1 2018-05-29 Richard Biener <rguenther@suse.de>
3 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
5 (stmt_vec_info_vec): Make pointer.
6 (init_stmt_vec_info_vec): Remove.
7 (free_stmt_vec_info_vec): Likewise.
8 (set_stmt_vec_info_vec): New function.
9 (free_stmt_vec_infos): Likewise.
10 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
11 (set_vinfo_for_stmt): Likewise.
12 (get_earlier_stmt): Likewise.
13 (get_later_stmt): Likewise.
14 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
15 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
16 (vec_info::~vec_info): Free stmt_vec_infos.
17 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
18 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
19 (pass_slp_vectorize::execute): Likewise.
20 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
21 (free_stmt_vec_info_vec): Likewise.
22 (set_stmt_vec_info_vec): New function.
23 (free_stmt_vec_infos): Likewise.
24 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
25 the global stmt_vec_info_vec.
26 * tree-parloops.c (gather_scalar_reductions): Use
27 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
30 2018-05-29 Richard Biener <rguenther@suse.de>
32 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
34 2018-05-29 Martin Liska <mliska@suse.cz>
35 David Malcolm <dmalcolm@redhat.com>
37 * vec.c (test_reverse): New.
38 (vec_c_tests): Add new test.
39 * vec.h (vl_ptr>::reverse): New function.
41 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
43 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
45 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
48 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
50 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
52 2018-05-28 Richard Biener <rguenther@suse.de>
54 PR tree-optimization/85933
55 * tree-vect-data-refs.c (vect_record_base_alignments): Only
56 look at stmts marked as vectorizable.
58 2018-05-28 Richard Biener <rguenther@suse.de>
60 PR tree-optimization/85934
61 * tree-vect-generic.c (expand_vector_operations_1): Hoist
62 vector boolean check before scalar optimization.
64 2018-05-28 Jakub Jelinek <jakub@redhat.com>
66 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
69 2018-05-28 Mark Wielaard <mark@klomp.org>
71 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
72 if it is an expression containing a minus sign.
74 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
76 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
78 2018-05-27 Paul Koning <ni1d@arrl.net>
80 * config/pdp11/pdp11.md (truncsihi2): Remove.
82 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
83 Chung-Ju Wu <jasonwucj@gmail.com>
85 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
87 (unaligned_store_dw): Ditto.
88 * config/nds32/nds32-memory-manipulation.c
89 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
90 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
91 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
92 (emit_setmem_word_loop): Rename to ...
93 (emit_setmem_doubleword_loop): ... this.
94 (nds32_gen_dup_4_byte_to_word_value): New function.
95 (nds32_gen_dup_8_byte_to_double_word_value): New function.
96 (nds32_expand_setmem_loop): Refine implementation.
97 (nds32_expand_setmem_loop_v3m): Ditto.
98 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
101 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
103 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
105 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
107 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
108 (nds32_init_machine_status): Initialize machine->attr_naked_p and
109 machine->attr_no_prologue_p.
110 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
111 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
112 (nds32_expand_epilogue): Consider attr_naked_p.
113 (nds32_expand_epilogue_v3pop): Likewise.
114 (nds32_can_use_return_insn): Likewise.
115 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
116 attr_no_prologue_p fields.
117 * config/nds32/nds32.opt (mret-in-naked-func): New option.
119 2018-05-27 Jakub Jelinek <jakub@redhat.com>
122 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
125 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
127 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
129 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
131 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
133 (*<floatsuffix>floatv2div2sf2): Rename to ...
134 (*float<floatunssuffix>v2div2sf2): ... this.
135 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
136 (float<floatunssuffix>v2div2sf2_mask): ... this.
137 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
138 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
139 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
141 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
143 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
145 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
147 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
149 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
151 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
152 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
153 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
154 gen_ufix_truncv8dfv8si2.
155 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
156 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
157 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
158 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
159 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
160 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
161 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
162 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
164 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
168 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
170 2018-05-25 Jim Wilson <jimw@sifive.com>
172 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
173 * config/riscv/riscv.c (struct machine_function): Add
174 interrupt_handler_p and attribute_checked_p fields.
175 (riscv_attribute_table): Add interrupt.
176 (riscv_interrupt_type_p): New.
177 (riscv_save_reg_p): Save extra regs for interrupt handler.
178 (riscv_use_save_libcall): Return false for interrupt handler.
179 (riscv_first_stack_step): Add forward declaration.
180 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
181 for interrupt handler with large frame. Use it for saved reg list.
182 (riscv_expand_prologue): Move flag_stack_usage_info support to
183 eliminate duplication.
184 (riscv_expand_epilogue): Generate mret for interrupt handler.
185 (riscv_epilogue_uses): New.
186 (riscv_can_use_return_insn): Return false for interrupt handler.
187 (riscv_function_ok_for_sibcall): Likewise.
188 (riscv_set_current_function): Add interrupt handler support.
189 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
190 * config/riscv/riscv.md (UNSPECV_MRET): New.
192 (riscv_frflags, riscv_fsflags): Use tab after opcode.
194 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
196 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
198 PR tree-optimization/85712
199 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
200 this candidate has already been replaced in-situ by a copy.
202 2018-05-25 Jason Merrill <jason@redhat.com>
204 PR c++/80485 - inline function non-zero address.
205 * symtab.c (nonzero_address): Check DECL_COMDAT.
207 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
210 * config/alpha/alpha.md (ashlsi3): New insn pattern.
211 (*ashlsi_se): Rename from *ashldi_se. Define as sign
212 extension of SImode operation. Use const123_operand predicate.
214 (*saddl_se_1): Ditto.
216 (*ssubl_se_1): Ditto.
217 * config/alpha/predicates.md (const123_operand): New predicate.
218 * config/alpha/constraints.md (P): Use IN_RANGE.
220 2018-05-25 Richard Biener <rguenther@suse.de>
222 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
224 (ref_maybe_used_by_stmt_p): Likewise.
225 (stmt_may_clobber_ref_p): Likewise.
226 (stmt_may_clobber_ref_p_1): Likewise.
227 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
229 (ref_maybe_used_by_stmt_p): Likewise.
230 (stmt_may_clobber_ref_p): Likewise.
231 (stmt_may_clobber_ref_p_1): Likewise.
232 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
233 the alias oracle to disambiguate DRs with stmts DR analysis
235 (vect_analyze_data_refs): Do not give up on not analyzable
236 DRs for BB vectorization. Remove code truncating the dataref
239 2018-05-25 Jakub Jelinek <jakub@redhat.com>
242 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
243 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
244 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
246 2018-05-25 Richard Biener <rguenther@suse.de>
248 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
249 function, combining stmt data ref gathering and fatal analysis
251 (vect_analyze_data_refs): Remove now redudnant code and simplify.
252 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
253 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
254 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
255 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
257 2018-05-25 Bin Cheng <bin.cheng@arm.com>
259 PR tree-optimization/85720
260 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
261 SCC if all partitions are builtins.
262 (version_loop_by_alias_check): New parameter. Generate cancelable
263 runtime alias check if all partitions are builtins.
264 (distribute_loop): Update call to above function.
266 2018-05-25 Bin Cheng <bin.cheng@arm.com>
268 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
269 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
270 (parm_default_def_partition_arg): Ditto.
271 (set_parm_default_def_partition): Ditto.
272 (get_parm_default_def_partitions): Ditto and make it static.
273 (get_undefined_value_partitions): Ditto and make it static.
274 (remove_ssa_form): Refactor call to init_var_map here.
275 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
276 computation for loop region.
277 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
278 (register_default_def): Delete.
279 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
280 (parm_default_def_partition_arg): Ditto.
281 (set_parm_default_def_partition): Ditto.
282 (get_parm_default_def_partitions): Ditto and make it static.
283 (get_undefined_value_partitions): Ditto and make it static.
284 (coalesce_with_default, coalesce_with_default): Update comment.
285 (create_coalesce_list_for_region): New func factored out from
286 create_outofssa_var_map.
287 (populate_coalesce_list_for_outofssa): New func factored out from
288 create_outofssa_var_map and coalesce_ssa_name.
289 (create_outofssa_var_map): Delete.
290 (coalesce_ssa_name): Refactor to support live range computation.
291 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
292 (get_parm_default_def_partitions): Delete.
293 (get_undefined_value_partitions): Ditto.
294 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
295 computation for loop region.
296 (new_tree_live_info, loe_visit_block): Ditto.
297 (live_worklist, set_var_live_on_entry): Ditto.
298 (calculate_live_on_exit, verify_live_on_entry): Ditto.
299 * tree-ssa-live.h (struct _var_map): New fields.
300 (init_var_map): Change decl.
301 (region_contains_p): New.
303 2018-05-25 Bin Cheng <bin.cheng@arm.com>
305 * tree-ssa-live.h (live_merge_and_clear): Delete.
307 2018-05-25 Richard Biener <rguenther@suse.de>
310 * tree-dump.c (dequeue_and_dump): Remove access to removed
311 operand 2 of a SWITCH_EXPR.
313 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
315 * doc/sourcebuild.texi (vect_double_cond_arith): Include
316 multiplication and division.
317 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
318 (cond_udiv@var{m}, cond_umod@var{m}): Document.
319 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
320 (cond_udiv_optab, cond_umod_optab): New optabs.
321 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
322 (IFN_COND_RDIV): New internal functions.
323 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
324 TRUNC_MOD_EXPR and RDIV_EXPR.
325 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
326 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
328 (SVE_INT_BINARY): Include mult.
329 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
330 (optab, sve_int_op): Handle mult.
331 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
333 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
334 for SVE_INT_BINARY_SD.
336 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
338 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
339 (optab, sve_int_op): Handle div and udiv.
340 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
341 for SVE_INT_BINARY_SD.
342 (*<optab><mode>3): New insn for the same.
344 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
346 * tree-vect-patterns.c: Include predict.h.
347 (vect_recog_divmod_pattern): Restrict check for division support
348 to when optimizing for size.
350 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
352 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
353 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
354 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
355 (gimple_match_op::set_op): Likewise.
356 (gimple_resimplify4): Declare.
357 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
358 (expr::gen_transform): Likewise.
359 (decision_tree::gen): Generate a simplification routine for 4 operands.
360 * gimple-match-head.c (gimple_simplify): Add an overload for
361 4 operands. In the top-level function, handle up to 4 call
362 arguments and call gimple_resimplify4.
363 (gimple_resimplify4): New function.
364 (build_call_internal): Pass a fourth operand.
365 (maybe_push_to_seq): Likewise.
366 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
367 Fold VEC_COND_EXPRs of an operation and a default value into
368 an IFN_COND_* function if possible.
369 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
371 (SVE_COND_FP_BINARY): Include them.
372 (optab, sve_fp_op): Handle them.
373 (SVE_INT_BINARY_REV): New code iterator.
374 (SVE_COND_FP_BINARY_REV): New int iterator.
375 (commutative): New int attribute.
376 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
378 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
380 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
381 (*cond_<optab><mode>): New patterns for reversed operands.
383 2018-05-25 Richard Biener <rguenther@suse.de>
385 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
386 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
387 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
388 (STMT_VINFO_GROUPED_ACCESS): Adjust.
389 * tree-vect-data-refs.c (everywhere): Adjust users.
390 * tree-vect-loop.c (everywhere): Likewise.
391 * tree-vect-slp.c (everywhere): Likewise.
392 * tree-vect-stmts.c (everywhere): Likewise.
393 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
395 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
397 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
399 (gcc_cv_as_section_exclude): ... this.
400 Try Solaris as #exclude syntax.
401 * configure: Regenerate.
402 * config.in: Regenerate.
403 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
405 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
406 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
408 * varasm.c (default_elf_asm_named_section): Don't check if
409 HAVE_GAS_SECTION_EXCLUDE is defined.
411 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
413 * doc/md.texi: Update the documentation of the cond_* optabs
414 to mention the new final operand. Fix GET_MODE_NUNITS call.
415 Describe the scalar case too.
416 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
417 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
419 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
420 (get_conditional_internal_fn): Update comment.
421 * tree-vect-loop.c (vectorizable_reduction): Pass the original
422 accumulator value as a final argument to conditional functions.
423 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
424 a define_expand and add an "else" operand. Assert for now that
425 the else operand is equal to operand 2. Use SVE_INT_BINARY and
426 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
427 (*cond_<optab><mode>): New patterns.
428 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
429 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
430 (UNSPEC_COND_EOR): Delete.
431 (optab): Remove associated mappings.
432 (SVE_INT_BINARY): New code iterator.
433 (sve_int_op): Remove int attribute and add "minus" to the code
435 (SVE_COND_INT_OP): Delete.
436 (SVE_COND_FP_OP): Rename to...
437 (SVE_COND_FP_BINARY): ...this.
439 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
441 * optabs.c (can_reuse_operands_p): New function.
442 (maybe_legitimize_operands): Try to reuse the results for
445 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
447 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
448 Add {q} suffix to insn mnemonic.
450 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
452 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
453 (msp430_warn_func_return): New.
455 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
457 * fold-const.c (tree_nonzero_bits): New function.
458 * fold-const.h (tree_nonzero_bits): Likewise.
459 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
460 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
462 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
466 * varasm.c (assemble_alias): Check ifunc_resolver only on
469 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
472 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
473 when memory input operand is handled.
475 2018-05-24 Luis Machado <luis.machado@linaro.org>
477 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
479 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
481 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
483 * match.pd: Delay FMA folds until after vectorization.
485 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
488 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
489 address check not strict.
491 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
493 * gimple-match.h (gimple_match_op): New class.
494 (mprts_hook): Replace parameters with a gimple_match_op *.
495 (maybe_build_generic_op): Likewise.
496 (gimple_simplified_result_is_gimple_val): Replace parameters with
497 a const gimple_match_op *.
498 (gimple_simplify): Replace code_helper * and tree * parameters with
499 a gimple_match_op * parameter.
500 (gimple_resimplify1): Replace code_helper *, tree and tree *
501 parameters with a gimple_match_op * parameter.
502 (gimple_resimplify2): Likewise.
503 (gimple_resimplify3): Likewise.
504 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
505 parameters with a gimple_match_op * parameter.
506 * gimple-match-head.c (gimple_simplify): Change prototypes of
507 auto-generated functions to take a gimple_match_op * instead of
508 separate code_helper * and tree * parameters. Make the same
509 change in the top-level overload and update calls to the
510 gimple_resimplify routines. Update calls to the auto-generated
511 functions and to maybe_push_res_to_seq in the publicly-facing
512 operation-specific gimple_simplify overloads.
513 (gimple_match_op::MAX_NUM_OPS): Define.
514 (gimple_resimplify1): Replace rcode and ops with a single res_op
515 parameter. Update call to gimple_simplify.
516 (gimple_resimplify2): Likewise.
517 (gimple_resimplify3): Likewise.
518 (mprts_hook): Replace parameters with a gimple_match_op *.
519 (maybe_build_generic_op): Likewise.
520 (build_call_internal): Replace type, nargs and ops with
522 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
523 with a single gimple_match_op *. Update calls to mprts_hook,
524 build_call_internal and gimple_simplified_result_is_gimple_val.
525 Factor out code that is common to the tree_code and combined_fn cases.
526 * genmatch.c (expr::gen_transform): Replace tem_code and
527 tem_ops with a gimple_match_op called tem_op. Update calls
528 to the gimple_resimplify functions and maybe_push_res_to_seq.
529 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
530 res_ops. Update call to the gimple_resimplify functions.
531 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
532 (decision_tree::gen): Make the functions take a gimple_match_op *
533 called res_op instead of separate res_code and res_ops parameters.
534 Update call accordingly.
535 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
536 and ops with a single res_op parameter. Update calls to
537 maybe_build_generic_op and maybe_push_res_to_seq.
538 (fold_stmt_1): Update calls to gimple_simplify and
539 replace_stmt_with_simplification.
540 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
541 and gimple_simplified_result_is_gimple_val.
542 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
544 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
545 with a gimple_match_op *.
546 (vn_nary_build_or_lookup): Likewise. Update call to
547 vn_nary_build_or_lookup_1.
548 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
549 gimple_match_op *. Update calls to the gimple_resimplify routines
550 and to gimple_simplified_result_is_gimple_val.
551 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
552 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
553 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
554 (visit_nary_op): Likewise.
555 (visit_reference_op_load): Likewise.
557 2018-05-23 Luis Machado <luis.machado@linaro.org>
559 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
560 modifier for printing the step amount.
562 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
565 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
568 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
570 * doc/sourcebuild.texi (Endianness): New subsubsection.
572 2018-05-23 Luis Machado <luis.machado@linaro.org>
574 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
575 <prefetch_dynamic_strides>: New const bool field.
576 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
577 prefetch_dynamic_strides.
578 (exynosm1_prefetch_tune): Likewise.
579 (thunderxt88_prefetch_tune): Likewise.
580 (thunderx_prefetch_tune): Likewise.
581 (thunderx2t99_prefetch_tune): Likewise.
582 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
584 (aarch64_override_options_internal): Update to set
585 PARAM_PREFETCH_DYNAMIC_STRIDES.
586 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
587 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
588 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
589 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
590 prefetch-dynamic-strides setting.
592 2018-05-23 Luis Machado <luis.machado@linaro.org>
594 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
595 <minimum_stride>: New const int field.
596 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
597 minimum_stride field defaulting to -1.
598 (exynosm1_prefetch_tune): Likewise.
599 (thunderxt88_prefetch_tune): Likewise.
600 (thunderx_prefetch_tune): Likewise.
601 (thunderx2t99_prefetch_tune): Likewise.
602 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
603 <default_opt_level>: Set to 3.
604 (aarch64_override_options_internal): Update to set
605 PARAM_PREFETCH_MINIMUM_STRIDE.
606 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
607 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
608 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
609 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
610 stride is constant and is below the minimum stride threshold.
612 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
614 * config/arm/arm-cpus.in (mode26): Delete.
615 (armv4): Delete mode26 reference.
616 * config/arm/arm.c (arm_configure_build_target): Delete use of
619 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
621 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
623 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
624 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
626 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
627 (floatunsdidf2): Ditto.
629 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
631 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
632 (fixuns_trunc<mode>si2_avx512f): Ditto.
633 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
634 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
635 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
637 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
639 PR rtl-optimization/79985
640 * df-scan.c (df_insn_refs_collect): Remove special case for
641 global registers and asm statements.
643 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
645 * extend.texi (Global Register Variables): Rewrite the bullet list.
646 Note that the register is available for allocation. Note that access
647 via inline asm must use constraints. Add note about async-signal
648 handlers. Remove paragraph about automagic register selection.
650 2018-05-23 Richard Biener <rguenther@suse.de>
652 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
653 of fixed offset from memset VN.
655 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
657 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
659 (alloc_cand_and_find_basis): Initialize first_interp field.
660 (slsr_process_mul): Modify first_interp field.
661 (slsr_process_add): Likewise.
662 (slsr_process_cast): Modify first_interp field for each new
664 (slsr_process_copy): Likewise.
665 (dump_candidate): Dump first_interp field.
666 (replace_mult_candidate): Process all interpretations, not just
668 (replace_rhs_if_not_dup): Likewise.
669 (replace_one_candidate): Likewise.
671 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
673 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
675 (aarch64_needs_frame_chain): New function.
676 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
678 2018-05-23 Sudakshina Das <sudi.das@arm.com>
681 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
682 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
683 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
684 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
685 as true for strict-align.
686 (aarch64_can_inline_p): Perform checks even when callee has no
687 attributes to check for strict alignment.
688 * doc/extend.texi (AArch64 Function Attributes): Document
690 * doc/invoke.texi: (AArch64 Options): Likewise.
692 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
694 PR tree-optimization/85853
695 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
696 the handling of the root of the node to...
697 (vect_slp_analyze_node_operations_1): ...this new function,
698 and run the whole thing with the child nodes' def types
699 set according to their SLP node's def type.
