13682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13683 Alan Hayward <alan.hayward@arm.com>
13684 David Sherwood <david.sherwood@arm.com>
13686 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13687 in a POINTER_PLUS_EXPR.
13689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13690 Alan Hayward <alan.hayward@arm.com>
13691 David Sherwood <david.sherwood@arm.com>
13693 * omp-simd-clone.c (simd_clone_subparts): New function.
13694 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13695 (ipa_simd_modify_function_body): Likewise.
13697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13698 Alan Hayward <alan.hayward@arm.com>
13699 David Sherwood <david.sherwood@arm.com>
13701 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13702 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13703 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13704 (expand_vector_condition, vector_element): Likewise.
13705 (subparts_gt): New function.
13706 (get_compute_type): Use subparts_gt.
13707 (count_type_subparts): Delete.
13708 (expand_vector_operations_1): Use subparts_gt instead of
13709 count_type_subparts.
13711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13712 Alan Hayward <alan.hayward@arm.com>
13713 David Sherwood <david.sherwood@arm.com>
13715 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13716 (vect_compile_time_alias): ...this new function. Do the calculation
13717 on poly_ints rather than trees.
13718 (vect_prune_runtime_alias_test_list): Update call accordingly.
13720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13721 Alan Hayward <alan.hayward@arm.com>
13722 David Sherwood <david.sherwood@arm.com>
13724 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13726 (vect_schedule_slp_instance): Likewise.
13728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13729 Alan Hayward <alan.hayward@arm.com>
13730 David Sherwood <david.sherwood@arm.com>
13732 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13733 constant and extern definitions for variable-length vectors.
13734 (vect_get_constant_vectors): Note that the number of units
13735 is known to be constant.
13737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13738 Alan Hayward <alan.hayward@arm.com>
13739 David Sherwood <david.sherwood@arm.com>
13741 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13742 of units as polynomial. Choose between WIDE and NARROW based
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-stmts.c (simd_clone_subparts): New function.
13750 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13753 Alan Hayward <alan.hayward@arm.com>
13754 David Sherwood <david.sherwood@arm.com>
13756 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13757 vectors as polynomial. Use build_index_vector for
13758 IFN_GOMP_SIMD_LANE.
13760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13761 Alan Hayward <alan.hayward@arm.com>
13762 David Sherwood <david.sherwood@arm.com>
13764 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13765 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13766 for variable-length vectors.
13767 (vectorizable_mask_load_store): Treat the number of units as
13768 polynomial, asserting that it is constant if the condition has
13769 already been enforced.
13770 (vectorizable_store, vectorizable_load): Likewise.
13772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13773 Alan Hayward <alan.hayward@arm.com>
13774 David Sherwood <david.sherwood@arm.com>
13776 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13777 of units as polynomial. Punt if we can't tell at compile time
13778 which vector contains the final result.
13780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13781 Alan Hayward <alan.hayward@arm.com>
13782 David Sherwood <david.sherwood@arm.com>
13784 * tree-vect-loop.c (vectorizable_induction): Treat the number
13785 of units as polynomial. Punt on SLP inductions. Use an integer
13786 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13787 cast of such a series for variable-length floating-point
13790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13791 Alan Hayward <alan.hayward@arm.com>
13792 David Sherwood <david.sherwood@arm.com>
13794 * tree.h (build_index_vector): Declare.
13795 * tree.c (build_index_vector): New function.
13796 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13797 of units as polynomial, forcibly converting it to a constant if
13798 vectorizable_reduction has already enforced the condition.
13799 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13800 to create a {1,2,3,...} vector.
13801 (vectorizable_reduction): Treat the number of units as polynomial.
13802 Choose vectype_in based on the largest scalar element size rather
13803 than the smallest number of units. Enforce the restrictions
13806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13807 Alan Hayward <alan.hayward@arm.com>
13808 David Sherwood <david.sherwood@arm.com>
13810 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13811 number of units as polynomial.
13813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13814 Alan Hayward <alan.hayward@arm.com>
13815 David Sherwood <david.sherwood@arm.com>
13817 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13818 * target.def (autovectorize_vector_sizes): Return the vector sizes
13819 by pointer, using vector_sizes rather than a bitmask.