701 2018-05-23 Richard Biener <rguenther@suse.de>
704 * tree-data-ref.c (create_runtime_alias_checks): Defer
705 and ignore overflow warnings.
707 2018-05-23 Yury Gribov <tetra2005@gmail.com>
709 PR tree-optimization/85822
710 * tree-vrp.c (is_masked_range_test): Fix handling of negative
713 2018-05-23 Richard Biener <rguenther@suse.de>
715 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
716 memset constants via native_interpret_expr.
718 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
721 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
723 (cgraph_node::create_alias): Likewise.
724 (cgraph_node::get_availability): Check ifunc_resolver instead
725 of looking up ifunc attribute.
726 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
727 * varasm.c (do_assemble_alias): Likewise.
728 (assemble_alias): Likewise.
729 (default_binds_local_p_3): Likewise.
730 * cgraph.h (cgraph_node): Add ifunc_resolver.
731 (cgraph_node::only_called_directly_or_aliased_p): Return false
733 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
735 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
736 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
737 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
738 instead of looking up ifunc attribute.
740 2018-05-22 Luis Machado <luis.machado@linaro.org>
742 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
744 2018-05-22 Martin Sebor <msebor@redhat.com>
747 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
748 only when expasion succeeds.
749 (expand_builtin_strcmp): Same.
750 (expand_builtin_strncmp): Same.
752 2018-05-22 Martin Sebor <msebor@redhat.com>
754 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
756 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
757 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
759 * config/aarch64/aarch64-ldpstp.md: Replace uses of
760 aarch64_mem_pair_operand with memory_operand and delete operand swapping
762 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
763 Add check for legitimate_address.
764 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
765 (aarch64_swap_ldrstr_operands): New.
766 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
769 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
770 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
772 * config/aarch64/aarch64.md: New patterns to generate stp
774 (store_pair_sw, store_pair_dw): New patterns to generate stp for
775 single words and double words.
776 (load_pair_sw, load_pair_dw): Likewise.
777 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
779 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
781 * config/aarch64/aarch64-ldpstp.md: Modify peephole
782 for different mode ldpstp and add peephole for merged zero stores.
784 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
786 (aarch64_gen_store_pair): Rename calls to match new patterns.
787 (aarch64_gen_load_pair): Rename calls to match new patterns.
788 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
789 (load_pair<DREG:mode><DREG2:mode>): ... This.
790 (store_pair<mode>): Rename to...
791 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
792 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
794 (V_INT_EQUIV): Handle SImode.
795 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
798 2018-05-22 Martin Sebor <msebor@redhat.com>
801 * calls.c (maybe_warn_nonstring_arg): Use string length to set
802 or ajust the presumed bound on an operation to avoid unnecessary
805 2018-05-22 Martin Sebor <msebor@redhat.com>
807 PR tree-optimization/85826
808 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
809 assuming that a DECL necesarily has a constant size.
811 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
814 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
816 2018-05-22 Richard Biener <rguenther@suse.de>
818 PR tree-optimization/85834
819 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
820 non-constant and non-zero memset arguments.
822 2018-05-22 Martin Liska <mliska@suse.cz>
825 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
827 2018-05-22 Richard Biener <rguenther@suse.de>
829 PR tree-optimization/85863
830 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
831 comparisons when vectype is specified.
832 (vectorizable_condition): Do not specify vectype for
833 vect_is_simple_cond when SLP vectorizing.
835 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
838 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
839 define __ibm128 as long double.
840 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
841 as a distinct type when IEEE 128-bit support is enabled.
842 (init_float128_ieee): Fix up conversions between IFmode and IEEE
843 128-bit types to use the correct functions.
844 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
845 convert between 128-bit floating point types that have different
846 modes but the same representation, instead of using gen_lowpart to
848 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
850 (IFKF_reg): New attributes to give the register constraints for
852 (extend<mode>tf2_internal): New insns to mark an explicit
853 conversion between 128-bit floating point types that have a
854 different mode but share the same representation.
856 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
858 PR tree-optimization/85814
859 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
860 a null return from get_strinfo when unsharing the next
861 strinfo in the chain.
863 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
866 * varasm.c (weak_finish): Clean up weak_decls.
868 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
870 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
871 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
872 UNSPEC_UADALP values.
873 * config/aarch64/iterators.md (ABAL): New int iterator.
876 (sur): Add mappings for the above.
877 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
879 (aarch64_<sur>abal<mode>_4): Likewise.
880 (aarch64_<sur>adalp<mode>_3): Likewise.
881 (<sur>sadv16qi): New define_expand.
883 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
885 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
886 (*movdf_internal): Ditto.
887 (*rcpsf2_sse): Ditto.
888 (*rsqrtsf2_sse): Ditto.
889 (*sqrt<mode>2_sse): Ditto.
891 2018-05-21 Tamar Christina <tamar.christina@arm.com>
893 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
895 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
896 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
897 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
898 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
900 * config/aarch64/arm_neon.h: Likewise.
901 * config/aarch64/iterators.md (VQ_I): New.
903 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
905 * config.gcc: Add arc/t-multilib-linux to tmake_file for
907 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
910 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
912 * config/nds32/constraints.md (S): New constraint.
913 * config/nds32/nds32.md (call_internal): Use constraint S.
914 (call_value_internal): Likewise.
915 (sibcall_internal): Likewise.
916 (sibcall_value_internal): Likewise.
918 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
919 Chung-Ju Wu <jasonwucj@gmail.com>
921 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
924 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
925 Chung-Ju Wu <jasonwucj@gmail.com>
927 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
928 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
929 (nds32_rtx_costs_impl): Simplify.
930 (nds32_address_cost_impl): Simplify.
931 (nds32_init_rtx_costs): New function.
932 (nds32_rtx_costs_speed_prefer): Likewise.
933 (nds32_rtx_costs_size_prefer): Likewise.
934 (nds32_address_cost_speed_prefer): Likewise.
935 (nds32_address_cost_speed_fwprop): Likewise.
936 (nds32_address_cost_size_prefer): Likewise.
937 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
938 * config/nds32/nds32.c (nds32_option_override): Use
939 nds32_init_rtx_costs function.
941 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
943 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
944 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
945 (TARGET_PIPELINE_N8): Likewise.
946 (TARGET_PIPELINE_N10): Likewise.
947 (TARGET_PIPELINE_N13): Likewise.
948 (TARGET_PIPELINE_GRAYWOLF): Likewise.
950 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
952 * config/nds32/nds32-fpu.md: Update copyright year.
954 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
956 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
958 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
960 * config/nds32/nds32.c
961 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
962 * config/nds32/nds32.opt (minline-asm-r15): New option.
964 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
966 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
968 * config/nds32/nds32.md (abssi2): New pattern.
970 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
972 * config/i386/i386.md (rex64namesuffix): New mode attribute.
973 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
974 Merge insn pattern from sse_cvtsi2ss<round_name> and
975 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
976 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
977 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
978 using SWI48 mode iterator.
979 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
980 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
981 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
982 pattern from sse_cvttss2si<round_saeonly_name>
983 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
984 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
985 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
986 using SWI48 mode iterator.
987 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
988 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
989 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
990 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
991 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
992 using SWI48 mode iterator.
993 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
994 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
995 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
996 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
997 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
999 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1000 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1001 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1002 pattern from sse_cvttsd2si<round_saeonly_name>
1003 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1005 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1007 * config/nds32/nds32-md-auxiliary.c
1008 (nds32_valid_smw_lwm_base_p): Refine.
1009 (nds32_output_smw_single_word): Refine.
1010 (nds32_output_smw_double_word): New.
1011 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1013 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1015 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1016 (nds32_output_stack_pop): Refine.
1017 (nds32_expand_unaligned_load): Refine.
1018 (nds32_expand_unaligned_store): Refine.
1020 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1021 Chung-Ju Wu <jasonwucj@gmail.com>
1023 * config/nds32/constants.md: Add TP_REGNUM constant.
1024 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1025 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1027 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1028 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1029 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1030 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1031 and PIC code generation.
1032 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1034 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1036 * config/nds32/nds32.md: Support TLS and PIC.
1037 * config/nds32/nds32.c: Support TLS and PIC.
1038 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1039 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1042 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1044 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1045 mode with E_ prefix.
1047 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1048 Chung-Ju Wu <jasonwucj@gmail.com>
1050 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1051 * config/nds32/nds32-md-auxiliary.c
1052 (symbolic_reference_mentioned_p): New.
1053 (nds32_legitimize_ict_address): New.
1054 (nds32_expand_ict_move): New.
1055 (nds32_indirect_call_referenced_p): New.
1056 (nds32_symbol_binds_local_p): Delete.
1057 (nds32_long_call_p): Modify.
1058 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1059 * config/nds32/nds32-protos.h
1060 (symbolic_reference_mentioned_p): Declare.
1061 (nds32_legitimize_ict_address): Declare.
1062 (nds32_expand_ict_move): Declare.
1063 (nds32_indirect_call_referenced_p): Declare.
1064 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1065 (nds32_relax_group): Use nds32_ict_const_p as condition.
1066 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1067 (nds32_asm_file_start): Output ict_model directive in asm code.
1068 (nds32_legitimate_address_p): Consider indirect call.
1069 (nds32_print_operand): Consider indirect call.
1070 (nds32_print_operand_address): Consider indirect call.
1071 (nds32_insert_attributes): Handle "indirect_call" attribute.
1072 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1073 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1074 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1075 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1076 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1077 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1078 (TARGET_ICT_MODEL_SMALL): Define.
1079 (TARGET_ICT_MODEL_LARGE): Define.
1080 * config/nds32/nds32.md (movsi): Consider ict model.
1081 (call, call_value): Consider ict model.
1082 (sibcall, sibcall_value): Consider ict model.
1083 * config/nds32/nds32.opt (mict-model): New option.
1084 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1087 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1088 Monk Chiang <sh.chiang04@gmail.com>
1089 Jim Wilson <jimw@sifive.com>
1091 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1092 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1093 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1094 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1095 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1096 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1097 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1098 compute save_libcall_adjustment properly.
1099 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1100 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1101 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1102 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1103 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1104 (ABI_SPEC): Handle mabi=ilp32e.
1105 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1106 (RVE): Add RVE mask.
1107 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1108 <-march>: Add rv32e as an example.
1110 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1113 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1114 (intra_create_variable_infos): Handle C++ constructors.
1116 2018-05-18 Martin Liska <mliska@suse.cz>
1118 * passes.def: Remove a redundant pass.
1120 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1123 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1125 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1127 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1129 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1131 (armv2, armv2a, armv3, armv3m): Delete architectures.
1132 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1133 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1134 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1136 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1137 (*mulsidi3adddi): Likewise.
1138 (mulsidi3): Likewise.
1139 (*mulsidi3_nov6): Likewise.
1140 (umulsidi3): Likewise.
1141 (umulsidi3_nov6): Likewise.
1142 (umaddsidi4): Likewise.
1143 (*umulsidi3adddi): Likewise.
1144 (smulsi3_highpart): Likewise.
1145 (*smulsi3_highpart_nov6): Likewise.
1146 (umulsi3_highpart): Likewise.
1147 (*umulsi3_highpart_nov6): Likewise.
1148 * config/arm/arm.h (arm_arch3m): Delete.
1149 * config/arm/arm.c (arm_arch3m): Delete.
1150 (arm_option_override_internal): Update armv3-related comment.
1151 (arm_configure_build_target): Delete use of isa_bit_mode32.
1152 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1153 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1154 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1156 (mulusa3): Likewise.
1157 * config/arm/arm-protos.h (arm_arch3m): Delete.
1158 * config/arm/arm-tables.opt: Regenerate.
1159 * config/arm/arm-tune.md: Likewise.
1160 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1161 deleted architectures.
1163 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1165 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1166 (armv5t, armv5te): New features.
1167 (ARMv5, ARMv5e): Delete fgroups.
1168 (ARMv5t, ARMv5te): Adjust for above changes.
1170 (armv5, armv5e): Delete arches.
1171 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1173 (*call_reg_arm): Likewise.
1174 (*call_value_reg_armv5): Likewise.
1175 (*call_value_reg_arm): Likewise.
1176 (*call_symbol): Likewise.
1177 (*call_value_symbol): Likewise.
1178 (*sibcall_insn): Likewise.
1179 (*sibcall_value_insn): Likewise.
1181 (prefetch): Likewise.
1182 (define_split and define_peephole2 dependent on arm_arch5):
1184 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1186 (TARGET_ARM_QBIT): Likewise.
1187 (TARGET_DSP_MULTIPLY): Likewise.
1188 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1189 (arm_arch5, arm_arch5e): Delete.
1190 (arm_arch5t, arm_arch5te): Declare.
1191 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1192 (arm_arch5t): Declare.
1193 (arm_option_reconfigure_globals): Update for the above.
1194 (arm_options_perform_arch_sanity_checks): Update comment, replace
1195 use of arm_arch5 with arm_arch5t.
1196 (use_return_insn): Likewise.
1197 (arm_emit_call_insn): Likewise.
1198 (output_return_instruction): Likewise.
1199 (arm_final_prescan_insn): Likewise.
1200 (arm_coproc_builtin_available): Likewise.
1201 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1202 arm_arch5e with arm_arch5t and arm_arch5te.
1203 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1204 (arm_arch5t, arm_arch5te): Declare.
1205 * config/arm/arm-tables.opt: Regenerate.
1206 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1207 * config/arm/t-multilib: Likewise.
1208 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1209 instead of arm_arch5.
1210 (*call_reg_thumb1): Likewise.
1211 (*call_value_reg_thumb1_v5): Likewise.
1212 (*call_value_reg_thumb1): Likewise.
1213 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1215 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1217 2018-05-18 Martin Liska <mliska@suse.cz>
1219 PR gcov-profile/84846
1220 * doc/gcov.texi: Document -t option of gcov tool.
1222 2018-05-18 Martin Liska <mliska@suse.cz>
1224 PR gcov-profile/84846
1225 * gcov.c (print_usage): Add new -t option.
1226 (process_args): Handle the option.
1227 (generate_results): Use stdout as output when requested by
1230 2018-05-18 Martin Liska <mliska@suse.cz>
1232 PR gcov-profile/84846
1233 * coverage.c (coverage_init): Write PWD to .gcno file.
1234 * doc/gcov.texi: Document how working directory is printed.
1235 * gcov-dump.c (dump_gcov_file): Print PWD.
1236 * gcov.c (output_intermediate_file): Likewise.
1237 (read_graph_file): Read PWD string.
1238 (output_lines): Print PWD.
1240 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1243 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1244 for retval and return false if all args to phi are zero.
1246 2018-05-18 Richard Biener <rguenther@suse.de>
1248 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1250 (evrp_dom_walker::before_dom_children): Call it.
1252 2018-05-18 Richard Biener <rguenther@suse.de>
1254 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1255 results when processing array refs with variable index.
1257 2018-05-18 Toon Moene <toon@moene.org>
1259 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1260 directly after that of -floop-interchange. Indicate that both
1261 options are enabled by default when specifying -O3.
1263 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1265 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1266 iterator. Delete separate integer-mode vec_set<mode> expander.
1267 (aarch64_simd_vec_setv2di): Delete.
1268 (vec_setv2di): Delete.
1269 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1270 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1271 the "w, r" alternative.
1273 2018-05-18 Martin Liska <mliska@suse.cz>
1275 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1276 * tree-pass.h (make_pass_lower_switch_O0): New function.
1277 * tree-switch-conversion.c (node_has_low_bound): Remove.
1278 (node_has_high_bound): Likewise.
1279 (node_is_bounded): Likewise.
1280 (class pass_lower_switch): Make it a template type and create
1282 (pass_lower_switch::execute): Add template argument.
1283 (make_pass_lower_switch): New function.
1284 (make_pass_lower_switch_O0): New function.
1285 (do_jump_if_equal): Remove.
1286 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1287 all the hard work to tree optimization passes.
1289 2018-05-18 Martin Liska <mliska@suse.cz>
1291 * dbgcnt.c (limit_low): Renamed from limit.
1292 (limit_high): New variable.
1293 (dbg_cnt_is_enabled): Check for upper limit.
1294 (dbg_cnt): Adjust dumping.
1295 (dbg_cnt_set_limit_by_index): Add new argument for high
1297 (dbg_cnt_set_limit_by_name): Likewise.
1298 (dbg_cnt_process_single_pair): Parse new format.
1299 (dbg_cnt_process_opt): Use strtok.
1300 (dbg_cnt_list_all_counters): Remove 'value' and add
1302 * doc/invoke.texi: Document changes.
1304 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1306 * doc/sourcebuild.texi (scalar_all_fma): Document.
1307 * tree.def (FMA_EXPR): Delete.
1308 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1309 * internal-fn.c (ternary_direct): New macro.
1310 (expand_ternary_optab_fn): Likewise.
1311 (direct_ternary_optab_supported_p): Likewise.
1312 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1313 * builtins.c (fold_builtin_fma): Delete.
1314 (fold_builtin_3): Don't call it.
1315 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1316 * expr.c (expand_expr_real_2): Likewise.
1317 * fold-const.c (operand_equal_p): Likewise.
1318 (fold_ternary_loc): Likewise.
1319 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1320 * gimple.c (DEFTREECODE): Likewise.
1321 * gimplify.c (gimplify_expr): Likewise.
1322 * optabs-tree.c (optab_for_tree_code): Likewise.
1323 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1324 * tree-eh.c (operation_could_trap_p): Likewise.
1325 (stmt_could_throw_1_p): Likewise.
1326 * tree-inline.c (estimate_operator_cost): Likewise.
1327 * tree-pretty-print.c (dump_generic_node): Likewise.
1328 (op_code_prio): Likewise.
1329 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1330 * tree-ssa-operands.c (get_expr_operands): Likewise.
1331 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1332 * fold-const-call.h (fold_fma): Delete.
1333 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1334 CFN_FNMA and CFN_FNMS.
1336 * genmatch.c (combined_fn): New enum.
1337 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1338 (commutative_op): New function.
1339 (commutate): Use it. Handle more than 2 operands.
1340 (dt_operand::gen_gimple_expr): Use commutative_op.
1341 (parser::parse_expr): Allow :c to be used with non-binary
1342 operators if the commutative operand is known.
1343 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1344 CFN_FMS, CFN_FNMA and CFN_FNMS.
1345 (backprop::process_assign_use): Remove FMA_EXPR handling.
1346 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1347 (gen_hsa_fma): New function.
1348 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1349 IFN_FNMA and IFN_FNMS.
1350 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1351 * gimple-fold.h (follow_all_ssa_edges): Declare.
1352 * gimple-fold.c (follow_all_ssa_edges): New function.
1353 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1354 gimple_build interface and use follow_all_ssa_edges to fold the result.
1355 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1356 instead of checking for optabs directly.
1357 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1358 rather than FMA_EXPRs.
1359 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1360 call to IFN_FMA instead of an FMA_EXPR.
1362 2018-05-17 Jim Wilson <jimw@sifive.com>
1364 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1365 sign extended promoted subreg, and the range does not have the sign bit
1366 set, then do a sign extend.
1368 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1369 test, check for sign extended subreg and/or constant operands, and
1370 do a sign extend in that case.
1372 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1374 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1375 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1377 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1378 Change logics_shift_reg to logics_shift_imm.
1379 (thunderx2t99_fp_loadpair_basic): Delete.
1380 (thunderx2t99_fp_storepair_basic): Delete.
1381 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1382 (thunderx2t99_asimd_polynomial): Delete.
1383 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1384 and neon_fp_mul_d_scalar_q.
1385 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1386 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1387 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1388 (thunderx2t99_asimd_lut): Add missing tbl types.
1389 (thunderx2t99_asimd_ext): Delete.
1390 (thunderx2t99_asimd_load1_1_mult): Delete.
1391 (thunderx2t99_asimd_load1_2_mult): Delete.
1392 (thunderx2t99_asimd_load1_ldp): New.
1393 (thunderx2t99_asimd_load1): New.
1394 (thunderx2t99_asimd_load2): Add missing *load2* types.
1395 (thunderx2t99_asimd_load3): New.
1396 (thunderx2t99_asimd_load4): New.
1397 (thunderx2t99_asimd_store1_1_mult): Delete.
1398 (thunderx2t99_asimd_store1_2_mult): Delete.
1399 (thunderx2t99_asimd_store2_mult): Delete.
1400 (thunderx2t99_asimd_store2_onelane): Delete.
1401 (thunderx2t99_asimd_store_stp): New.
1402 (thunderx2t99_asimd_store1): New.
1403 (thunderx2t99_asimd_store2): New.
1404 (thunderx2t99_asimd_store3): New.
1405 (thunderx2t99_asimd_store4): New.
1407 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1409 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1410 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1412 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1413 Segher Boessenkool <segher@kernel.crashing.org>
1416 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1419 2018-05-17 Richard Biener <rguenther@suse.de>
1421 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1422 for pruning loop and prune defs feeding only already visited PHIs.
1424 2018-05-17 Richard Biener <rguenther@suse.de>
1426 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1428 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1429 Richard Biener <rguenther@suse.de>
1431 PR tree-optimization/85793
1432 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1433 for VMAT_ELEMENTWISE.
1435 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1437 * internal-fn.h (lookup_internal_fn): Declare
1438 * internal-fn.c (lookup_internal_fn): New function.
1439 * gimple.c (gimple_build_call_from_tree): Handle calls to
1441 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1442 internal function names.
1443 * tree-pretty-print.c (dump_generic_node): Likewise.
1444 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1446 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1448 * gimple-fold.h (gimple_build): Make the function forms take
1449 combined_fn rather than built_in_function.
1450 (gimple_simplify): Likewise.
1451 * gimple-match-head.c (gimple_simplify): Likewise.
1452 * gimple-fold.c (gimple_build): Likewise.
1453 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1454 rather than gimple_build_call_internal.
1455 (get_initial_defs_for_reduction): Likewise.
1456 (vect_create_epilog_for_reduction): Likewise.
1457 (vectorizable_live_operation): Likewise.
1459 2018-05-17 Martin Liska <mliska@suse.cz>
1461 * gimple-ssa-sprintf.c (format_directive): Do not use
1462 space in between 'G_' and '('.
1464 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1467 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1468 even if the mask is not all ones.
1471 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1473 (ix86_gimple_fold_builtin): Likewise.
1476 * config/i386/i386.c: Include tree-vector-builder.h.
1477 (ix86_vector_shift_count): New function.
1478 (ix86_fold_builtin): Fold shift builtins by scalar count.
1479 (ix86_gimple_fold_builtin): Likewise.
1481 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1482 _mm512_setzero): New intrinsics.
1484 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1485 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1487 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1488 code generation for cases where splatting a value is not useful.
1489 * simplify-rtx.c (simplify_ternary_operation): Simplify
1490 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1491 mode to a vec_concat.
1493 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1495 * config.gcc: Support "goldmont-plus".
1496 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1498 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1499 PROCESSOR_GOLDMONT_PLUS.
1500 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1501 (processor_target_table): Add "goldmont-plus".
1502 (PTA_GOLDMONT_PLUS): Define.
1503 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1504 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1505 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1506 (fold_builtin_cpu): Add "goldmont-plus".
1507 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1508 (ix86_option_override_internal): Add "goldmont-plus".
1509 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1510 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1511 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1512 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1514 2018-05-17 Richard Biener <rguenther@suse.de>
1516 PR tree-optimization/85757
1517 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1518 remove defs that only feed that PHI from further processing.
1520 2018-05-16 Jim Wilson <jimw@sifive.com>
1522 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1524 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1526 2018-05-16 Mark Wielaard <mark@klomp.org>
1528 * dwarf2out.c (count_index_strings): New function.
1529 (output_indirect_strings): Call count_index_strings and generate
1530 header for dwarf_version >= 5.
1532 2018-05-16 Mark Wielaard <mark@klomp.org>
1534 * dwarf2out.c (dwarf_FORM): New function.
1535 (set_indirect_string): Use dwarf_FORM.
1536 (reset_indirect_string): Likewise.
1537 (size_of_die): Likewise.
1538 (value_format): Likewise.
1539 (output_die): Likewise.
1540 (add_skeleton_AT_string): Likewise.
1541 (output_macinfo_op): Likewise.
1542 (index_string): Likewise.
1543 (output_index_string_offset): Likewise.
1544 (output_index_string): Likewise.
1545 (count_index_strings): Likewise.
1547 2018-05-16 Carl Love <cel@us.ibm.com>
1549 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1550 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1552 2018-05-16 Martin Jambor <mjambor@suse.cz>
1554 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1555 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1557 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1559 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1560 (fnma<mode>4): Likewise.
1561 (fms<mode>4): Likewise.
1562 (fnms<mode>4): Likewise.
1563 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1564 (aarch64_fnma<mode>4): Likewise.
1565 (aarch64_fms<mode>4): Likewise.
1566 (aarch64_fnms<mode>4): Likewise.
1567 (aarch64_fnmadd<mode>4): Likewise.
1569 2018-05-16 Jason Merrill <jason@redhat.com>
1571 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1573 2018-05-16 Richard Biener <rguenther@suse.de>
1575 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1576 (dump_stmt_cost): Declare.
1577 (add_stmt_cost): Dump cost we add.
1578 (add_stmt_costs): New function.
1579 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1581 (vect_analyze_stmt): Adjust prototype.
1582 (vectorizable_condition): Likewise.
1583 (vectorizable_live_operation): Likewise.
1584 (vectorizable_reduction): Likewise.
1585 (vectorizable_induction): Likewise.
1586 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1587 cost vector to pass to vectorizable_ and record afterwards.
1588 (vect_model_reduction_cost): Take cost vector argument and adjust.
1589 (vect_model_induction_cost): Likewise.
1590 (vectorizable_reduction): Likewise.
1591 (vectorizable_induction): Likewise.
1592 (vectorizable_live_operation): Likewise.
1593 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1594 SLP_TREE_NUMBER_OF_VEC_STMTS.
1595 (vect_analyze_slp_cost_1): Remove.
1596 (vect_analyze_slp_cost): Likewise.
1597 (vect_slp_analyze_node_operations): Take visited args and
1598 a target cost vector. Avoid processing already visited stmt sets.
1599 (vect_slp_analyze_operations): Use a local cost vector to gather
1600 costs and register those of non-discarded instances.
1601 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1602 (vect_schedule_slp_instance): Remove copying of
1603 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1605 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1606 adding cost. Record cost entry location.
1607 (vect_prologue_cost_for_slp_op): Function to compute cost of
1608 a constant or invariant generated for SLP vect in the prologue,
1609 split out from vect_analyze_slp_cost_1.
1610 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1611 (vect_model_promotion_demotion_cost): Likewise.
1612 (vect_model_store_cost): Likewise, make static.
1613 (vect_model_load_cost): Likewise.
1614 (vectorizable_bswap): Add cost vector arg and adjust.
1615 (vectorizable_call): Likewise.
1616 (vectorizable_simd_clone_call): Likewise.
1617 (vectorizable_conversion): Likewise.
1618 (vectorizable_assignment): Likewise.
1619 (vectorizable_shift): Likewise.
1620 (vectorizable_operation): Likewise.
1621 (vectorizable_store): Likewise.
1622 (vectorizable_load): Likewise.
1623 (vectorizable_condition): Likewise.
1624 (vectorizable_comparison): Likewise.
1625 (can_vectorize_live_stmts): Likewise.
1626 (vect_analyze_stmt): Likewise.
1627 (vect_transform_stmt): Adjust calls to vectorizable_*.
1628 * tree-vectorizer.c: Include gimple-pretty-print.h.
1629 (dump_stmt_cost): New function.
1631 2018-05-16 Richard Biener <rguenther@suse.de>
1633 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1634 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1635 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1636 (check_name): New callback.
1637 (dse_classify_store): Track cycles via a visited bitmap of PHI
1638 defs and simplify handling of in-loop and across loop dead stores
1639 and properly fail for loop-variant refs. Handle byte-tracking with
1640 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1643 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1645 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1646 (vect_get_mask_type_for_stmt): Likewise.
1647 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1649 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1650 to determine the statement's vector type and the vector type that
1651 should be used for calculating nunits. Deal with cases in which
1652 the type has to be deferred.
1653 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1654 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1655 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1656 (vect_determine_vf_for_stmt): New functions, split out from...
1657 (vect_determine_vectorization_factor): ...here.
1658 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1659 (vect_get_mask_type_for_stmt): New functions, split out from
1660 vect_determine_vectorization_factor.
1662 2018-05-16 Richard Biener <rguenther@suse.de>
1664 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1665 verify the [VEC_]COND_EXPR embedded comparison.
1667 2018-05-15 Martin Sebor <msebor@redhat.com>
1669 PR tree-optimization/85753
1670 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1671 RECORD_TYPE in addition to ARRAY_TYPE.
1673 2018-05-15 Martin Sebor <msebor@redhat.com>
1676 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1678 2018-05-15 Richard Biener <rguenther@suse.de>
1680 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1681 add by_clobber_p one. Change algorithm to collect all defs
1682 representing uses we need to walk and try reducing them to
1683 a single one before failing.
1684 (dse_dom_walker::dse_optimize_stmt): Adjust.
1686 2018-05-13 Mark Wielaard <mark@klomp.org>
1688 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1689 (size_of_loc_descr): Likewise.
1690 (output_loc_operands): Likewise.
1691 (output_loc_operands_raw): Likewise.
1692 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1693 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1694 (hash_loc_operands): Likewise.
1695 (compare_loc_operands): Likewise.
1697 2018-05-14 Mark Wielaard <mark@klomp.org>
1699 * dwarf2out.c (count_index_addrs): New function.
1700 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1702 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1704 PR tree-optimization/83648
1705 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1706 return value as malloc candidate.
1708 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1711 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1712 as true in call to suggest_attribute.
1714 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1716 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1717 -mreadonly-in-sdata.
1719 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1721 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1723 (aarch64_crypto_aesd_fused): Likewise.
1725 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1727 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1728 (movsi_aarch64): Likewise.
1729 (load_pairsi): Likewise.
1730 (load_pairdi): Likewise.
1731 (store_pairsi): Likewise.
1732 (store_pairdi): Likewise.
1733 (load_pairsf): Likewise.
1734 (load_pairdf): Likewise.
1735 (store_pairsf): Likewise.
1736 (store_pairdf): Likewise.
1737 (zero_extend): Likewise.
1738 (trunc): Swap alternatives.
1739 (fcvt_target): Add '?' to prefer w over r.
1741 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1744 * config/i386/i386.md: Disallow non-commutative arithmetics in
1745 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1746 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1747 in the peephole2 before it.
1749 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1751 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1752 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1753 (ix86_handle_option): Handle -mcldemote.
1754 * config.gcc: New header.
1755 * config/i386/cldemoteintrin.h: New file.
1756 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1757 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1759 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1760 OPTION_MASK_ISA_CLDEMOTE.
1761 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1762 (ix86_valid_target_attribute_inner_p): Ditto.
1763 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1764 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1765 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1766 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1767 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1769 * config/i386/i386.opt: Add -mcldemote.
1770 * config/i386/x86intrin.h: New header.
1771 * doc/invoke.texi: Add -mcldemote.
1773 2018-05-14 Richard Biener <rguenther@suse.de>
1775 * doc/match-and-simplify.texi: Adjust :s documentation.
1777 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1779 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1780 intended memcpy size.
1781 (REORDER_45): Likewise.
1783 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1785 * sort.cc: New file.
1786 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1787 * vec.c (qsort_chk): Use gcc_qsort.
1788 * Makefile.in (OBJS-libcommon): Add sort.o.
1789 (build/sort.o): New target. Use it...
1790 (BUILD_RTL): ... here, and...
1791 (build/gencfn-macros): ... here, and...
1792 (build/genmatch): ... here.
1794 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1795 Chung-Ju Wu <jasonwucj@gmail.com>
1797 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1798 * config/nds32/nds32-graywolf.md: New file.
1799 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1800 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1802 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1803 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1804 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1805 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1806 * config/nds32/pipelines.md: Include n15 settings.
1808 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1809 Chung-Ju Wu <jasonwucj@gmail.com>
1811 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1812 * config/nds32/nds32-n13.md: New file.
1813 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1814 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1816 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1817 * config/nds32/nds32.md (pipeline_model): Add n13.
1818 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1819 * config/nds32/pipelines.md: Include n13 settings.
1821 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1822 Chung-Ju Wu <jasonwucj@gmail.com>
1824 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1825 * config/nds32/nds32-n10.md: New file.
1826 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1827 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1829 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1830 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1831 * config/nds32/nds32.md (pipeline_model): Add n10.
1832 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1833 * config/nds32/pipelines.md: Include n10 settings.
1835 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1836 Kito Cheng <kito.cheng@gmail.com>
1837 Chung-Ju Wu <jasonwucj@gmail.com>
1839 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1840 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1841 Add enum values for DSP extension instructions.
1842 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1844 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1845 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1847 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1848 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1849 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1850 * config/nds32/nds32-intrinsic.md: Likewise.
1851 * config/nds32/nds32_intrinsic.h: Likewise.
1852 * config/nds32/nds32-md-auxiliary.c: Likewise.
1853 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1854 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1855 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1856 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1857 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1858 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1859 TYPE_DMAC in switch statement.
1860 * config/nds32/nds32.c: New checking and implementation for DSP
1861 extension instructions.
1862 * config/nds32/nds32.h: Likewise.
1863 * config/nds32/nds32.md: Likewise.
1864 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1865 * config/nds32/predicates.md: Implement new predicates for DSP
1868 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1870 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1871 Reformat alternatives and attributes so it is easier to identify
1872 which constraints/attributes go with which instruction.
1873 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1874 (mov<mode>_softfloat32, FMOVE64): Likewise.
1875 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1876 (mov<mode>_softfloat64, FMOVE64): Likewise.
1878 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1880 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1882 (Basic PowerPC Built-in Functions): The new name of the
1883 subsection previously known as "PowerPC Built-in Functions".
1884 (Basic PowerPC Built-in Functions Available on all Configurations):
1886 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1887 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1888 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1889 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1891 2018-05-11 Martin Jambor <mjambor@suse.cz>
1894 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1897 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1900 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1902 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1904 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1905 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1906 (ix86_handle_option): Handle -mwaitpkg.
1907 * config.gcc: New header.
1908 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1909 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1910 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1912 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1913 OPTION_MASK_ISA_WAITPKG.
1914 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1915 (ix86_option_override_internal): Add PTA_WAITPKG.
1916 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1917 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1918 IX86_BUILTIN_TPAUSE.
1919 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1920 __builtin_ia32_umwait and __builtin_ia32_tpause.
1921 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1922 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1923 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1924 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1925 UNSPECV_TPAUSE): New.
1926 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1927 * config/i386/i386.opt: Add -mwaitpkg.
1928 * config/i386/waitpkgintrin.h: New file.
1929 * config/i386/x86intrin.h: New header.
1930 * doc/invoke.texi: Add -mwaitpkg.
1932 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1935 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1937 (cortex-m0): Use armv6s-m isa.
1938 (cortex-m0plus): Likewise.
1939 (cortex-m1): Likewise.
1940 (cortex-m0.small-multiply): Likewise.
1941 (cortex-m0plus.small-multiply): Likewise.
1942 (cortex-m1.small-multiply): Likewise.
1944 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1945 Jakub Jelinek <jakub@redhat.com>
1947 PR tree-optimization/85692
1948 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1949 source permute as well.
1951 2018-05-11 Martin Liska <mliska@suse.cz>
1954 * doc/extend.texi: Document LLVM style format for no_sanitize
1957 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1959 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1960 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1961 (mode_supports_vsx_dform_quad): Likewise.
1962 (mode_supports_vmx_dform): Move these functions to be next to the
1963 other mode_supports functions.
1964 (mode_supports_dq_form): Likewise.
1965 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1966 mode_supports_dq_form.
1967 (reg_offset_addressing_ok_p): Likewise.
1968 (offsettable_ok_by_alignment): Likewise.
1969 (rs6000_legitimate_offset_address_p): Likewise.
1970 (legitimate_lo_sum_address_p): Likewise.
1971 (rs6000_legitimize_address): Likewise.
1972 (rs6000_legitimize_reload_address): Likewise.
1973 (rs6000_secondary_reload_inner): Likewise.
1974 (rs6000_preferred_reload_class): Likewise.
1975 (rs6000_output_move_128bit): Likewise.
1977 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1979 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1980 Generate SImode target register for null target.
1981 <case IX86_BUILTIN_XGETBV>: Ditto.
1982 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1983 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1985 2018-05-10 Carl Love <cel@us.ibm.com>
1987 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1988 dcbtt and dcbtstt if operands[2] is 0.
1990 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1993 * config/i386/sse.md (usadv64qi): New expander.
1995 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1997 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1998 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1999 -maltivec=be support.
2000 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2001 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2002 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2003 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2004 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2005 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2006 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2007 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2008 altivec_vsumsws): Adjust.
2009 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2010 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2011 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2013 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2014 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2015 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2016 (altivec_lve<VI_char>x): Delete expand.
2017 (*altivec_lve<VI_char>x_internal): Rename to...
2018 (altivec_lve<VI_char>x): ... this.
2019 (altivec_lvxl_<mode>): Delete expand.
2020 (*altivec_lvxl_<mode>_internal): Rename to ...
2021 (altivec_lvxl_<mode>): ... this.
2022 (altivec_stvxl_<mode>): Delete expand.
2023 (*altivec_stvxl_<mode>_internal): Rename to ...
2024 (altivec_stvxl_<mode>): ... this.
2025 (altivec_stve<VI_char>x): Delete expand.
2026 (*altivec_stve<VI_char>x_internal): Rename to ...
2027 (altivec_stve<VI_char>x): ... this.
2028 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2029 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2030 reduc_plus_scal_<mode>): Adjust.
2031 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2033 (rs6000_cpu_cpp_builtins): Adjust.
2034 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2035 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2036 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2037 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2038 -maltivec=be support.
2039 (rs6000_split_vec_extract_var): Adjust.
2040 (rs6000_split_v4si_init): Adjust.
2041 (swap_selector_for_mode): Delete.
2042 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2043 altivec_expand_stvex_be): Delete.
2044 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2045 -maltivec=be support.
2046 (rs6000_gimple_fold_builtin): Ditto.
2047 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2049 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2050 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2051 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2052 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2053 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2054 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2055 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2056 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2057 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2058 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2059 anonymous split): Adjust.
2060 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2061 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2063 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2065 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2066 when --with-gxx-include-dir is also specified.
2067 * configure: Regenerate.
2069 2018-05-09 Jim Wilson <jimw@sifive.com>
2072 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2073 * config/riscv/t-withmultilib: New.
2074 * config/riscv/withmultilib.h: New.
2075 * doc/install.texi: Document RISC-V --with-multilib-list support.
2077 2018-05-09 Richard Biener <rguenther@suse.de>
2079 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2081 (vect_bb_vectorization_profitable_p): Adjust. Compute
2082 actual scalar cost using the cost vector and the add_stmt_cost
2085 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2087 PR rtl-optimization/85645
2088 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2089 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2091 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2093 PR rtl-optimization/85645
2094 * shrink-wrap.c (spread_components): Return a boolean saying if
2095 anything was changed.
2096 (try_shrink_wrapping_separate): Iterate spread_components until
2097 nothing changes anymore.