13820 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13821 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13822 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13824 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13825 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13826 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13827 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13828 * omp-general.c (omp_max_vf): Likewise.
13829 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13830 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13831 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13832 * tree-vect-slp.c (vect_slp_bb): Likewise.
13833 * doc/tm.texi: Regenerate.
13834 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13836 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13837 the vector size as a poly_uint64 rather than an unsigned int.
13838 (current_vector_size): Change from an unsigned int to a poly_uint64.
13839 (get_vectype_for_scalar_type): Update accordingly.
13840 * tree.h (build_truth_vector_type): Take the size and number of
13841 units as a poly_uint64 rather than an unsigned int.
13842 (build_vector_type): Add a temporary overload that takes
13843 the number of units as a poly_uint64 rather than an unsigned int.
13844 * tree.c (make_vector_type): Likewise.
13845 (build_truth_vector_type): Take the number of units as a poly_uint64
13846 rather than an unsigned int.
13848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13849 Alan Hayward <alan.hayward@arm.com>
13850 David Sherwood <david.sherwood@arm.com>
13852 * target.def (get_mask_mode): Take the number of units and length
13853 as poly_uint64s rather than unsigned ints.
13854 * targhooks.h (default_get_mask_mode): Update accordingly.
13855 * targhooks.c (default_get_mask_mode): Likewise.
13856 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13857 * doc/tm.texi: Regenerate.
13859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13860 Alan Hayward <alan.hayward@arm.com>
13861 David Sherwood <david.sherwood@arm.com>
13863 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13864 * omp-general.c (omp_max_vf): Likewise.
13865 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13866 (expand_omp_simd): Handle polynomial safelen.
13867 * omp-low.c (omplow_simd_context): Add a default constructor.
13868 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13869 (lower_rec_simd_input_clauses): Update accordingly.
13870 (lower_rec_input_clauses): Likewise.
13872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13873 Alan Hayward <alan.hayward@arm.com>
13874 David Sherwood <david.sherwood@arm.com>
13876 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13877 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13878 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13879 (vect_analyze_slp_cost): Likewise.
13880 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13881 (vect_model_load_cost): Likewise.
13883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13884 Alan Hayward <alan.hayward@arm.com>
13885 David Sherwood <david.sherwood@arm.com>
13887 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13888 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13889 from an unsigned int * to a poly_uint64_pod *.
13890 (calculate_unrolling_factor): New function.
13891 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13894 Alan Hayward <alan.hayward@arm.com>
13895 David Sherwood <david.sherwood@arm.com>
13897 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13898 from an unsigned int to a poly_uint64.
13899 (_loop_vec_info::slp_unrolling_factor): Likewise.
13900 (_loop_vec_info::vectorization_factor): Change from an int
13902 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13903 (vect_get_num_vectors): New function.
13904 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13905 (vect_get_num_copies): Use vect_get_num_vectors.
13906 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13907 to an unsigned int *.
13908 (vect_analyze_data_refs): Change min_vf from an int * to a
13910 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13911 than an unsigned HOST_WIDE_INT.
13912 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13913 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13914 to an unsigned int *.
13915 (vect_analyze_data_ref_dependences): Likewise.
13916 (vect_compute_data_ref_alignment): Handle polynomial vf.
13917 (vect_enhance_data_refs_alignment): Likewise.
13918 (vect_prune_runtime_alias_test_list): Likewise.
13919 (vect_shift_permute_load_chain): Likewise.
13920 (vect_supportable_dr_alignment): Likewise.
13921 (dependence_distance_ge_vf): Take the vectorization factor as a
13922 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13923 (vect_analyze_data_refs): Change min_vf from an int * to a
13925 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13926 vfm1 as a poly_uint64 rather than an int. Make the same change
13927 for the returned bound_scalar.
13928 (vect_gen_vector_loop_niters): Handle polynomial vf.
13929 (vect_do_peeling): Likewise. Update call to
13930 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13931 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13933 * tree-vect-loop.c (vect_determine_vectorization_factor)
13934 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13935 (vect_get_known_peeling_cost): Likewise.