2099 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2101 PR rtl-optimization/85645
2102 * regrename.c (build_def_use): Also kill the chains that include the
2103 destination of a REG_CFA_REGISTER note.
2105 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2107 PR rtl-optimization/85645
2108 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2109 insn that has a REG_CFA_REGISTER note.
2111 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2113 * cfgexpand.c (expand_clobber): New function.
2114 (expand_gimple_stmt_1): Use it.
2115 * tree-vect-stmts.c (vect_clobber_variable): New function,
2117 (vectorizable_simd_clone_call): ...here.
2118 (vectorizable_store): Emit a clobber either side of an
2119 IFN_STORE_LANES sequence.
2120 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2122 2018-05-09 Tom de Vries <tom@codesourcery.com>
2125 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2126 (define_insn "trap_if_false"): Add exit after trap.
2128 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2130 PR rtl-optimization/85638
2131 * bb-reorder.c: Include common/common-target.h.
2132 (create_forwarder_block): New function extracted from...
2133 (fix_up_crossing_landing_pad): ...here. Rename into...
2134 (dw2_fix_up_crossing_landing_pad): ...this.
2135 (sjlj_fix_up_crossing_landing_pad): New function.
2136 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2137 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2138 from both partitions and exit the loop after one iteration.
2140 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2143 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2145 (Basic PowerPC Built-in Functions): The new name of the
2146 subsection previously known as "PowerPC Built-in Functions".
2147 (Basic PowerPC Built-in Functions Available on all Configurations):
2149 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2151 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2153 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2155 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2158 2018-05-08 Jim Wilson <jimw@sifive.com>
2160 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2161 (LD_EMUL_SUFFIX): New.
2162 (LINK_SPEC): Use it.
2164 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2166 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2168 (Basic PowerPC Built-in Functions): The new name of the
2169 subsection previously known as "PowerPC Built-in Functions".
2170 (Basic PowerPC Built-in Functions Available on all Configurations):
2172 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2174 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2176 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2178 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2181 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2184 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2185 after cmpelim optimization.
2187 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2189 * config.gcc: Support "goldmont".
2190 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2191 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2193 * config/i386/i386.c (m_GOLDMONT): Define.
2194 (processor_target_table): Add "goldmont".
2195 (PTA_GOLDMONT): Define.
2196 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2197 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2198 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2199 (fold_builtin_cpu): Add "goldmont".
2200 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2201 (ix86_option_override_internal): Add "goldmont".
2202 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2203 (processor_type): Add PROCESSOR_GOLDMONT.
2204 * config/i386/i386.md: Add CPU "glm".
2205 * config/i386/glm.md: New file.
2206 * config/i386/x86-tune.def: Add m_GOLDMONT.
2207 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2209 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2212 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2214 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2215 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2216 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2219 * config/i386/i386.c (ix86_fold_builtin): Handle
2220 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2223 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2224 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2226 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2229 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2230 (check_arch): Likewise.
2231 (check_fpu): Return the result rather than printing it.
2232 (end arch): Fix operator precedence.
2233 (end cpu): Likewise.
2234 (END): Print the result from check_fpu.
2236 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2237 Alan Hayward <alan.hayward@arm.com>
2238 David Sherwood <david.sherwood@arm.com>
2240 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2241 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2242 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2243 (*fcmuo<mode>_and): New patterns.
2245 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2247 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2248 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2249 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2250 (cmp_op, sve_imm_con): New code attributes.
2251 (SVE_COND_INT_CMP, imm_con): Delete.
2252 (cmp_op): Remove above unspecs from int attribute.
2253 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2255 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2256 comparison-specific unspecs.
2257 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2258 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2259 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2260 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2261 (*vec_fcm<cmp_op><mode>): Rename to...
2262 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2263 (*vec_fcmuo<mode>): Rename to...
2264 (*fcmuo<mode>): ...this and adjust likewise.
2265 (*pred_fcm<cmp_op><mode>): New pattern.
2266 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2267 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2269 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2271 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2272 (aarch64_emit_sve_predicated_cond): New function.
2273 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2274 (aarch64_emit_unspec_cond_or): Replace with...
2275 (aarch64_emit_sve_or_conds): ...this new function. Use
2276 aarch64_emit_sve_ptrue_op for the individual comparisons and
2277 aarch64_emit_binop to OR them together.
2278 (aarch64_emit_inverted_unspec_cond): Replace with...
2279 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2280 aarch64_emit_sve_ptrue_op for the comparison and
2281 aarch64_emit_unop to invert the result.
2282 (aarch64_expand_sve_vec_cmp_float): Update after the above
2283 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2285 2018-05-07 Nathan Sidwell <nathan@acm.org>
2287 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2288 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2289 (Backwards Compatibility): Likewise.
2291 2018-05-07 Luis Machado <luis.machado@linaro.org>
2295 2018-05-07 Luis Machado <luis.machado@linaro.org>
2297 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2298 <prefetch_dynamic_strides>: New const bool field.
2299 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2300 prefetch_dynamic_strides.
2301 (exynosm1_prefetch_tune): Likewise.
2302 (thunderxt88_prefetch_tune): Likewise.
2303 (thunderx_prefetch_tune): Likewise.
2304 (thunderx2t99_prefetch_tune): Likewise.
2305 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2306 (aarch64_override_options_internal): Update to set
2307 PARAM_PREFETCH_DYNAMIC_STRIDES.
2308 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2309 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2310 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2311 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2312 prefetch-dynamic-strides setting.
2314 2018-05-07 Luis Machado <luis.machado@linaro.org>
2316 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2317 <minimum_stride>: New const int field.
2318 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2319 minimum_stride field.
2320 (exynosm1_prefetch_tune): Likewise.
2321 (thunderxt88_prefetch_tune): Likewise.
2322 (thunderx_prefetch_tune): Likewise.
2323 (thunderx2t99_prefetch_tune): Likewise.
2324 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2325 (aarch64_override_options_internal): Update to set
2326 PARAM_PREFETCH_MINIMUM_STRIDE.
2327 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2328 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2329 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2330 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2331 stride is constant and is below the minimum stride threshold.
2333 2018-05-07 Luis Machado <luis.machado@linaro.org>
2335 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2338 2018-05-07 Luis Machado <luis.machado@linaro.org>
2340 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2341 <prefetch_dynamic_strides>: New const bool field.
2342 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2343 prefetch_dynamic_strides.
2344 (exynosm1_prefetch_tune): Likewise.
2345 (thunderxt88_prefetch_tune): Likewise.
2346 (thunderx_prefetch_tune): Likewise.
2347 (thunderx2t99_prefetch_tune): Likewise.
2348 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2349 (aarch64_override_options_internal): Update to set
2350 PARAM_PREFETCH_DYNAMIC_STRIDES.
2351 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2352 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2353 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2354 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2355 prefetch-dynamic-strides setting.
2357 2018-05-07 Luis Machado <luis.machado@linaro.org>
2359 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2360 <minimum_stride>: New const int field.
2361 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2362 minimum_stride field.
2363 (exynosm1_prefetch_tune): Likewise.
2364 (thunderxt88_prefetch_tune): Likewise.
2365 (thunderx_prefetch_tune): Likewise.
2366 (thunderx2t99_prefetch_tune): Likewise.
2367 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2368 (aarch64_override_options_internal): Update to set
2369 PARAM_PREFETCH_MINIMUM_STRIDE.
2370 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2371 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2372 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2373 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2374 stride is constant and is below the minimum stride threshold.
2376 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2379 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2380 the type is addressable. Don't force op into register if it has
2383 2018-05-05 Roland McGrath <mcgrathr@google.com>
2386 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2387 any section for which we don't know a specific type it should have,
2388 regardless of name. Previously this was done only for the exact
2389 names ".init_array", ".fini_array", and ".preinit_array".
2390 (default_elf_asm_named_section): Add comment about
2391 relationship with default_section_type_flags and SECTION_NOTYPE.
2392 (get_section): Don't consider it a type conflict if one side has
2393 SECTION_NOTYPE and the other doesn't, as long as neither has the
2394 SECTION_BSS et al used in the default_section_type_flags logic.
2396 2018-05-05 Tom de Vries <tom@codesourcery.com>
2399 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2400 (workaround_barsyncs): New function.
2401 (nvptx_reorg): Use workaround_barsyncs.
2402 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2403 (define_expand "nvptx_membar_cta"): New define_expand.
2404 (define_insn "*nvptx_membar_cta"): New insn.
2406 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2408 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2409 To improve optimization opportunities.
2410 * builtin-types.def: The new needed builtin types for the above.
2412 2018-05-04 Richard Biener <rguenther@suse.de>
2414 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2415 * gimple-ssa-store-merging.c
2416 (imm_store_chain_info::output_merged_store): Remove redundant create,
2417 release split_store vector contents on failure.
2418 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2419 scalar stmt vector on cache hit.
2421 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2423 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2425 * config.gcc (powerpc-xilinx-eabi*): Remove.
2426 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2428 (fusion_addis_mem_combo_load): Ditto.
2429 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2431 (rs6000_cpu_cpp_builtins): Ditto.
2432 * config/rs6000/rs6000-linux.c
2433 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2434 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2435 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2437 (rs6000_setup_reg_addr_masks): Ditto.
2438 (rs6000_init_hard_regno_mode_ok): Ditto.
2439 (rs6000_option_override_internal): Ditto.
2440 (legitimate_lo_sum_address_p): Ditto.
2441 (rs6000_legitimize_address): Ditto.
2442 (rs6000_legitimize_reload_address): Ditto.
2443 (rs6000_legitimate_address_p): Ditto.
2444 (abi_v4_pass_in_fpr): Ditto.
2445 (setup_incoming_varargs): Ditto.
2446 (rs6000_gimplify_va_arg): Ditto.
2447 (rs6000_split_multireg_move): Ditto.
2448 (rs6000_savres_strategy): Ditto.
2449 (rs6000_emit_prologue_components): Ditto.
2450 (rs6000_emit_epilogue_components): Ditto.
2451 (rs6000_emit_prologue): Ditto.
2452 (rs6000_emit_epilogue): Ditto.
2453 (rs6000_elf_file_end): Ditto.
2454 (rs6000_function_value): Ditto.
2455 (rs6000_libcall_value): Ditto.
2456 * config/rs6000/rs6000.h: Ditto.
2457 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2458 (TARGET_MINMAX): ... this. New.
2459 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2460 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2461 (*movsi_internal1_single): Delete.
2462 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2463 mfpu=, mxilinx-fpu): Delete.
2464 * config/rs6000/singlefp.h: Delete.
2465 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2466 * config/rs6000/t-rs6000: Ditto.
2467 * config/rs6000/t-xilinx: Delete.
2468 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2469 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2470 (VStype_simple): Delete.
2471 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2472 * config/rs6000/xfpu.h: Delete.
2473 * config/rs6000/xfpu.md: Delete.
2474 * config/rs6000/xilinx.h: Delete.
2475 * config/rs6000/xilinx.opt: Delete.
2476 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2477 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2479 2018-05-04 Tom de Vries <tom@codesourcery.com>
2482 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2485 2018-05-04 Richard Biener <rguenther@suse.de>
2488 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2489 (expand_complex_div_wide): Likewise.
2490 (expand_complex_operations_1): Likewise.
2491 (expand_complex_libcall): Preserve EH info of the original stmt.
2492 (tree_lower_complex): Handle removed blocks.
2493 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2494 on complex multiplication and division libcall builtins.
2496 2018-05-04 Richard Biener <rguenther@suse.de>
2499 * fold-const.c (negate_expr_p): Restrict negation of operand
2500 zero of a division to when we know that can happen without
2502 (fold_negate_expr_1): Likewise.
2504 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2507 * real.h (real_nextafter): Declare.
2508 * real.c (real_nextafter): New function.
2509 * fold-const-call.c (fold_const_nextafter): New function.
2510 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2511 CASE_CFN_NEXTTOWARD.
2512 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2513 even when arg1_mode is different from arg0_mode.
2515 2018-05-03 Nathan Sidwell <nathan@acm.org>
2517 * doc/extend.texi (Deprecated Features): Remove
2519 (Backwards Compatibility): Likewise.
2520 * doc/invoke.texi (C++ Language Options): Likewise.
2521 (C++ Dialect Options): Likewise.
2523 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2526 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2527 _mm512_mask_mullox_epi64): New intrinsics.
2529 2018-05-03 Tom de Vries <tom@codesourcery.com>
2532 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2533 dump files): Add offload-tree.
2535 2018-05-03 Richard Biener <rguenther@suse.de>
2537 PR tree-optimization/85615
2538 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2539 to loops not nested in BBs loop father to avoid creating multi-entry
2542 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2544 PR tree-optimization/70291
2545 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2546 arguments. Change return type to tree. Emit libcall as a new
2547 statement rather than replacing existing one when inplace_p is true.
2548 (expand_complex_multiplication_components): New function.
2549 (expand_complex_multiplication): Expand floating-point complex
2550 multiplication using the above.
2551 (expand_complex_division): Rename inner_type parameter to type.
2552 Update expand_complex_libcall call-site.
2553 (expand_complex_operations_1): Update expand_complex_multiplication
2554 and expand_complex_division call-sites.
2556 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2559 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2560 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2561 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2562 the highest significant bit of the shift count mask is clear. In
2563 check whether and[sq]i3 is needed verify that all significant bits
2564 of the shift count other than the highest are set.
2566 2018-05-02 Tom de Vries <tom@codesourcery.com>
2569 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2570 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2571 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2572 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2573 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2574 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2575 __builtin_goacc_parlevel_size.
2577 2018-05-02 Richard Biener <rguenther@suse.de>
2579 PR tree-optimization/85597
2580 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2581 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2584 2018-05-02 Tom de Vries <tom@codesourcery.com>
2587 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2588 dump files): Add ltrans-tree.
2590 2018-05-02 Tom de Vries <tom@codesourcery.com>
2593 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2594 dump files): Add wpa-ipa.
2596 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2598 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2599 powerpc*-*-linux*paired* target.
2600 * config/rs6000/750cl.h: Delete.
2601 * config/rs6000/paired.h: Delete.
2602 * config/rs6000/paired.md: Delete.
2603 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2605 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2606 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2607 comment. Remove paired float support.
2608 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2609 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2611 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2612 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2614 * config/rs6000/rs6000.c: Remove paired float support.
2615 (paired_expand_vector_init, paired_expand_vector_move,
2616 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2617 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2618 paired_expand_builtin, paired_expand_predicate_builtin,
2619 paired_init_builtins): Delete.
2620 * config/rs6000/rs6000.h: Remove paired float support.
2621 * config/rs6000/rs6000.md: Remove paired float support.
2622 (move_from_CR_ov_bit): Delete.
2623 * config/rs6000/rs6000.opt (mpaired): Delete.
2624 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2625 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2627 2018-05-02 Richard Biener <rguenther@suse.de>
2630 * gimplify.c (gimplify_save_expr): When in SSA form allow
2631 SAVE_EXPRs to compute to SSA vars.
2633 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2636 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2637 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2638 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2639 clobber operands[2], instead use a new pseudo. Formatting fixes.
2641 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2643 PR tree-optimization/85586
2644 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2645 exit early for statements in the same group if the accesses are
2648 2018-05-02 Tom de Vries <tom@codesourcery.com>
2651 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2654 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2656 PR tree-optimization/85143
2657 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2659 2018-05-01 Tom de Vries <tom@codesourcery.com>
2662 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2663 not found" error message.
2665 2018-05-01 Tom de Vries <tom@codesourcery.com>
2668 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2669 * vec.c (test_ordered_remove_if): New function.
2670 (vec_c_tests): Call test_ordered_remove_if.
2671 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2672 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2673 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2674 VEC_ORDERED_REMOVE_IF.
2676 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2678 PR tree-optimization/82665
2679 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2680 pointer subtraction where arguments come from a memchr call.
2682 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2684 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2685 --push-state --as-needed and --pop-state instead of --as-needed and
2686 --no-as-needed if ld supports it.
2687 * configure: Regenerated.
2690 * doc/install.texi2html: Replace _002d with - and _002a with * in
2691 generated html files using sed.
2693 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2696 * gcc-rich-location.c (blank_line_before_p): New function.
2697 (use_new_line): New function.
2698 (gcc_rich_location::add_fixit_insert_formatted): New function.
2699 * gcc-rich-location.h
2700 (gcc_rich_location::add_fixit_insert_formatted): New function.
2702 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2704 * selftest.c (assert_streq): Rename "expected" and "actual" to
2705 "val1" and "val2". Extend NULL-handling to cover both inputs
2706 symmetrically, while still requiring both to be non-NULL for a pass.
2707 * selftest.h (assert_streq): Rename "expected" and "actual" to
2709 (ASSERT_EQ): Likewise.
2710 (ASSERT_EQ_AT): Likewise.
2711 (ASSERT_KNOWN_EQ): Likewise.
2712 (ASSERT_KNOWN_EQ_AT): Likewise.
2713 (ASSERT_NE): Likewise.
2714 (ASSERT_MAYBE_NE): Likewise.
2715 (ASSERT_MAYBE_NE_AT): Likewise.
2716 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2717 the assertion to pass.
2718 (ASSERT_STREQ_AT): Likewise.
2720 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2722 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2723 interaction with -pie.
2725 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2727 * selftest.h: Fix alphabetization of per-source-file selftest
2730 2018-04-30 Jason Merrill <jason@redhat.com>
2732 PR c++/61982 - dead stores to destroyed objects.
2733 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2736 2018-04-30 Jason Merrill <jason@redhat.com>
2738 * tree.c (build_clobber): New.
2739 * tree.h: Declare it.
2740 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2742 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2744 * diagnostic-show-locus.c (layout::layout): Update for
2745 location_get_source_line returning a char_span.
2746 (struct char_span): Move to input.h.
2747 (struct correction): Update for fields in char_span becoming
2749 (struct source_line): Update for location_get_source_line
2750 returning a char_span.
2751 (layout::print_line): Likewise.
2752 * edit-context.c (edited_file::print_content): Likewise.
2753 (edited_file::print_diff_hunk): Likewise.
2754 (edited_file::print_run_of_changed_lines): Likewise.
2755 (edited_file::get_num_lines): Likewise.
2756 (edited_line::edited_line): Likewise.
2757 * final.c (asm_show_source): Likewise.
2758 * input.c (location_get_source_line): Convert return type
2759 from const char * to char_span, losing the final "line_len"
2761 (dump_location_info): Update for the above.
2762 (get_substring_ranges_for_loc): Likewise. Use a char_span
2763 when handling the literal within the line.
2764 (test_reading_source_line): Update for location_get_source_line
2765 returning a char_span.
2766 * input.h (class char_span): Move here from
2767 diagnostic-show-locus.c, converting from a struct to a class.
2768 Make data members private.
2769 (char_span::operator bool): New.
2770 (char_span::length): New.
2771 (char_span::get_buffer): New.
2772 (char_span::operator[]): New.
2773 (char_span::subspan): Make const.
2774 (char_span::xstrdup): New.
2775 (location_get_source_line): Convert return type from const char *
2776 to char_span, losing the final "line_size" param.
2778 2018-04-30 Jan Hubicka <jh@suse.cz>
2780 * lto-wrapper.c (ltrans_priorities): New static var.
2781 (cmp_priority): New.
2782 (run_gcc): Read priorities and if doing parallel build order
2783 the Makefile by them.
2785 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2787 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2789 2018-04-30 Richard Biener <rguenther@suse.de>
2791 * tree-cfg.c (verify_address): Remove base argument, add
2792 flag whether to check TREE_ADDRESSABLE and do that.
2793 (verify_expr): Remove.
2794 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2795 (verify_gimple_assign_single): Likewise.
2796 (verify_gimple_switch): Likewise.
2797 (verify_expr_location_1): Dereference tp once. Add (disabled)
2798 piece from verify_expr.
2799 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2801 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2803 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2805 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2807 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2808 (small_data_pattern): Likewise.
2809 (arc_rewrite_small_data): Likewise.
2810 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2811 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2812 (get_symbol_alignment): New function.
2813 (legitimate_small_data_address_p): Likewise.
2814 (legitimate_scaled_address): Update, call
2815 legitimate_small_data_address_p.
2816 (output_sdata): New static variable.
2817 (arc_print_operand): Update how we handle small data operands.
2818 (arc_print_operand_address): Likewise.
2819 (arc_legitimate_address_p): Update, use
2820 legitimate_small_data_address_p.
2821 (arc_rewrite_small_data_p): Remove.
2822 (arc_rewrite_small_data_1): Likewise.
2823 (arc_rewrite_small_data): Likewise.
2824 (small_data_pattern): Likewise.
2825 (compact_sda_memory_operand): Update to use
2826 legitimate_small_data_address_p and get_symbol_alignment.
2827 (prepare_move_operands): Don't rewite sdata pattern.
2828 (prepare_extend_operands): Remove.
2829 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2831 (zero_extendqisi2): Likewise.
2832 (zero_extendhisi2): Likewise.
2833 (extendqihi2): Likewise.
2834 (extendqisi2): Likewise.
2835 (extendhisi2): Likewise.
2839 * config/arc/constraints.md (Usd): Change it to memory constraint.
2841 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2843 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2844 as source of std instructions.
2845 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2846 allow 6-bit constants as source for store instructions.
2847 (movdi_insn): Update instruction pattern to allow 6-bit constants
2848 as source for store instructions.
2850 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2852 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2854 2018-04-30 Nathan Sidwell <nathan@acm.org>
2855 Sandra Loosemore <sandra@codesourcery.com>
2857 * dumpfile.c (dump_open): Allow '-' for stdout.
2858 * doc/invoke.texi (Developer Options): Document dump filename
2859 determination early. Document stdin/stdout selection.
2861 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2863 Microblaze Target: PIC data text relative
2865 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2866 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2868 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2869 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2870 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2871 New addressing mode for data-text relative position indepenedent code.
2872 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2873 'ADDRESS_SYMBOLIC_TXT_REL'.
2874 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2875 (microblaze_legitimate_pic_operand): Exclude function calls from
2876 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2877 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2879 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2880 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2881 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2882 for 'address + offset'.
2883 (microblaze_expand_prologue): Add new function prologue call for
2885 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2886 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2887 table in case of TARGET_PIC_DATA_TEXT_REL.
2888 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2889 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2890 Add new macros 'UNSPEC_TEXT',
2891 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2892 + exclude function calls from 'UNSPEC_PLT' in case of data text
2894 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2895 new target hook for generating address diff vector tables in case of
2897 * doc/tm.texi : Regenerate.
2898 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2899 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2900 of addr diff vector generation.
2901 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2902 target hook definition.
2903 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2904 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2905 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2906 option for data text relative.
2908 2018-04-30 Richard Biener <rguenther@suse.de>
2910 * tree-chrec.h (evolution_function_is_constant_p): Remove
2912 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2914 2018-04-30 Richard Biener <rguenther@suse.de>
2917 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2919 2018-04-30 Richard Biener <rguenther@suse.de>
2921 PR tree-optimization/28364
2922 PR tree-optimization/85275
2923 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2924 copying first exit test.
2926 2018-04-28 Mark Wielaard <mark@klomp.org>
2928 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2930 (dwarf_AT): Handle DW_AT_addr_base.
2931 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2933 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2936 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2937 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2938 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2939 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2941 2018-04-28 Richard Biener <rguenther@suse.de>
2943 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2944 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2945 to reflect use. Only add interesting stmts.
2947 2018-04-27 Martin Jambor <mjambor@suse.cz>
2950 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2951 the jump function allows for passing through aggregate values.
2953 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2955 * input.h (in_system_header_at): Convert from macro to inline
2957 (from_macro_expansion_at): Likewise.
2958 (from_macro_definition_at): Likewise.
2960 2018-04-27 Jeff Law <law@redhat.com>
2962 * config.gcc: Mark tile* targets as deprecated/obsolete.
2964 2018-04-27 Richard Biener <rguenther@suse.de>
2966 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2969 2018-04-27 Richard Biener <rguenther@suse.de>
2971 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2973 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2975 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2976 with Yd constraint. Set "preferred_for_speed" attribute from
2977 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2979 (*movdi_internal): Ditto.
2980 (movti_interunit splitters): Remove
2981 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2982 (movdi_interunit splitters): Ditto.
2983 * config/i386/constraints.md (Ye): Remove.
2984 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2986 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2989 * config/aarch64/constraints.md (Usg): Limit to 31.
2992 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2994 PR tree-optimization/85529
2995 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2996 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2997 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2998 zero extension or masking of the MSB bit.
2999 (optimize_range_tests): Add FIRST_BB argument, pass it through
3000 to optimize_range_tests_var_bound.
3001 (maybe_optimize_range_tests, reassociate_bb): Adjust
3002 optimize_range_tests callers.
3004 2018-04-26 Richard Biener <rguenther@suse.de>
3005 Jakub Jelinek <jakub@redhat.com>
3007 * cgraph.h (symbol_table): Just declare debug method here.
3008 * symtab.c (symbol_table::debug): Define.
3010 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3012 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3014 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3016 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3017 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3018 (*movdi_internal): Substitute Yi and Yj constraint with x
3019 and Ym and Yn constraint with y constraint. Update "isa"
3020 attribute and set "preferred_for_speed" attribute from
3021 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3022 (*movsi_internal): Ditto.
3023 (*movdf_internal): Ditto.
3024 (*movsf_internal): Ditto.
3025 (*zero_extendsidi2): Ditto.
3026 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3027 (sse2_loadld): Ditto.
3028 (*vec_extract<ssevecmodelower>_0): Ditto.
3029 (*vec_extractv4si_0_zext_sse4): Ditto.
3030 (vec_concatv2di): Ditto.
3031 (*vec_dup<mode>): Ditto.
3032 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3033 * config/i386/constraints.md (Yi): Remove.
3038 2018-04-26 Nathan Sidwell <nathan@acm.org>
3040 * dumpfile.c (dump_open): New.
3041 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3042 (dump_finish): Detect stdio/stderr by value not name.
3044 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3046 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3048 2018-04-26 Tom de Vries <tom@codesourcery.com>
3051 * config/nvptx/nvptx.c (verify_neutering_jumps)
3052 (verify_neutering_labels): New function
3053 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3055 2018-04-26 Tom de Vries <tom@codesourcery.com>
3058 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3059 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3062 2018-04-26 Richard Biener <rguenther@suse.de>
3063 Tom de Vries <tom@codesourcery.com>
3066 * lto-streamer-out.c (output_function): Fixup loops if required to match
3067 discovery done in the reader.
3069 2018-04-26 Richard Biener <rguenther@suse.de>
3071 PR tree-optimization/85116
3072 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3073 have a loop exit from the single latch predecessor. Remove
3074 case of header with just condition.
3075 (ch_base::copy_headers): Exclude infinite loops from any
3077 (pass_ch::execute): Record exits.
3079 2018-04-26 Richard Biener <rguenther@suse.de>
3081 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3082 prologue cost vector and pass it to vect_get_load_cost.
3083 (vect_get_peeling_costs_all_drs): Likewise.
3084 (vect_peeling_hash_get_lowest_cost): Likewise.
3085 (vect_enhance_data_refs_alignment): Likewise.
3087 2018-04-26 Richard Biener <rguenther@suse.de>
3090 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3091 checking of integer<->pointer conversions.
3092 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3093 sign-/zero-extending pointer types.
3094 (expand_omp_for_static_chunk): Likewise.
3096 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3097 Jean Lee <xiaoyur347@gmail.com>
3099 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3100 (TARGET_ASAN_SHADOW_OFFSET): Define.
3101 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3102 true for -fsanitize=address.
3104 2018-04-25 Mark Wielaard <mark@klomp.org>
3106 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3109 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3111 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3112 than "alu", remove explicit "memory" and "imm_disp" attributes.
3113 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3116 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3117 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3118 gen_lowpart_no_emit.
3120 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3123 * config/i386/i386.c (ix86_expand_builtin): Change memory
3124 operand to XI, extend p0 to Pmode.
3125 * config/i386/i386.md: Change unspec volatile and operand
3126 1 mode to XI, change operand 0 mode to P.
3128 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3130 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3131 GET_MODE_MASK before any checking.
3132 (nds32_can_use_bset_p): Likewise.
3133 (nds32_can_use_btgl_p): Likewise.
3135 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3137 * config/nds32/nds32-doubleword.md: New define_split pattern for
3138 illegal register number.
3140 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3142 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3144 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3146 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3148 2018-04-25 Richard Biener <rguenther@suse.de>
3150 * lto-streamer.h (LTO_major_version): Bump to 8.
3152 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3154 * BASE-VER: Set to 9.0.0.
3156 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3158 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3159 in __abskf2 and __powikf2.
3161 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3164 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3165 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3166 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3167 Use the above on operand 2. Reindent.
3168 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3170 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3173 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3175 * config/i386/i386.opt (mcet): Removed.
3176 * doc/install.texi: Remove -mcet documentation.
3177 * doc/invoke.texi: Likewise.
3179 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3182 * doc/install.texi: Remove -mcet from bootstrap-cet.
3184 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3187 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3188 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3192 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3193 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3194 containing a CONST_VECTOR.
3196 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3198 * doc/install.texi: Update newlib dependency for nvptx.
3200 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3203 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3204 instead of INTVAL when shifting x left.
3206 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3208 PR tree-optimization/85478
3209 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3210 vect_grouped_store_supported for single element vectors.
3212 2018-04-24 Richard Biener <rguenther@suse.de>
3215 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3216 load cost increase to the case of non-constant step.
3218 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3221 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3222 destination if any_malformed_asm.
3224 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3227 * expr.c (store_field): In the bitfield case, if the value comes from
3228 a function call and is returned in registers by means of a PARALLEL,
3229 do not change the mode of the temporary unless BLKmode and VOIDmode.
3231 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3233 PR rtl-optimization/85423
3234 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3235 dependencies to debug insns when the previous insn is non-debug.
3237 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3239 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3240 enums into a single definition.
3241 (fls): Fix predicates and printing.
3244 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3246 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3247 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3248 and short u6 immediate.
3249 (check_if_valid_sleep_operand): Remove.
3250 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3252 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3254 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3255 flag_always_save_lp condition.
3256 * config/nds32/nds32.opt (malways-save-lp): New option.
3258 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3260 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3261 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3262 * config/nds32/nds32.h
3263 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3264 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3266 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3268 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3269 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3271 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3272 Chung-Ju Wu <jasonwucj@gmail.com>
3274 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3276 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3277 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3279 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3281 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3283 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3285 * config/nds32/nds32-protos.h (nds32_data_alignment,
3286 nds32_local_alignment): Declare.
3287 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3288 nds32_local_alignment): New functions.
3289 (TARGET_CONSTANT_ALIGNMENT): Define.
3290 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3292 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3294 * config/nds32/nds32.c
3295 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3296 (TARGET_MODES_TIEABLE_P): Likewise.
3298 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3300 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3303 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3304 Chung-Ju Wu <jasonwucj@gmail.com>
3306 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3307 for unaligned access.
3308 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3310 * config/nds32/nds32-intrinsic.md: Likewise.
3311 * config/nds32/nds32_intrinsic.h: Likewise.
3312 * config/nds32/nds32.h (nds32_builtins): Likewise.
3313 * config/nds32/nds32.opt (munaligned-access): New option.
3314 * config/nds32/nds32.c (nds32_asm_file_start): Display
3315 flag_unaligned_access status.
3317 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3319 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3320 -mno-relax is present.
3321 * config/riscv/linux.h (LINK_SPEC): Ditto.
3323 2018-04-20 Martin Sebor <msebor@redhat.com>
3326 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3328 (gimple_fold_builtin_stxcpy_chk): Same.
3329 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3331 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3334 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3335 __powikf2 when long double is IEEE 128-bit.
3337 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3339 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3340 step to make sure stack always aligned.
3342 2018-04-20 Carl Love <cel@us.ibm.com>
3345 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3346 size check for arg0.
3348 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3349 Tom de Vries <tom@codesourcery.com>
3352 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3353 Emit insns for calls too.
3354 (nvptx_find_par): Always look for worker-level predecessor insn.
3355 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3357 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3358 (nvptx_process_pars): Propagate frames for calls.
3360 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3363 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3365 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3366 (ix86_handle_option): Don't handle OPT_mibt.
3367 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3369 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3371 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3372 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3373 (ix86_target_macros): Define __CET__ with flag_cf_protection
3374 for -fcf-protection.
3375 * config/i386/i386.c (isa2_opts): Remove -mibt.
3376 * config/i386/i386.h (TARGET_IBT): Removed.
3377 (TARGET_IBT_P): Likewise.
3378 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3379 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3380 * config/i386/i386.opt (mcet): Update help message.
3383 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3384 -mcet as an alias for -mshstk.
3386 2018-04-20 Richard Biener <rguenther@suse.de>
3389 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3390 complexity by forcing a single use of the multiply operand.
3392 2018-04-20 Martin Jambor <mjambor@suse.cz>
3395 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3396 recursion dependency to only apply to non-clones.
3398 2018-04-20 Martin Jambor <mjambor@suse.cz>
3401 * ipa-cp.c (create_specialized_node): Check that clones of
3402 self-recursive edges exist during IPA-CP.
3404 2018-04-19 Toon Moene <toon@moene.org>
3406 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3409 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3411 PR tree-optimization/85467
3412 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3413 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3414 VECTOR_CST element to type.
3416 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3419 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3420 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3421 (builtin_longjmp): Likewise.
3422 (save_stack_nonlocal): New pattern.
3423 (restore_stack_nonlocal): Likewise.
3425 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3428 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3429 Replace ASM_OUTPUT_LABEL with fprintf.
3431 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3434 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3435 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3436 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3437 define __IBT__ and __SHSTK__ for -fcf-protection.
3438 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3440 (ix86_trampoline_init): Likewise.
3441 (x86_output_mi_thunk): Likewise.
3442 (ix86_notrack_prefixed_insn_p): Likewise.
3443 (ix86_option_override_internal): Don't disallow -fcf-protection.
3444 * config/i386/i386.md (rdssp<mode>): Also enable for
3446 (incssp<mode>): Likewise.
3447 (nop_endbr): Likewise.
3448 * config/i386/i386.opt (mcet): Change help message to built-in
3452 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3453 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3454 enable CET built-in functions.
3456 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3458 * common/config/i386/i386-common.c
3459 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3460 OPTION_MASK_ISA_MOVDIRI_UNSET,
3461 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3462 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3463 * config.gcc (movdirintrin.h): New header.
3464 * config/i386/cpuid.h (bit_MOVDIRI,
3465 bit_MOVDIR64B): New bits.
3466 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3468 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3469 (VOID, PVOID, PCVOID)): New function types.
3470 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3471 __builtin_ia32_directstoreu_u64,
3472 __builtin_ia32_movdir64b): New builtins.
3473 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3474 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3476 (ix86_valid_target_attribute_inner_p): Ditto.
3477 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3478 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3479 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3480 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3481 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3482 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3483 (movdiri<mode>, movdir64b_<mode>): New.
3484 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3485 * config/i386/immintrin.h: Include movdirintrin.h.
3486 * config/i386/movdirintrin.h: New file.
3487 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3489 2018-04-19 Richard Biener <rguenther@suse.de>
3492 * cfg.c (clear_bb_flags): When loop state says we have
3493 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3495 2018-04-19 Richard Biener <rguenther@suse.de>
3497 PR tree-optimization/84737
3498 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3499 copying restrict info.
3500 (vect_setup_realignment): Use it.
3501 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3502 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3503 the first DR to all generated stores.
3504 (vectorizable_load): Likewise for loads.
3506 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3508 PR tree-optimization/85446
3509 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3510 the integral and pointer types to have the same precision.
3512 * doc/install.texi: Document --disable-cet being the default and
3515 2018-04-18 Martin Liska <mliska@suse.cz>
3517 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3520 2018-04-18 Martin Liska <mliska@suse.cz>
3523 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3526 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3527 arguments if they are comparable.
3529 2018-04-18 Martin Liska <mliska@suse.cz>
3532 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3535 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3538 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3541 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3542 ENDBR after calling __morestack.
3544 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3547 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3548 by using gcc_base_ver to generate a gcc_driver_version, and use
3549 it when generating GCC_DRIVER_NAME.
3550 * configure: Regenerate.
3552 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3555 * config.gcc: Obsolete powerpc*-*-*spe*.
3557 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3560 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3561 (stabstr_D): Change type of unum from unsigned int to
3562 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3565 2018-04-17 Jim Wilson <jimw@sifive.com>
3568 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3569 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3570 Set arg_pointer_offset after using pretend_args_size.
3572 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3574 PR rtl-optimization/85431
3575 * dse.c (record_store): Ignore zero width stores.
3578 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3579 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3580 __builtin_stack_restore rather than after it.
3581 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3582 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3583 argument instead of virtual_dynamic_stack_rtx.
3585 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3587 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3589 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3590 Add note to error message to explain internal mapping of overloaded
3591 built-in function name to non-overloaded built-in function name.
3592 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3595 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3598 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3599 where the inputs overlap with the output.
3601 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3603 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3604 (=v, v) alternative and explicit "memory" attribute.
3605 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3606 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3608 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3609 "sselog1" type instead of "sselog".
3610 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3611 "sselog". Remove explicit "memory" attribute.
3612 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3613 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3615 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3616 "sselog1" type instead of "sselog". Remove explicit "memory"
3618 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3619 use "sselog1" type instead of "sselog". Remove explicit "memory"
3621 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3622 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3624 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3625 "sselog1" type instead of "sselog". Remove explicit "memory"
3627 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3628 use "sselog1" type instead of "sselog". Remove explicit "memory"
3632 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3635 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3638 2018-04-17 Martin Jambor <mjambor@suse.cz>
3641 * ipa-cp.c (create_specialized_node): Call
3642 expand_all_artificial_thunks if necessary.
3644 2018-04-17 Martin Liska <mliska@suse.cz>
3647 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3648 in message, remote space in between '_G' and '('.
3650 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3653 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3654 avx512f_vmcmp<mode>3<round_saeonly_name>,
3655 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3656 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3657 avx512f_rndscale<mode><round_saeonly_name>,
3658 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3659 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3660 Use %<iptr>2 instead of %2 for -masm=intel.
3661 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3662 avx512f_vcvttss2usi<round_saeonly_name>,
3663 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3665 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3666 avx512f_vcvttsd2usi<round_saeonly_name>,
3667 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3668 Use %q1 instead of %1 for -masm=intel.
3669 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3670 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3671 of %3 for -masm=intel.
3672 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3674 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3676 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3678 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3679 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3681 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3683 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3684 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3685 %g1 and one with %0 and %1.
3686 (avx512er_vmrcp28<mode><round_saeonly_name>,
3687 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3689 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3690 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3691 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3692 of %0 and %{%4%} for -masm=intel.
3693 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3694 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3695 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3696 order of %0 and %{%5%}%{z%} for -masm=intel.
3698 2018-04-17 Jan Hubicka <jh@suse.cz>
3701 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3703 2018-04-17 Martin Liska <mliska@suse.cz>
3706 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3707 for target_clone error message. Make default implementation
3708 clone to be a local declaration.
3709 (separate_attrs): Add new argument and check for an empty
3711 (expand_target_clones): Handle it.
3712 (ipa_target_clone): Make redirection just for target_clones
3715 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3716 Tom de Vries <tom@codesourcery.com>
3719 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3720 tiled basic blocks without omp continue statements.