13936 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13937 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13938 (vect_transform_loop): Likewise. Use the lowest possible VF when
13939 updating the upper bounds of the loop.
13940 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13941 rather than an int.
13942 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13943 polynomial unroll factors.
13944 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13945 (vect_make_slp_decision): Likewise.
13946 (vect_supported_load_permutation_p): Likewise, and polynomial
13948 (vect_analyze_slp_cost): Handle polynomial vf.
13949 (vect_slp_analyze_node_operations): Likewise.
13950 (vect_slp_analyze_bb_1): Likewise.
13951 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13952 than an unsigned HOST_WIDE_INT.
13953 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13954 (vectorizable_load): Handle polynomial vf.
13955 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13957 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13960 Alan Hayward <alan.hayward@arm.com>
13961 David Sherwood <david.sherwood@arm.com>
13963 * match.pd: Handle bit operations involving three constants
13964 and try to fold one pair.
13966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13968 * tree-vect-loop-manip.c: Include gimple-fold.h.
13969 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13970 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13971 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13972 Add a path that uses a step of VF instead of 1, but disable it
13974 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13975 and niters_no_overflow parameters. Update calls to
13976 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13977 Create a new SSA name if the latter choses to use a ste other
13978 than zero, and return it via niters_vector_mult_vf_var.
13979 * tree-vect-loop.c (vect_transform_loop): Update calls to
13980 vect_do_peeling, vect_gen_vector_loop_niters and
13981 slpeel_make_loop_iterate_ntimes.
13982 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13983 (vect_gen_vector_loop_niters): Update declarations after above changes.
13985 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13987 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13988 128-bit round to integer instructions.
13989 (ceil<mode>2): Likewise.
13990 (btrunc<mode>2): Likewise.
13991 (round<mode>2): Likewise.
13993 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13995 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13996 unaligned VSX load/store on P8/P9.
13997 (expand_block_clear): Allow the use of unaligned VSX
13998 load/store on P8/P9.
14000 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14002 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14004 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14005 swap associated with both a load and a store.
14007 2018-01-02 Andrew Waterman <andrew@sifive.com>
14009 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14010 * config/riscv/riscv.md (clear_cache): Use it.
14012 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14014 * web.c: Remove out-of-date comment.
14016 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14018 * expr.c (fixup_args_size_notes): Check that any existing
14019 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14020 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14021 (emit_single_push_insn): ...here.
14023 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14025 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14026 (const_vector_encoded_nelts): New function.
14027 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14028 (const_vector_int_elt, const_vector_elt): Declare.
14029 * emit-rtl.c (const_vector_int_elt_1): New function.
14030 (const_vector_elt): Likewise.
14031 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14032 of CONST_VECTOR_ELT.
14034 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14036 * expr.c: Include rtx-vector-builder.h.
14037 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14038 directly on the tree encoding.
14039 (const_vector_from_tree): Likewise.
14040 * optabs.c: Include rtx-vector-builder.h.
14041 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14042 sequence of "u" values.
14043 * vec-perm-indices.c: Include rtx-vector-builder.h.
14044 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14045 directly on the vec_perm_indices encoding.
14047 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14049 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14050 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14051 * rtx-vector-builder.h: New file.
14052 * rtx-vector-builder.c: Likewise.
14053 * rtl.h (rtx_def::u2): Add a const_vector field.
14054 (CONST_VECTOR_NPATTERNS): New macro.
14055 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14056 (CONST_VECTOR_DUPLICATE_P): Likewise.
14057 (CONST_VECTOR_STEPPED_P): Likewise.
14058 (CONST_VECTOR_ENCODED_ELT): Likewise.
14059 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14060 (unwrap_const_vec_duplicate): Likewise.
14061 (const_vec_series_p): Check for a non-duplicated vector encoding.
14062 Say that the function only returns true for integer vectors.
14063 * emit-rtl.c: Include rtx-vector-builder.h.
14064 (gen_const_vec_duplicate_1): Delete.