3722 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3725 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3726 vec_extract expression as having side effects to make sure it gets
3729 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3732 * config/i386/i386.c (get_builtin_code_for_version): Check
3735 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3738 * gcc/config.gcc: Support "skylake".
3739 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3741 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3742 (processor_target_table): Add "skylake".
3743 (ix86_option_override_internal): Add "skylake".
3744 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3745 PROCESSOR_CANNONLAKE.
3746 (get_builtin_code_for_version): Fix priority for
3747 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3748 PROCESSOR_SKYLAKE-AVX512.
3749 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3750 (processor_type): Add PROCESSOR_SKYLAKE.
3752 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3753 Jason Merrill <jason@redhat.com>
3756 * convert.c (convert_to_integer_1): Use direct recursion for
3757 enumeral types and types with a precision less than the number
3758 of bits in their mode.
3760 2018-04-16 Julia Koval <julia.koval@intel.com>
3763 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3764 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3766 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3769 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3770 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3771 and -mno-direct-move.
3773 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3776 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3777 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3778 Ensure negative shifts result in {0}.
3780 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3782 PR rtl-optimization/79916
3783 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3784 regs (if any) to define how to gnerate SD moves when LRA is in
3787 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3789 PR rtl-optimization/85393
3790 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3791 * except.c (expand_dw2_landing_pad_for_region): Make static.
3792 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3793 a label and unconditional jump to old_bb, rather than
3794 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3797 PR rtl-optimization/85376
3798 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3799 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3800 instead of a specific value.
3802 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3803 Bin Cheng <bin.cheng@arm.com>
3805 PR tree-optimization/82965
3806 PR tree-optimization/83991
3807 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3808 by_profile_only parameter.
3809 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3810 information if the loop was predicted to iterate too many times.
3811 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3813 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3816 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3819 2018-04-13 Martin Liska <mliska@suse.cz>
3820 Jakub Jelinek <jakub@redhat.com>
3823 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3824 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3825 * builtins.c (expand_builtin_memory_copy_args): Use
3826 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3827 handle dest_addr == pc_rtx.
3829 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3832 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3833 asked to not generate direct moves.
3834 (fix_trunc<mode>si2_stfiwx): Similar.
3835 (fix_trunc<mode>si2_internal): Similar.
3837 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3840 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3841 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3842 lookup if dest in some wider mode is known to be const0_rtx and
3843 if so, record permanent equivalence for it to be ZERO_EXTEND of
3844 the narrower mode destination.
3846 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3848 * lto-streamer-out.c (output_function): Revert 259346.
3849 * omp-expand.c (expand_oacc_for): Likewise.
3851 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3853 PR rtl-optimization/85354
3854 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3855 * sel-sched.c (sel_global_init): ... here.
3857 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3860 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3861 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3862 mode for PE-COFF targets.
3863 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3864 (i386_pe_asm_lto_end): Likewise.
3865 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3866 (TARGET_ASM_LTO_END): Likewise.
3867 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3868 (i386_pe_asm_lto_start): New function.
3869 (i386_pe_asm_lto_end): Likewise.
3871 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3872 Richard Biener <rguenther@suse.de>
3875 * lto-streamer-out.c (output_function): Fix CFG loop state before
3877 * omp-expand.c (expand_oacc_for): Handle calls to internal
3878 functions like regular functions.
3880 2018-04-12 Richard Biener <rguenther@suse.de>
3883 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3884 for the early LTO debug to properly generate references to it
3885 during DIE emission. Do not re-use that for the skeleton for
3887 (dwarf2out_early_finish): Likewise.
3889 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3892 * config/i386/sse.md
3893 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3894 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3895 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3896 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3897 and output is a reg, avoid creating invalid lowpart subreg, but
3898 instead split into a 512-bit move. Don't split if not AVX512VL,
3899 input is xmm16+ reg and output is a mem.
3900 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3901 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3902 xmm16+ reg and output is a mem.
3904 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3906 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3907 also for flag_dwarf2_cfi_asm.
3909 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3911 PR rtl-optimization/85342
3912 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3913 a bool scalar var inside of the loop instead. Don't try to update
3914 recog_data.operand after failed apply_change_group.
3916 2018-04-12 Tom de Vries <tom@codesourcery.com>
3919 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3920 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3921 array with flexible array member as array without given dimension.
3922 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3923 argument for undefined param to true.
3925 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3928 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3929 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3930 from PowerPC section.
3931 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3932 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3933 help text that is too long.
3934 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3935 help text that is too long.
3936 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3937 help text that is too long.
3939 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3941 * config/alpha/alpha.md (stack_probe_internal): Rename
3942 from "probe_stack". Update all callers.
3944 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3946 PR rtl-optimization/84566
3947 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3948 sched_macro_fuse_insns.
3950 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3953 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3954 (compute_block_dependences): ... from here.
3956 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3958 PR tree-optimization/85331
3959 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3960 from int to HOST_WIDE_INT.
3962 2018-04-11 Martin Jambor <mjambor@suse.cz>
3965 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3966 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3967 not the same as the source val.
3968 (cgraph_edge_brings_value_p): New parameter.
3969 (gather_edges_for_value): Pass destination value to
3970 cgraph_edge_brings_value_p.
3971 (perhaps_add_new_callers): Likewise.
3972 (get_info_about_necessary_edges): Likewise and exclude values brought
3973 only by self-recursive edges.
3974 (create_specialized_node): Redirect only clones of self-calling edges.
3975 (+self_recursive_pass_through_p): New function.
3976 (find_more_scalar_values_for_callers_subset): Use it.
3977 (find_aggregate_values_for_callers_subset): Likewise.
3978 (known_aggs_to_agg_replacement_list): Removed.
3979 (decide_whether_version_node): Re-calculate known constants for all
3980 remaining context clones.
3982 2018-04-11 Richard Biener <rguenther@suse.de>
3985 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3986 from early DWARF output.
3987 (dwarf2out_early_finish): Output line info unconditionally into
3988 early DWARF and add reference to it.
3990 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3993 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3994 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3995 other than V2DFmode using iptr mode attribute.
3996 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3998 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4000 PR rtl-optimization/84659
4001 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4003 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4006 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4008 (output_loc_list): Pass address of a dummy size variable even in the
4009 locview handling loop.
4010 (index_location_lists): Add comment on why skip_loc_list_entry can't
4013 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4016 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4019 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4022 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4023 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4024 and -mstring-compare-inline-limit.
4026 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4029 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4030 for stack clash protection in a register whenever we need it to be in
4033 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4035 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4036 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4038 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4041 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4043 (mlong-double-): Ditto.
4044 * config/rs6000/sysv4.opt (msdata=): Ditto.
4045 (mtls-size=): Ditto.
4047 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4049 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4050 erroneous entries for
4051 "vector int vec_ldl (int, long int *)", and
4052 "vector unsigned int vec_ldl (int, unsigned long int *)".
4053 Add comments and entries for
4054 "vector bool char vec_ldl (int, bool char *)",
4055 "vector bool short vec_ldl (int, bool short *)",
4056 "vector bool int vec_ldl (int, bool int *)",
4057 "vector bool long long vec_ldl (int, bool long long *)",
4058 "vector pixel vec_ldl (int, pixel *)",
4059 "vector long long vec_ldl (int, long long *)",
4060 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4061 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4062 type tree bool_long_long_type_node and correct definition of
4063 bool_V2DI_type_node to make reference to this new type tree.
4064 (rs6000_mangle_type): Replace erroneous reference to
4065 bool_long_type_node with bool_long_long_type_node.
4066 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4067 comments to emphasize sign distinctions for char and int types and
4068 replace RS6000_BTI_bool_long constant with
4069 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4070 use of RS6000_BTI_pixel.
4071 (bool_long_type_node): Remove this macro definition.
4072 (bool_long_long_type_node): New macro definition
4074 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4076 PR rtl-optimization/85300
4077 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4078 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4079 simplify_unary_operation fails.
4081 2018-04-10 Martin Liska <mliska@suse.cz>
4083 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4084 cgraph_edge and ipa_ref.
4086 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4090 * config/i386/sse.md
4091 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4092 computation of the VEC_MERGE selector from mask.
4093 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4094 Fix decoding of the VEC_MERGE selector into mask.
4096 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4098 PR tree-optimization/85286
4099 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4101 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4103 * final.c (final_1): Set insn_last_address as well as
4104 insn_current_address.
4106 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4109 * explow.c (emit_stack_probe): Call validize_mem on memory location
4110 before passing it to gen_probe_stack. Create address operand and
4111 legitimize it for the probe_stack_address case.
4113 2018-04-09 Jan Hubicka <jh@suse.cz>
4116 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4117 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4118 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4120 (free_lang_data): Rebuild type inheritance graph.
4122 2018-04-09 Martin Sebor <msebor@redhat.com>
4124 * invoke.texi (-finline-small-functions): Mention other optimization
4126 (-findirect-inlining, -fpartial-inlining): Same.
4127 (-finline-functions-called-once): Same.
4128 (-freorder-blocks-and-partition): Same.
4130 2018-04-09 Jan Hubicka <jh@suse.cz>
4133 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4134 jumps; choose last target that matches the criteria (i.e.
4135 no partition changes for non-crossing jumps).
4136 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4137 support for redirecting crossing jumps to non-crossing.
4139 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4141 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4142 also for naked functions.
4144 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4146 * config/arc/arc.md (add_shift): New pattern.
4147 (add_shift2): Likewise.
4148 (sub_shift): Likewise.
4149 (sub_shift_cmp0_noout): Likewise.
4150 (compare_si_ashiftsi): Likewise.
4151 (xbfu_cmp0_noout): New combine pattern.
4152 (xbfu_cmp0"): Likewise.
4153 (movsi_set_cc_insn): Place the predicable variant first.
4154 (commutative_binary_cmp0_noout): Remove clobber.
4155 (commutative_binary_cmp0): New pattern.
4156 (noncommutative_binary_cmp0): Likewise.
4157 (noncommutative_binary_cmp0_noout): Likewise.
4158 (noncommutative_binary_comparison_result_used): Removed.
4159 (rsub_cmp0): New pattern.
4160 (rsub_cmp0_noout): Likewise.
4161 (extzvsi): Changed, keep only meaningful variants.
4162 (SQH, SEZ): New iterators.
4163 (SQH_postfix): New mode attribute.
4164 (SEZ_prefix): New code attribute.
4165 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4166 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4167 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4169 (noncommutative_operator): Check the availability of barrel
4172 2018-04-09 Richard Biener <rguenther@suse.de>
4174 PR tree-optimization/85284
4175 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4176 Only use the niter constraining form of simple_iv when the exit
4179 2018-04-09 Tom de Vries <tom@codesourcery.com>
4182 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4183 (define_expand "*memory_barrier"): New define_expand.
4184 (define_insn "memory_barrier"): New insn.
4186 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4188 PR rtl-optimization/80463
4189 PR rtl-optimization/83972
4190 PR rtl-optimization/83480
4192 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4193 correct producer for the insn.
4194 (tidy_control_flow): Fixup seqnos in case of debug insns.
4196 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4198 PR rtl-optimization/83913
4200 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4201 different sched-times when merging exprs.
4203 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4205 PR rtl-optimization/83962
4207 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4208 tidy_fallthru_edge and tidy_control_flow.
4210 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4212 PR rtl-optimization/83530
4214 * sel-sched.c (force_next_insn): New global variable.
4215 (remove_insn_for_debug): When force_next_insn is true, also leave only
4216 next insn in the ready list.
4217 (sel_sched_region): When the region wasn't scheduled, make another pass
4218 over it with force_next_insn set to 1.
4220 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4222 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4224 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4225 for interrupt control.
4226 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4227 functions for interrupt control.
4228 * config/nds32/nds32-intrinsic.md: Likewise.
4229 * config/nds32/nds32_intrinsic.h: Likewise.
4230 * config/nds32/nds32.h (nds32_builtins): Likewise.
4232 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4234 * config/nds32/nds32.c (nds32_init_machine_status,
4235 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4236 strict_aligned_p field.
4237 (nds32_expand_to_rtl_hook): New function.
4238 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4239 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4241 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4242 Chung-Ju Wu <jasonwucj@gmail.com>
4244 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4245 * config/nds32/nds32-n7.md: New file.
4246 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4247 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4249 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4250 * config/nds32/nds32.md (pipeline_model): Add n7.
4251 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4252 * config/nds32/pipelines.md: Include n7 settings.
4254 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4255 Chung-Ju Wu <jasonwucj@gmail.com>
4257 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4258 * config/nds32/nds32-e8.md: New file.
4259 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4260 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4262 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4263 * config/nds32/nds32.md (pipeline_model): Add e8.
4264 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4265 * config/nds32/pipelines.md: Include e8 settings.
4267 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4268 Chung-Ju Wu <jasonwucj@gmail.com>
4270 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4271 * config/nds32/nds32-n8.md: New file.
4272 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4273 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4275 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4276 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4277 * config/nds32/nds32.md (pipeline_model): Add n8.
4278 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4279 * config/nds32/pipelines.md: Include n8 settings.
4281 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4282 Chung-Ju Wu <jasonwucj@gmail.com>
4284 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4285 * config/nds32/nds32-n9-2r1w.md: New file.
4286 * config/nds32/nds32-n9-3r2w.md: New file.
4287 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4288 nds32_register_ports): New or modify for cpu n9.
4289 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4291 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4292 * config/nds32/nds32-utils.c: New file.
4293 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4294 TARGET_MUL_SLOW): Define.
4295 * config/nds32/nds32.md (pipeline_model): New attribute.
4296 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4297 New options that support cpu n9.
4298 * config/nds32/pipelines.md: Include n9 settings.
4299 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4301 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4303 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4304 information if necessary.
4305 (output_cond_branch_compare_zero): Likewise.
4306 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4307 (nds32_target_alignment): Refine for alignment.
4308 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4309 (FUNCTION_BOUNDARY): Modify.
4310 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4312 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4314 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4316 * config/nds32/constants.md (unspec_volatile_element): Add values for
4317 TLB operation and data prefetch.
4318 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4319 functions for TLB operation and data prefetch.
4320 * config/nds32/nds32-intrinsic.md: Likewise.
4321 * config/nds32/nds32_intrinsic.h: Likewise.
4322 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4323 (nds32_print_operand): Likewise.
4324 * config/nds32/nds32.h (nds32_builtins): Likewise.
4326 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4327 Andrew Pinski <pinsika@gcc.gnu.org>
4330 * match.pd: Use constant_boolean_node of correct type instead of
4331 boolean_true_node or boolean_false_node for simplifying
4332 pointer comparisons to zero.
4334 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4336 PR tree-optimization/80021
4337 * tree.c (verify_type_variant): Make error call in verify_variant_match
4338 translatable and remove final full stop.
4340 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4342 * config/nds32/constants.md (unspec_volatile_element): Add
4343 UNSPEC_VOLATILE_EH_RETURN.
4344 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4345 nds32_output_stack_pop): Support dwarf exception handling process.
4346 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4347 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4348 exception handling process.
4349 (nds32_compute_stack_frame): Likewise.
4350 (nds32_return_addr_rtx): Likewise.
4351 (nds32_initial_elimination_offset): Likewise.
4352 (nds32_expand_prologue): Likewise.
4353 (nds32_expand_epilogue): Likewise.
4354 (nds32_dynamic_chain_address): New function.
4355 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4357 (DYNAMIC_CHAIN_ADDRESS): Define.
4358 (EH_RETURN_DATA_REGNO): Define.
4359 (EH_RETURN_STACKADJ_RTX): Define.
4360 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4361 patterns for dwarf exception handling.
4363 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4365 * config/nds32/nds32.h: Clean up obsolete macros.
4367 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4369 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4370 Add enum values for particular instructions.
4371 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4372 particular intrinsic functions.
4373 * config/nds32/nds32-intrinsic.md: Likewise.
4374 * config/nds32/nds32_intrinsic.h: Likewise.
4375 * config/nds32/nds32.h (nds32_builtins): Likewise.
4376 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4377 (btst, ave): New patterns for particular instructions.
4379 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4381 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4382 Add enum values for atomic load/store and memory sync.
4383 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4385 * config/nds32/nds32-intrinsic.md: Likewise.
4386 * config/nds32/nds32_intrinsic.h: Likewise.
4387 * config/nds32/nds32.h (nds32_builtins): Likewise.
4389 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4391 PR tree-optimization/85257
4392 * fold-const.c (native_encode_vector): If not all elts could fit
4393 and off is -1, return 0 rather than offset.
4394 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4395 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4396 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4397 adjust buffer in native_interpret_expr call.
4399 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4401 * config/nds32/constants.md (unspec_volatile_element): Add cache
4402 control enum values.
4403 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4404 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4405 * config/nds32/nds32.c (nds32_cctl_names): New.
4406 (nds32_print_operand): Handle cache control register names.
4407 * config/nds32/nds32.h (nds32_builtins): New enum values.
4408 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4410 * config/nds32/nds32.md (type): Add mmu.
4411 * config/nds32/pipelines.md (simple_insn): Add mmu.
4413 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4415 * config/nds32/nds32.md (type): Remove call.
4416 * config/nds32/pipelines.md (simple_insn): Likewise.
4418 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4420 * config/nds32/constants.md (unspec_volatile_element): Add
4421 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4422 UNSPEC_VOLATILE_FMFCFG.
4423 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4424 description for fmfcfg and fmfcsr.
4425 (bdesc_1arg): Add fmtcsr.
4426 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4427 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4428 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4429 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4430 unspec_fmfcfg): New patterns.
4431 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4432 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4433 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4434 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4435 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4436 __nds32__fmfcfg): Define.
4438 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4440 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4441 intrinsic register names.
4442 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4443 intrinsic register enum values and macros.
4445 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4447 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4448 for load/store addressing form.
4449 (nds32_print_operand_address): Likewise.
4451 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4454 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4455 based on LABEL_REF. Remove useless assertion.
4456 (pic_address_needs_scratch): Fix formatting.
4457 (sparc_legitimize_pic_address): Minor tweaks.
4458 (sparc_delegitimize_address): Adjust assertion accordingly.
4459 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4460 into symbolic_operand.
4461 (movsi_high_pic_label_ref): Likewise.
4462 (movsi_lo_sum_pic_label_ref): Likewise.
4463 (movdi_pic_label_ref): Likewise.
4464 (movdi_high_pic_label_ref): Likewise.
4465 (movdi_lo_sum_pic_label_ref): Likewise.
4467 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4469 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4470 custom LIB_SPEC setup.
4472 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4473 Kito Cheng <kito.cheng@gmail.com>
4475 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4476 * config/riscv/freebsd.h: New.
4478 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4480 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4481 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4484 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4485 Kito Cheng <kito.cheng@gmail.com>
4487 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4488 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4489 * config/nds32/nds32-protos.h (nds32_output_call,
4490 nds32_output_return): Declare.
4491 * config/nds32/nds32.md: Refine all the call and return patterns.
4493 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4496 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4497 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4499 PR rtl-optimization/84872
4500 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4501 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4504 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4506 * expr.c (copy_blkmode_to_reg): Revert 254862.
4507 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4509 2018-04-06 Richard Biener <rguenther@suse.de>
4512 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4513 after seeing a component reference with an adjacent field. Treat
4514 refs to arrays at struct end of external decls similar to
4515 refs to unconstrained commons.
4517 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4520 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4521 look through SAVE_EXPRs with non-side-effects argument. Adjust
4523 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4526 2018-04-06 Richard Biener <rguenther@suse.de>
4529 * alias.c (find_base_term): New wrapper around find_base_term
4530 unwinding CSELIB_VAL_PTR changes.
4531 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4534 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4536 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4538 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4539 constant definitions.
4540 ("nop"): lr 0,0 -> nopr r0
4541 ("nop_lr0", "nop_lr1"): New insn definitions.
4543 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4545 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4546 NDS32_V3PUSH_AVAILABLE_P macro.
4548 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4549 Chung-Ju Wu <jasonwucj@gmail.com>
4551 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4552 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4553 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4554 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4555 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4556 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4557 * config/nds32/constraints.md: New constraints and checking for hard
4558 float configuration.
4559 * config/nds32/iterators.md: New mode iterator and attribute for hard
4560 float configuration.
4561 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4563 * config/nds32/nds32-fpu.md: New file.
4564 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4565 deal with hard float code generation.
4566 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4568 (abi_type, float_reg_number): New enum type.
4569 * config/nds32/nds32-predicates.c: New predicates for hard float.
4570 * config/nds32/nds32-protos.h: Declare functions for hard float.
4571 * config/nds32/nds32.c: Implementation for hard float configuration.
4572 * config/nds32/nds32.h: Definitions for hard float configuration.
4573 * config/nds32/nds32.md: Include hard float machine description and
4574 modify patterns for hard float configuration.
4575 * config/nds32/nds32.opt: New options for hard float configuration.
4576 * config/nds32/predicates.md: New predicates for hard float
4579 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4581 * common/config/nds32/nds32-common.c
4582 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4584 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4587 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4588 CONSTRUCTOR_ELT (ctor, ...)->value.
4590 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4593 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4595 2018-04-05 Tom de Vries <tom@codesourcery.com>
4598 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4601 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4602 Kito Cheng <kito.cheng@gmail.com>
4604 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4605 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4606 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4607 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4609 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4610 Kito Cheng <kito.cheng@gmail.com>
4612 * config/nds32/constraints.md (Ufe): New memory constraint.
4613 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4614 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4615 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4617 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4618 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4620 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4622 * config/nds32/nds32.md: Use optimize_size in the condition for
4623 alu-shift instructions.
4625 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4627 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4629 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4631 * config/nds32/nds32.md (negsi2): Refine pattern.
4633 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4634 Chung-Ju Wu <jasonwucj@gmail.com>
4636 * config/nds32/iterators.md (shift_rotate): New code iterator.
4637 (shift): New code attribute.
4638 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4639 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4640 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4641 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4642 bit-wise operations.
4643 (andsi3, *andsi3): Ditto.
4644 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4645 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4646 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4647 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4648 nds32_ior_operand, nds32_xor_operand): New predicates.
4650 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4652 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4653 (addsi3, subsi3): ... this.
4655 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4657 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4659 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4661 * config/nds32/nds32.md: Adjust indention.
4663 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4665 * config/nds32/nds32.md (feature): New attribute.
4667 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4669 * config/nds32/nds32.md (subtype): New attribute.
4671 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4674 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4675 expansion to perform a bitwise AND of the argument followed by a
4676 boolean negation of the result.
4678 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4680 PR rtl-optimization/84878
4681 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4682 the basic block. Assert the use reference is not artificial and that
4683 it has an associated insn.
4685 2018-04-04 Michael Matz <matz@suse.de>
4687 * builtins.c (compute_objsize): Pass correct operand
4688 to array_at_struct_end_p.
4690 2018-04-04 Richard Biener <rguenther@suse.de>
4693 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4694 from contexts for DINFO_LEVEL_TERSE and below.
4696 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4698 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4699 resiter_operand condition.
4700 * config/nds32/nds32.md (*move<mode>): Ditto.
4702 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4703 Monk Chiang <sh.chiang04@gmail.com>
4705 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4707 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4709 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4711 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4712 Kito Cheng <kito.cheng@gmail.com>
4714 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4715 nds32_cond_code_str, output_cond_branch,
4716 output_cond_branch_compare_zero, nds32_expand_cbranch,
4717 nds32_expand_cstore, nds32_expand_movcc,
4718 nds32_output_cbranchsi4_equality_zero,
4719 nds32_output_cbranchsi4_equality_reg,
4720 nds32_output_cbranchsi4_equality_reg_or_const_int,
4721 nds32_output_cbranchsi4_greater_less_zero: New functions.
4722 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4723 nds32_expand_cstore, nds32_expand_movcc,
4724 nds32_output_cbranchsi4_equality_zero,
4725 nds32_output_cbranchsi4_equality_reg,
4726 nds32_output_cbranchsi4_equality_reg_or_const_int,
4727 nds32_output_cbranchsi4_greater_less_zero): Declare.
4728 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4729 nds32_rimm11s_operand): New predicates.
4730 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4731 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4734 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4736 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4737 * config/nds32/nds32.md: Ditto.
4738 * config/nds32/pipelines.md: Ditto.
4740 2018-04-04 Richard Biener <rguenther@suse.de>
4742 PR tree-optimization/85168
4743 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4744 propagating abnormals.
4746 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4748 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4750 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4751 Kito Cheng <kito.cheng@gmail.com>
4753 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4754 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4755 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4756 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4757 * config/nds32/nds32.md (sibcall_internal): New.
4758 (sibcall_register): Remove.
4759 (sibcall_immediate): Remove.
4760 (sibcall_value_internal): New.
4761 (sibcall_value_register): Remove.
4762 (sibcall_value_immediate): Remove.
4763 * config/nds32/predicates.md (nds32_general_register_operand): New.
4764 (nds32_call_address_operand): New.
4766 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4768 PR rtl-optimization/85167
4769 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4770 bb_defs if *split_p, instead preinitialize it to NULL.
4772 PR tree-optimization/85156
4773 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4774 evaluating the argument multiple times.
4776 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4778 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4780 (_mm_cvtpd_ps): Likewise.
4781 (_mm_cvttpd_epi32): Likewise.
4782 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4783 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4784 vector, pixel, and bool following altivec.h include.
4786 2018-04-03 Martin Sebor <msebor@redhat.com>
4788 * doc/extend.texi (Common Function Attributes): Clarify.
4789 (const attribute): Likewise.
4790 (pure attribute): Likewise.
4792 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4795 * config/i386/i386.c (ix86_expand_vector_set): Use
4796 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4798 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4800 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4801 instructions when changing rounding bits to preserve precision bits
4802 in the x87 control word.
4804 2018-04-03 Martin Liska <mliska@suse.cz>
4806 PR tree-optimization/82491
4807 * rtl.h (strip_offset_and_add): Replace += suboffset with
4808 poly_uint64 () + suboffset.
4810 2018-03-29 Martin Liska <mliska@suse.cz>
4811 Martin Jambor <mjambor@suse.cz>
4814 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4815 param_type is not an integral or pointer type.
4817 2018-04-03 Richard Biener <rguenther@suse.de>
4819 * sese.h (recompute_all_dominators): Remove.
4821 2018-04-02 Martin Sebor <msebor@redhat.com>
4823 * doc/invoke.texi (-Wrestrict): Fix typos.
4825 2018-04-02 Jim Wilson <jimw@sifive.com>
4827 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4828 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4829 (<optab>di3, <optab>si3_extend): Likewise.
4830 (<optab>si3_mask, <optab>si3_mask_1): New.
4831 (<optab>di3_mask, <optab>di3_mask_1): New.
4832 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4833 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4834 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4836 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4838 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4841 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4843 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4844 (nds32_canonicalize_comparison): New function.
4846 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4847 Kito Cheng <kito.cheng@gmail.com>
4848 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4850 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4851 * config/nds32/constants.md (unspec_volatile_element): Add
4852 UNSPEC_VOLATILE_RELAX_GROUP.
4853 * config/nds32/nds32-relax-opt.c: New file.
4854 * config/nds32/nds32-predicates.c
4855 (nds32_symbol_load_store_p): New function.
4856 * config/nds32/nds32-protos.h
4857 (nds32_symbol_load_store_p): Declare function.
4858 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4859 * config/nds32/nds32.c
4860 (nds32_register_pass): New function to register pass.
4861 (nds32_register_passes): New function to register passes.
4862 * config/nds32/nds32.md (relax_group): New pattern.
4863 * config/nds32/nds32.opt (mrelax-hint): New option.
4864 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4866 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4868 * config/nds32/t-nds32: Modify files dependency.
4870 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4872 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4873 (PROFILE_HOOK): Define its implementation.
4875 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4877 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4878 type and 32-bit size.
4880 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4883 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4884 (V_128_256): New mode iterator.
4885 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4886 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4887 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4889 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4890 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4892 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4895 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4896 NaN inputs correctly.
4898 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4901 * config/rs6000/vsx.md (??r): New mode attribute.
4902 (*vsx_mov<mode>_64bit): Use it.
4903 (*vsx_mov<mode>_32bit): Likewise.
4905 2018-03-30 Martin Sebor <msebor@redhat.com>
4907 PR tree-optimization/84818
4908 * builtins.c (check_access): Use warning_n.
4910 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4913 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4915 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4918 2018-03-30 Julia Koval <julia.koval@intel.com>
4921 * x86-tune.def (movx, partial_reg_dependency): Enable for
4924 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4927 * lra-constraints.c (process_alt_operands): Move setting
4928 this_alternative_matches below.
4930 2018-03-29 Martin Liska <mliska@suse.cz>
4933 * doc/invoke.texi: Document how LTO works with debug info.
4934 Describe auto-load support of binutils. Mention 'x86-64'
4935 as valid option value of -march option.
4937 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4939 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4942 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4943 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4944 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4947 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4950 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4951 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4952 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4953 for RS6000_BTM_POWERPC64.
4954 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4955 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4956 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4961 2018-03-28 Carl Love <cel@us.ibm.com>
4964 2017-09-27 Carl Love <cel@us.ibm.com>
4966 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4967 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4968 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4971 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4973 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4974 instead of __vector bool.
4975 (_mm_max_pu8): Likewise.
4976 (_mm_min_pi16): Likewise.
4978 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4981 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4982 (DIVWEUO): Likewise.
4984 (DIVDEUO): Likewise.
4985 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4986 DIVWEUO and DIVDEUO.
4987 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4988 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4989 (div_extend): Likewise.
4990 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4992 (__builtin_divweuo): Likewise.
4993 (__builtin_divdeo): Likewise.
4994 (__builtin_divdeuo): Likewise.
4996 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4999 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5000 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5002 PR tree-optimization/82004
5003 * gimple-match-head.c (optimize_pow_to_exp): New function.
5004 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5005 Don't fold to exp if optimize_pow_to_exp is false.
5007 2018-03-28 Martin Liska <mliska@suse.cz>
5010 * calls.c (initialize_argument_information): Fix trailing space.
5011 * common.opt: Fix typo and provide better explanation for
5012 -fsanitize-coverage option.
5013 * config/i386/i386.opt: Fix typo.
5015 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5016 Martin Liska <mliska@suse.cz>
5019 * gimplify.c (asan_poison_variable): Don't do the check for
5020 gimplify_omp_ctxp here.
5021 (gimplify_decl_expr): Do it here.
5022 (gimplify_target_expr): Likewise.
5024 2018-03-28 Martin Liska <mliska@suse.cz>
5027 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5028 chkp_type_bounds_count if MPX is not enabled.
5030 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5032 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5034 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5037 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5038 function to create the function decl for complex long double
5039 multiply and divide for -mabi=ieeelongdouble.
5040 (init_float128_ieee): Call it.
5042 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5045 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5046 -fcf-protection=branch -mibt.
5047 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5049 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5052 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5054 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5057 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5058 extern array declarations.
5060 2018-03-27 Richard Biener <rguenther@suse.de>
5063 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5064 explicit single_use checks.
5066 2018-03-27 Richard Biener <rguenther@suse.de>
5068 PR tree-optimization/85082
5069 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5072 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5074 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5075 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5076 Turn on fasynchronous-unwind-tables and funwind-tables.
5078 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5081 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5082 (*bmi_blsr_<mode>_ccz): Ditto.
5084 2018-03-26 Tom de Vries <tom@codesourcery.com>
5086 PR tree-optimization/85063
5087 * omp-general.c (offloading_function_p): New function. Factor out
5089 * omp-offload.c (pass_omp_target_link::gate): ... here.
5090 * omp-general.h (offloading_function_p): Declare.
5091 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5092 with attribute omp declare target for offloading functions.
5094 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5096 PR tree-optimization/84005
5097 * tree-data-ref.h (get_base_for_alignment): Declare.
5098 * tree-data-ref.c (get_base_for_alignment_1): New function.
5099 (get_base_for_alignment): Likewise.
5100 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5101 get_base_for_alignment to find a suitable base object, instead
5102 of always using drb->base_address.
5104 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5107 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5108 known size by default.
5110 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5113 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5114 and non BLKmode operands.
5116 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5119 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5120 Clean up attributes.
5122 2018-03-23 Richard Biener <rguenther@suse.de>
5125 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5126 we are going to emit early debug for LTO.
5128 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5131 * function.c (match_asm_constraints_1): Don't optimize if input
5132 doesn't satisfy general_operand predicate for output's mode.
5135 * alias.c (write_dependence_p): Don't require for x_canonicalized
5136 non-VOIDmode if x has VOIDmode.
5139 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5140 just don't try to optimize it rather than assert it never happens.
5142 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5144 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5145 macro expansions for definition of ST_INTERNAL_<mode> and
5146 LD_INTERNAL_<mode> builtins.
5147 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5149 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5151 (altivec_expand_st_builtin): Likewise.
5152 (altivec_expand_builtin): Remove calls to deleted functions.
5153 (rs6000_address_for_altivec): Delete this function.
5154 * config/rs6000/vector.md: Remove expands for
5155 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5157 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5160 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5161 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5162 re-computing once computed.
5163 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5164 (arm_init_machine_status): Initialize
5165 machine->static_chain_stack_bytes.
5167 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5170 * doc/extend.texi: Add four new prototypes for vec_ld.
5171 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5172 definitions for more logical presentation.
5173 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5174 entries for V1TI variants of __builtin_altivec_ld builtin.
5175 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5176 handling of V1TI variant of LVX icode pattern.
5177 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5178 (rs6000_gimple_fold_builtin): Likewise.
5179 (altivec_init_builtins): Add code to define
5180 __builtin_altivec_lvx_v1ti function.
5182 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5185 * function.c (match_asm_constraints_1): Don't do the optimization
5186 if input isn't a REG, SUBREG, MEM or constant.
5188 2018-03-22 Tom de Vries <tom@codesourcery.com>
5190 PR tree-optimization/84956
5191 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5192 bb_has_abnormal_pred.
5194 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5197 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5198 DECL_INITIAL (decl) to decl at the end.
5199 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5202 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5204 * doc/extend.texi (__builtin_tgmath): Document when complex
5205 integer types are treated as _Complex _Float64.
5207 2018-03-21 Tom de Vries <tom@codesourcery.com>
5209 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5211 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5213 PR tree-optimization/84960
5214 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5215 if it is ENTRY block, move them into single succ of ENTRY in that case.
5217 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5219 PR tree-optimization/84811
5220 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5221 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5222 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5223 (known_subrange_p): Update accordingly. Cast each value involved
5224 in the size comparison, rather than casting the result of the
5227 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5229 PR tree-optimization/84982
5230 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5231 by flipping the least significant bit rather than all bits from
5232 bitpos to bitpos + bitsize - 1.
5234 2018-03-21 Nathan Sidwell <nathan@acm.org>
5236 * doc/extend.texi (Deprecated Features): Remove mention of
5237 long-deleted deprecations.
5239 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5242 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5243 * configure: Regenerate.
5245 2018-03-21 Tom de Vries <tom@codesourcery.com>
5247 PR tree-optimization/83126
5248 * tree-parloops.c (num_phis): New function.
5249 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5251 2018-03-21 Nathan Sidwell <nathan@acm.org>
5253 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5254 mention anon-struct/union members and trailing attributes.
5256 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5258 PR tree-optimization/84969
5259 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5260 builtin memset partitions if they set different rhs values.
5262 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5264 PR rtl-optimization/84989
5265 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5266 VEC_DUPLICATE with scalar result mode.
5268 2018-03-21 Martin Liska <mliska@suse.cz>
5271 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5272 not intended return statement.
5274 2018-03-21 Martin Liska <mliska@suse.cz>
5277 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5278 (chkp_find_bound_slots_1): Limit number of iterations.
5280 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5283 * Minor grammar fixes for x86 options.
5285 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5288 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5289 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5291 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5294 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5295 (altivec_lvx_<mode>_1op): Likewise.
5296 (altivec_stvx_<mode>_2op): Likewise.
5297 (altivec_stvx_<mode>_1op): Likewise.
5298 (altivec_lvx_<VM2:mode>): New define_expand.
5299 (altivec_stvx_<VM2:mode>): Likewise.
5300 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5301 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5302 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5303 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5304 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5305 (rs6000_gen_lvx): Likewise.
5306 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5307 (altivec_expand_stv_builtin): Likewise.
5308 (altivec_expand_builtin): Likewise.
5309 * config/rs6000/vector.md: Likewise.
5311 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5314 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5317 2018-03-20 Richard Biener <rguenther@suse.de>
5320 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5321 sign-conversions as zero, fall back to standard scalar_stmt
5324 2018-03-20 Martin Liska <mliska@suse.cz>
5327 * predict.c (rebuild_frequencies): Handle case when we have
5328 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5330 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5333 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5334 flag_section_anchors.
5335 * varasm.c (use_blocks_for_decl_p): Remove hack for
5336 dw2_force_const_mem.
5339 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5341 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5342 be created, use lowpart_subreg of operands[0] rather than operands[0]
5344 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5345 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5346 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5347 and n constraint instead of aarch64_shift_imm_di and Usd.
5348 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5349 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5351 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5354 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5355 to favor GPR over NEON registers.
5356 (<shift>di3_neon): Likewise.
5358 2018-03-20 Tom de Vries <tom@codesourcery.com>
5361 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5362 (nvptx_process_pars): Emit bar.sync asap and alap.
5364 2018-03-20 Tom de Vries <tom@codesourcery.com>
5367 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5368 seen_label if seen_label is already set.
5370 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5373 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5374 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5375 Use 1U instead of 1. Formatting fixes.
5378 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5379 instead of TREE_TYPE (s1) for the return value.
5381 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5383 PR tree-optimization/84946
5384 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5385 bitsize + bitsize in poly_uint64 rather than poly_int64.
5388 * dwarf2asm.c: Include fold-const.c.
5389 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5390 of decl rather than decl itself.
5392 PR rtl-optimization/84643
5393 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5395 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5398 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5399 calling assemble_variable.
5401 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5404 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5405 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5407 2018-03-19 Jim Wilson <jimw@sifive.com>
5410 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5411 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5412 (riscv_first_stack_step): Likewise.
5413 (riscv_option_override): Use STACK_BOUNDARY instead of
5415 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5417 (BIGGEST_ALIGNMENT): Set to 128.
5418 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5419 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5422 2018-03-19 Richard Biener <rguenther@suse.de>
5424 PR tree-optimization/84933
5425 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5426 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5428 2018-03-19 Richard Biener <rguenther@suse.de>
5430 PR tree-optimization/84859
5431 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5432 (cond_if_else_store_replacement): Perform sinking operation on
5433 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5434 Generalize what a BB with a single eligible store is.
5436 2018-03-19 Richard Biener <rguenther@suse.de>
5438 PR tree-optimization/84929
5439 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5440 chrec_is_positive against non-chrec arg.
5442 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5445 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5447 2018-03-18 Martin Liska <mliska@suse.cz>
5449 PR rtl-optimization/84635
5450 * regrename.c (build_def_use): Use matches_mode only when
5453 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5455 PR tree-optimization/84913
5456 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5457 vectorize chains of COND_EXPRs.
5459 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5461 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5463 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5465 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5467 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5469 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5471 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5472 Kito Cheng <kito.cheng@gmail.com>
5474 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5475 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5476 (nds32_adjust_reg_alloc_order): New function.
5477 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5479 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5481 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5482 nds32_print_operand, nds32_print_operand_address): Use
5483 HOST_WIDE_INT_PRINT_DEC instead.
5485 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5487 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5489 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5492 * config/i386/i386.c (initial_ix86_tune_features,
5493 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5495 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5496 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5497 rather than 1u << ix86_tune. Formatting fix.