14065 (gen_const_vector): Call gen_const_vec_duplicate instead of
14066 gen_const_vec_duplicate_1.
14067 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14068 (gen_const_vec_duplicate): Use rtx_vector_builder.
14069 (gen_const_vec_series): Likewise.
14070 (gen_rtx_CONST_VECTOR): Likewise.
14071 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14072 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14073 Build a new vector rather than modifying a CONST_VECTOR in-place.
14074 (handle_special_swappables): Update call accordingly.
14075 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14076 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14077 Build a new vector rather than modifying a CONST_VECTOR in-place.
14078 (handle_special_swappables): Update call accordingly.
14080 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14082 * simplify-rtx.c (simplify_const_binary_operation): Use
14083 CONST_VECTOR_ELT instead of XVECEXP.
14085 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14087 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14088 the selector elements to be different from the data elements
14089 if the selector is a VECTOR_CST.
14090 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14091 ssizetype for the selector.
14093 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14095 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14096 before testing each element individually.
14097 * tree-vect-generic.c (lower_vec_perm): Likewise.
14099 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14101 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14102 * selftest-run-tests.c (selftest::run_tests): Call it.
14103 * vector-builder.h (vector_builder::operator ==): New function.
14104 (vector_builder::operator !=): Likewise.
14105 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14106 (vec_perm_indices::all_from_input_p): New function.
14107 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14108 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14109 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14110 instead of reading the VECTOR_CST directly. Detect whether both
14111 vector inputs are the same before constructing the vec_perm_indices,
14112 and update the number of inputs argument accordingly. Use the
14113 utility functions added above. Only construct sel2 if we need to.
14115 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14117 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14118 the broadcast of the low byte.
14119 (expand_mult_highpart): Use an explicit encoding for the permutes.
14120 * optabs-query.c (can_mult_highpart_p): Likewise.
14121 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14122 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14123 (vectorizable_bswap): Likewise.
14124 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14125 explicit encoding for the power-of-2 permutes.
14126 (vect_permute_store_chain): Likewise.
14127 (vect_grouped_load_supported): Likewise.
14128 (vect_permute_load_chain): Likewise.
14130 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14132 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14133 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14134 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14135 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14136 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14137 (vect_gen_perm_mask_any): Likewise.
14139 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14141 * int-vector-builder.h: New file.
14142 * vec-perm-indices.h: Include int-vector-builder.h.
14143 (vec_perm_indices): Redefine as an int_vector_builder.
14144 (auto_vec_perm_indices): Delete.
14145 (vec_perm_builder): Redefine as a stand-alone class.
14146 (vec_perm_indices::vec_perm_indices): New function.
14147 (vec_perm_indices::clamp): Likewise.
14148 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14149 (vec_perm_indices::new_vector): New function.
14150 (vec_perm_indices::new_expanded_vector): Update for new
14151 vec_perm_indices class.
14152 (vec_perm_indices::rotate_inputs): New function.
14153 (vec_perm_indices::all_in_range_p): Operate directly on the
14154 encoded form, without computing elided elements.
14155 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14156 encoding. Update for new vec_perm_indices class.
14157 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14158 the given vec_perm_builder.
14159 (expand_vec_perm_var): Update vec_perm_builder constructor.
14160 (expand_mult_highpart): Use vec_perm_builder instead of
14161 auto_vec_perm_indices.
14162 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14163 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14164 or double series encoding as appropriate.
14165 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14166 vec_perm_indices instead of auto_vec_perm_indices.
14167 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14168 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14169 (vect_permute_store_chain): Likewise.
14170 (vect_grouped_load_supported): Likewise.
14171 (vect_permute_load_chain): Likewise.
14172 (vect_shift_permute_load_chain): Likewise.
14173 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14174 (vect_transform_slp_perm_load): Likewise.
14175 (vect_schedule_slp_instance): Likewise.
14176 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14177 (vectorizable_mask_load_store): Likewise.
14178 (vectorizable_bswap): Likewise.
14179 (vectorizable_store): Likewise.
14180 (vectorizable_load): Likewise.
14181 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14182 vec_perm_indices instead of auto_vec_perm_indices. Use
14183 tree_to_vec_perm_builder to read the vector from a tree.
14184 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14185 vec_perm_builder instead of a vec_perm_indices.
14186 (have_whole_vector_shift): Use vec_perm_builder and
14187 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14188 truncation to calc_vec_perm_mask_for_shift.
14189 (vect_create_epilog_for_reduction): Likewise.
14190 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14191 from auto_vec_perm_indices to vec_perm_indices.
14192 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14193 instead of changing individual elements.
14194 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14195 the vector in d.perm.
14196 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14197 from auto_vec_perm_indices to vec_perm_indices.
14198 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14199 instead of changing individual elements.
14200 (arm_vectorize_vec_perm_const): Use new_vector to install
14201 the vector in d.perm.
14202 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14203 Update vec_perm_builder constructor.
14204 (rs6000_expand_interleave): Likewise.
14205 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14206 (rs6000_expand_interleave): Likewise.
14208 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14210 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14211 to qimode could truncate the indices.
14212 * optabs.c (expand_vec_perm_var): Likewise.
14214 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14216 * Makefile.in (OBJS): Add vec-perm-indices.o.
14217 * vec-perm-indices.h: New file.
14218 * vec-perm-indices.c: Likewise.
14219 * target.h (vec_perm_indices): Replace with a forward class
14221 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14222 * optabs.h: Include vec-perm-indices.h.
14223 (expand_vec_perm): Delete.
14224 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14225 (expand_vec_perm_const): Declare.
14226 * target.def (vec_perm_const_ok): Replace with...
14227 (vec_perm_const): ...this new hook.
14228 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14229 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14230 * doc/tm.texi: Regenerate.
14231 * optabs.def (vec_perm_const): Delete.
14232 * doc/md.texi (vec_perm_const): Likewise.
14233 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14234 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14235 expand_vec_perm for constant permutation vectors. Assert that
14236 the mode of variable permutation vectors is the integer equivalent
14237 of the mode that is being permuted.
14238 * optabs-query.h (selector_fits_mode_p): Declare.
14239 * optabs-query.c: Include vec-perm-indices.h.
14240 (selector_fits_mode_p): New function.
14241 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14242 is defined, instead of checking whether the vec_perm_const_optab
14243 exists. Use targetm.vectorize.vec_perm_const instead of
14244 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14245 fit in the vector mode before using a variable permute.
14246 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14247 vec_perm_indices instead of an rtx.
14248 (expand_vec_perm): Replace with...
14249 (expand_vec_perm_const): ...this new function. Take the selector
14250 as a vec_perm_indices rather than an rtx. Also take the mode of
14251 the selector. Update call to shift_amt_for_vec_perm_mask.
14252 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14253 Use vec_perm_indices::new_expanded_vector to expand the original
14254 selector into bytes. Check whether the indices fit in the vector
14255 mode before using a variable permute.
14256 (expand_vec_perm_var): Make global.
14257 (expand_mult_highpart): Use expand_vec_perm_const.
14258 * fold-const.c: Includes vec-perm-indices.h.
14259 * tree-ssa-forwprop.c: Likewise.
14260 * tree-vect-data-refs.c: Likewise.
14261 * tree-vect-generic.c: Likewise.
14262 * tree-vect-loop.c: Likewise.
14263 * tree-vect-slp.c: Likewise.
14264 * tree-vect-stmts.c: Likewise.
14265 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14267 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14268 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14269 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14270 (aarch64_vectorize_vec_perm_const): ...this new function.
14271 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14272 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14273 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14274 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14275 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14276 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14277 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14279 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14280 check for NEON modes.
14281 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14282 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14283 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14284 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14286 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14287 the old VEC_PERM_CONST conditions.
14288 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14289 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14290 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14291 (ia64_vectorize_vec_perm_const_ok): Merge into...
14292 (ia64_vectorize_vec_perm_const): ...this new function.
14293 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14294 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14295 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14296 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14297 * config/mips/mips.c (mips_expand_vec_perm_const)
14298 (mips_vectorize_vec_perm_const_ok): Merge into...
14299 (mips_vectorize_vec_perm_const): ...this new function.
14300 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14301 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14302 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14303 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14304 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14305 (rs6000_expand_vec_perm_const): Delete.
14306 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14308 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14309 (altivec_expand_vec_perm_const_le): Take each operand individually.
14310 Operate on constant selectors rather than rtxes.
14311 (altivec_expand_vec_perm_const): Likewise. Update call to
14312 altivec_expand_vec_perm_const_le.
14313 (rs6000_expand_vec_perm_const): Delete.
14314 (rs6000_vectorize_vec_perm_const_ok): Delete.
14315 (rs6000_vectorize_vec_perm_const): New function.
14316 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14317 an element count and rtx array.
14318 (rs6000_expand_extract_even): Update call accordingly.
14319 (rs6000_expand_interleave): Likewise.
14320 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14321 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14322 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14323 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14324 (rs6000_expand_vec_perm_const): Delete.
14325 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14326 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14327 (altivec_expand_vec_perm_const_le): Take each operand individually.
14328 Operate on constant selectors rather than rtxes.
14329 (altivec_expand_vec_perm_const): Likewise. Update call to
14330 altivec_expand_vec_perm_const_le.
14331 (rs6000_expand_vec_perm_const): Delete.
14332 (rs6000_vectorize_vec_perm_const_ok): Delete.
14333 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14334 reference to the SPE evmerge intructions.
14335 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14336 an element count and rtx array.
14337 (rs6000_expand_extract_even): Update call accordingly.
14338 (rs6000_expand_interleave): Likewise.
14339 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14340 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14342 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14344 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14346 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14347 vector mode and that that mode matches the mode of the data
14349 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14350 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14351 directly using expand_vec_perm_1 when forcing selectors into
14353 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14355 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14357 * optabs-query.h (can_vec_perm_p): Delete.
14358 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14359 * optabs-query.c (can_vec_perm_p): Split into...
14360 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14361 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14362 particular selector is valid.
14363 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14364 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14365 (vect_grouped_load_supported): Likewise.
14366 (vect_shift_permute_load_chain): Likewise.
14367 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14368 (vect_transform_slp_perm_load): Likewise.
14369 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14370 (vectorizable_bswap): Likewise.
14371 (vect_gen_perm_mask_checked): Likewise.
14372 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14373 implementations of variable permutation vectors into account
14374 when deciding which selector to use.
14375 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14376 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14377 with a false third argument.
14378 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14379 to test whether the constant selector is valid and can_vec_perm_var_p
14380 to test whether a variable selector is valid.
14382 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14384 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14385 * optabs-query.c (can_vec_perm_p): Likewise.
14386 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14387 instead of vec_perm_indices.
14388 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14389 (vect_gen_perm_mask_checked): Likewise,
14390 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14391 (vect_gen_perm_mask_checked): Likewise,
14393 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14395 * optabs-query.h (qimode_for_vec_perm): Declare.
14396 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14397 (qimode_for_vec_perm): ...this new function.
14398 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14400 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14402 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14403 does not have a conditional at the top.
14405 2018-01-02 Richard Biener <rguenther@suse.de>
14407 * ipa-inline.c (big_speedup_p): Fix expression.
14409 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14412 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14415 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14419 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14420 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14421 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14422 cond_taken_branch_cost 3->4.
14424 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14426 PR tree-optimization/83581
14427 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14428 TODO_cleanup_cfg if any changes have been made.
14430 PR middle-end/83608
14431 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14432 convert_modes if target mode has the right side, but different mode
14435 PR middle-end/83609
14436 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14437 last argument when extracting from CONCAT. If either from_real or
14438 from_imag is NULL, use expansion through memory. If result is not
14439 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14440 the parts directly to inner mode, if even that fails, use expansion
14443 PR middle-end/83623
14444 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14445 check for bswap in mode rather than HImode and use that in expand_unop
14448 Copyright (C) 2018 Free Software Foundation, Inc.
14450 Copying and distribution of this file, with or without modification,
14451 are permitted in any medium without royalty provided the copyright
14452 notice and this notice are preserved.