5498 (ix86_option_override_internal): Change ix86_arch_mask from
5499 unsigned int to unsigned HOST_WIDE_INT, initialize to
5500 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5501 (ix86_function_specific_restore): Likewise.
5503 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5506 * postreload.c (reload_combine_recognize_pattern): Perform
5507 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5508 truncate_int_for_mode the result for the destination's mode.
5511 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5512 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5514 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5517 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5520 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5521 regno_allocno_class_array and sorted_pseudos.
5522 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5523 insns where regno is used.
5525 2018-03-16 Martin Liska <mliska@suse.cz>
5528 * multiple_target.c (create_dispatcher_calls): Redirect
5529 reference in the symbol table.
5531 2018-03-16 Martin Liska <mliska@suse.cz>
5534 * multiple_target.c (create_dispatcher_calls): Redirect also
5537 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5541 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5542 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5543 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5545 2018-03-16 Julia Koval <julia.koval@intel.com>
5547 * doc/invoke.texi (Skylake Server): Add CLWB.
5548 Cannonlake): Remove CLWB.
5550 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5552 PR tree-optimization/84841
5553 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5555 (FLOAT_ONE_CONST_TYPE): Define.
5556 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5557 (sort_by_operand_rank): Put entries with higher constant_type last
5558 rather than first to match comments.
5560 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5562 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5565 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5568 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5569 check and use address of target always.
5571 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5574 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5575 (indirect_thunk_bnd_needed): Likewise.
5576 (indirect_thunks_used): Likewise.
5577 (indirect_thunks_bnd_used): Likewise.
5578 (indirect_return_needed): New.
5579 (indirect_return_bnd_needed): Likewise.
5580 (output_indirect_thunk_function): Add a bool argument for
5582 (output_indirect_thunk_function): Don't generate alias for
5583 function return thunk.
5584 (ix86_code_end): Call output_indirect_thunk_function to generate
5585 function return thunks.
5586 (ix86_output_function_return): Set indirect_return_bnd_needed
5587 and indirect_return_needed instead of indirect_thunk_bnd_needed
5588 and indirect_thunk_needed.
5590 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5592 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5593 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5594 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5596 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5597 Paul Hua <paul.hua.gm@gmail.com>
5600 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5602 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5604 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5605 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5608 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5611 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5612 instead of GET_MODE_SIZE when comparing Units.
5614 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5617 * varasm.c (hash_section): Return an unchangeble hash value
5618 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5619 Return !aarch64_can_use_per_function_literal_pools_p ().
5621 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5624 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5625 rather than address of cmode as last argument to prepare_cmp_insn.
5627 2018-03-15 Julia Koval <julia.koval@intel.com>
5629 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5630 F_AVX512VNNI, F_AVX512BITALG): New.
5632 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5635 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5636 insn for floating-point loads and stores.
5638 2018-03-14 Carl Love <cel@us.ibm.com>
5640 * config/rs6000/rs6000-c.c: Add macro definitions for
5641 ALTIVEC_BUILTIN_VEC_PERMXOR.
5642 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5643 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5644 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5645 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5647 * config/doc/extend.texi: Add prototypes for vec_permxor.
5649 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5652 * diagnostic-show-locus.c (class layout_point): Convert m_line
5653 from int to linenum_type.
5654 (line_span::comparator): Use linenum "compare" function when
5655 comparing line numbers.
5656 (test_line_span): New function.
5657 (layout_range::contains_point): Convert param "row" from int to
5659 (layout_range::intersects_line_p): Likewise.
5660 (layout::will_show_line_p): Likewise.
5661 (layout::print_source_line): Likewise.
5662 (layout::should_print_annotation_line_p): Likewise.
5663 (layout::print_annotation_line): Likewise.
5664 (layout::print_leading_fixits): Likewise.
5665 (layout::annotation_line_showed_range_p): Likewise.
5666 (struct line_corrections): Likewise for field m_row.
5667 (line_corrections::line_corrections): Likewise for param "row".
5668 (layout::print_trailing_fixits): Likewise.
5669 (layout::get_state_at_point): Likewise.
5670 (layout::get_x_bound_for_row): Likewise.
5671 (layout::print_line): Likewise.
5672 (diagnostic_show_locus): Likewise for locals "last_line" and
5674 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5675 * input.c (selftest::test_linenum_comparisons): New function.
5676 (selftest::input_c_tests): Call it.
5677 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5678 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5679 * selftest.h (ASSERT_GT): New macro.
5680 (ASSERT_GT_AT): New macro.
5681 (ASSERT_LT): New macro.
5682 (ASSERT_LT_AT): New macro.
5684 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5686 PR rtl-optimization/84780
5687 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5689 2018-03-14 Martin Liska <mliska@suse.cz>
5691 * tree.c (record_node_allocation_statistics): Use
5692 get_stats_node_kind.
5693 (get_stats_node_kind): New function extracted from
5694 record_node_allocation_statistics.
5695 (free_node): Use get_stats_node_kind.
5697 2018-03-14 Richard Biener <rguenther@suse.de>
5699 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5700 that the value-set of ANTIC_IN doesn't grow.
5703 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5705 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5706 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5708 2018-03-14 Julia Koval <julia.koval@intel.com>
5710 * config.gcc (icelake-client, icelake-server): New.
5712 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5713 (initial_ix86_arch_features): Ditto.
5714 (PTA_SKYLAKE): Add SGX.
5715 (PTA_ICELAKE): Remove.
5716 (PTA_ICELAKE_CLIENT): New.
5717 (PTA_ICELAKE_SERVER): New.
5718 (ix86_option_override_internal): Split up icelake on icelake client and
5720 (get_builtin_code_for_version): Ditto.
5721 (fold_builtin_cpu): Ditto.
5722 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5723 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5724 * config/i386/i386.h (processor_type): Ditto.
5725 * doc/invoke.texi: Ditto.
5727 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5730 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5731 INTEGER_CST offset, add it together with bitpos / 8 and
5732 sign extend based on POINTER_SIZE.
5736 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5739 * config/i386/constraints.md (Yc): New register constraint.
5740 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5741 Use Yc constraint for alternative 2 of operand 0. Remove
5742 preferred_for_speed attribute.
5744 2018-03-14 Richard Biener <rguenther@suse.de>
5746 PR tree-optimization/84830
5747 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5748 with the old one to avoid oscillations.
5750 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5753 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5755 (assign_by_spills): Return a flag of reload assignment failure.
5756 Do not process the reload assignment failures. Do not spill other
5757 reload pseudos if they has the same reg class. Update n if
5759 (lra_assign): Add a return arg. Set up from the result of
5760 assign_by_spills call.
5761 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5762 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5763 usage_insns if it is not NULL.
5764 (spill_hard_reg_in_range): New function.
5765 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5766 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5767 function prototypes.
5768 (lra_assign): Change prototype.
5769 * lra.c (lra): Add code to deal with fails by splitting hard reg
5772 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5774 * config/riscv/riscv.opt (mrelax): New option.
5775 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5776 "norelax" when riscv_mrelax is disabled.
5777 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5779 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5782 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5783 reassociation for int modes.
5785 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5787 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5788 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5790 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5791 * config/aarch64/aarch64-sve.md
5792 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5793 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5794 (*extend<mode><Vwide>2): Rename to...
5795 (aarch64_sve_extend<mode><Vwide>2): ...this.
5796 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5797 renaming the old pattern to...
5798 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5800 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5801 define_expand, renaming the old pattern to...
5802 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5803 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5804 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5805 account when deciding which SVE instruction the optab should use.
5806 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5808 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5810 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5811 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5812 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5813 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5814 (tlsdesc_small_<mode>): Turn a define_expand and use
5815 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5816 (tlsdesc_small_advsimd_<mode>): ...this.
5817 (tlsdesc_small_sve_<mode>): New pattern.
5819 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5821 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5822 (UNSPEC_UMUL_HIGHPART): New constants.
5823 (MUL_HIGHPART): New int iteraor.
5824 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5825 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5827 (*<su>mul<mode>3_highpart): New define_insn.
5829 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5832 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5835 2018-03-13 Martin Liska <mliska@suse.cz>
5838 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5840 (sem_item_optimizer::~sem_item_optimizer): Release it.
5841 (sem_item_optimizer::merge_classes): Register variable aliases.
5842 (sem_item_optimizer::fixup_pt_set): New function.
5843 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5844 * ipa-icf.h: Declare new variables and functions.
5846 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5849 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5850 integer_pow2p@2 and test integer_pow2p in condition.
5851 (A < 0 ? C : 0): Similarly for @1.
5854 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5855 characters starting at p contain '\0' character, don't look beyond
5859 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5860 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5863 * reg-stack.c (change_stack): Change update_end var from int to
5864 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5865 also call set_block_for_insn on the newly added insns and rescan.
5868 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5869 on the last operand.
5872 * tree.c (stabilize_reference_1): Return save_expr (e) for
5873 STATEMENT_LIST even if it doesn't have side-effects.
5875 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5877 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5879 2018-03-12 Renlin Li <renlin.li@arm.com>
5881 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5882 aarch64_output_scalar_simd_mov_immediate.
5884 2018-03-12 Martin Sebor <msebor@redhat.com>
5886 PR tree-optimization/83456
5887 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5888 for perfectly overlapping calls to memcpy.
5889 (gimple_fold_builtin_memory_chk): Same.
5890 (gimple_fold_builtin_strcpy): Handle no-warning.
5891 (gimple_fold_builtin_stxcpy_chk): Same.
5892 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5894 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5896 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5897 parameter. Use it for SFmode.
5898 (rs6000_function_arg_advance_1): Adjust.
5899 (rs6000_function_arg): Adjust.
5900 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5902 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5904 PR rtl-optimization/84169
5905 PR rtl-optimization/84780
5906 * combine.c (can_combine_p): Check for a 2-insn combination whether
5907 the destination register is used between the two insns, too.
5909 2018-03-12 Richard Biener <rguenther@suse.de>
5911 PR tree-optimization/84803
5912 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5913 for refs DR analysis didn't process.
5915 2018-03-12 Richard Biener <rguenther@suse.de>
5917 PR tree-optimization/84777
5918 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5919 force-vectorize loops ignore whether we are optimizing for size.
5921 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5923 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5924 (TARGET_MD_ASM_ADJUST): Define.
5926 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5927 Kito Cheng <kito.cheng@gmail.com>
5928 Chung-Ju Wu <jasonwucj@gmail.com>
5930 * config/nds32/nds32.c (nds32_compute_stack_frame,
5931 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5932 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5933 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5934 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5935 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5936 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5937 * config/nds32/nds32.md (prologue, epilogue): Use macro
5938 NDS32_V3PUSH_AVAILABLE_P to do checking.
5940 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5943 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5944 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5945 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5946 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5947 addition of most attributes on !orig_type_die or the attribute not
5948 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5950 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5951 Chung-Ju Wu <jasonwucj@gmail.com>
5953 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5955 * config/nds32/nds32.opt (mvh): New option.
5957 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5958 Chung-Ju Wu <jasonwucj@gmail.com>
5960 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5962 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5963 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5966 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5967 Chung-Ju Wu <jasonwucj@gmail.com>
5969 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5971 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5972 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5974 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5975 Kito Cheng <kito.cheng@gmail.com>
5976 Chung-Ju Wu <jasonwucj@gmail.com>
5978 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5979 UNSPEC_FFMISM and UNSPEC_FLMISM.
5980 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5981 for ffb, ffmism and flmism.
5982 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5983 (unspec_ffmism): Ditto.
5984 (unspec_flmism): Ditto.
5985 (nds32_expand_builtin_impl): Check if string extension is available.
5986 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5987 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5989 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5992 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5995 * lra-assigns.c (assign_by_spills): Return a flag of reload
5996 assignment failure. Do not process the reload assignment
5997 failures. Do not spill other reload pseudos if they has the same
5999 (lra_assign): Add a return arg. Set up from the result of
6000 assign_by_spills call.
6001 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6002 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6003 usage_insns if it is not NULL.
6004 (spill_hard_reg_in_range): New function.
6005 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6006 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6007 function prototypes.
6008 (lra_assign): Change prototype.
6009 * lra.c (lra): Add code to deal with fails by splitting hard reg
6012 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6015 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6017 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6020 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6021 (dw_val_node): Add val_symbolic_view.
6022 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6023 (symview_upper_bound): New.
6024 (new_line_info_table): Initialize symviews_since_reset.
6025 (dwarf2out_source_line): Count symviews_since_reset and set
6026 symview_upper_bound.
6027 (dw_val_equal_p): Handle symview.
6028 (add_AT_symview): New.
6029 (print_dw_val): Handle symview.
6030 (attr_checksum, attr_checksum_ordered): Likewise.
6031 (same_dw_val_p, size_of_die): Likewise.
6032 (value_format, output_die): Likewise.
6033 (add_high_low_attributes): Use add_AT_symview for entry_view.
6034 (dwarf2out_finish): Reset symview_upper_bound, clear
6037 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6040 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6041 Add strict argument and use it.
6042 (rs6000_split_multireg_move): Update for new strict argument.
6043 (mem_operand_gpr): Disallow all non-offsettable addresses.
6044 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6046 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6049 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6050 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6051 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6054 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6055 decl, use remap_type if we want to use the type.
6057 2018-03-09 Martin Sebor <msebor@redhat.com>
6059 PR tree-optimization/84526
6060 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6062 (builtin_access::generic_overlap): Be prepared to handle non-array
6065 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6067 PR rtl-optimization/84682
6068 * lra-constraints.c (process_address_1): Check is_address flag
6069 for address constraints.
6070 (process_alt_operands): Likewise.
6071 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6072 preprocess_constraints.
6073 * recog.h (preprocess_constraints): Add oploc parameter.
6075 * recog.c (preprocess_constraints): Test address_operand for
6076 CT_ADDRESS constraints.
6078 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6081 * lra-assigns.c (assign_by_spills): Return a flag of reload
6082 assignment failure. Do not process the reload assignment
6083 failures. Do not spill other reload pseudos if they has the same
6085 (lra_assign): Add a return arg. Set up from the result of
6086 assign_by_spills call.
6087 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6088 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6089 usage_insns if it is not NULL.
6090 (spill_hard_reg_in_range): New function.
6091 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6092 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6093 function prototypes.
6094 (lra_assign): Change prototype.
6095 * lra.c (lra): Add code to deal with fails by splitting hard reg
6098 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6101 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6102 Accept complain bool parameter. Only emit errors if it is true.
6103 (arm_parse_cpu_option_name): Likewise.
6104 (arm_target_thumb_only): Adjust callers of the above.
6105 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6106 prototype to take a default true bool parameter.
6107 (arm_parse_arch_option_name): Likewise.
6109 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6110 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6114 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6115 * configure: Regenerate.
6116 * configure.ac ("linker --version-script option"): New.
6117 ("linker soname option"): New.
6119 2018-03-09 Richard Biener <rguenther@suse.de>
6121 PR tree-optimization/84775
6122 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6123 immediate uses of predicate stmts and mark them modified.
6126 PR tree-optimization/84178
6127 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6129 (version_loop_for_if_conversion): Delay update_ssa call.
6130 (tree_if_conversion): Delay update_ssa until after predicate
6133 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6136 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6137 when the function accesses prior frames.
6139 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6142 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6143 gen_llsym, otherwise call maybe_gen_llsym.
6146 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6147 has ',' character inside of it.
6149 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6152 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6153 as clobbering CC_REGNUM.
6155 2018-03-08 Richard Biener <rguenther@suse.de>
6158 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6159 (follow_copies_to_constant): Do not follow SSA names registered
6162 2018-03-08 Richard Biener <rguenther@suse.de>
6164 PR tree-optimization/84178
6165 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6167 (version_loop_for_if_conversion): Delay update_ssa call.
6168 (tree_if_conversion): Delay update_ssa until after predicate
6171 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6173 PR tree-optimization/84178
6174 * tree-if-conv.c (release_bb_predicate): Remove the
6175 the assertion that the stmts have NULL use_ops.
6176 Discard the statements, asserting that they haven't
6177 yet been added to a BB.
6179 2018-03-08 Richard Biener <rguenther@suse.de>
6181 PR tree-optimization/84746
6182 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6183 (phi_translate): Pass in destination ANTIC_OUT set.
6184 (phi_translate_1): Likewise. For a simplified result lookup
6185 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6186 (phi_translate_set): Adjust.
6187 (do_pre_regular_insertion): Likewise.
6188 (do_pre_partial_partial_insertion): Likewise.
6190 2018-03-08 Martin Liska <mliska@suse.cz>
6192 PR gcov-profile/84735
6193 * doc/gcov.texi: Document usage of profile files.
6194 * gcov-io.h: Document changes in the format.
6196 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6200 * dwarf2out.c (struct dw_line_info_table): Update comments for
6202 (FORCE_RESET_NEXT_VIEW): New.
6203 (FORCE_RESETTING_VIEW_P): New.
6204 (RESETTING_VIEW_P): Check for -1 too.
6205 (ZERO_VIEW_P): Likewise.
6206 (new_line_info_table): Force-reset next view.
6207 (dwarf2out_begin_function): Likewise.
6208 (dwarf2out_source_line): Simplify zero_view_p initialization.
6209 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6210 view directly. Omit view when omitting .loc at line 0.
6212 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6214 PR tree-optimization/84740
6215 * tree-switch-conversion.c (process_switch): Call build_constructors
6216 only if info.phi_count is non-zero.
6218 PR tree-optimization/84739
6219 * tree-tailcall.c (find_tail_calls): Check call arguments against
6220 DECL_ARGUMENTS (current_function_decl) rather than
6221 DECL_ARGUMENTS (func) when checking for tail recursion.
6223 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6225 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6226 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6227 Volker Reichelt's entry and add entries for people that perform
6228 GCC fuzzy testing and report numerous bugs.
6230 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6233 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6234 readonly data in sdata, if that is disabled.
6235 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6236 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6237 -mreadonly-in-sdata option.
6239 2018-03-07 Martin Sebor <msebor@redhat.com>
6241 PR tree-optimization/84468
6242 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6243 basic block when looking for nul assignment.
6245 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6248 * except.h (output_function_exception_table): Adjust prototype.
6249 * except.c (output_function_exception_table): Remove FNNAME parameter
6250 and add SECTION parameter. Ouput one part of the table at a time.
6251 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6252 the first part of the exception table and emit unwind directives.
6253 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6254 (i386_pe_seh_cold_init): Likewise.
6255 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6256 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6257 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6258 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6259 * config/i386/winnt.c: Include except.h.
6260 (struct seh_frame_state): Add reg_offset, after_prologue and
6261 in_cold_section fields.
6262 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6263 (i386_pe_seh_cold_init): New function.
6264 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6265 to seh->in_cold_section.
6266 (seh_emit_push): Record the offset of the push.
6267 (seh_emit_save): Record the offet of the save.
6268 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6269 Test seh->after_prologue to disregard the epilogue.
6270 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6271 (i386_pe_end_cold_function): New function.
6273 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6276 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6277 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6280 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6282 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6283 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6286 * multiple_target.c: Include tree-inline.h and intl.h.
6287 (expand_target_clones): Diagnose and fail if node->definition and
6288 !tree_versionable_function_p (node->decl).
6290 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6292 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6294 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6295 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6296 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6298 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6301 * combine.c (try_combine): Use reg_or_subregno instead of handling
6302 just paradoxical SUBREGs and REGs.
6304 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6306 * config/arc/arc.c (arc_finalize_pic): Remove function.
6307 (arc_must_save_register): We use single base PIC register, remove
6308 checks to save/restore the PIC register.
6309 (arc_expand_prologue): Likewise.
6310 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6312 (arc_verify_short): Likewise.
6313 (arc_attr_type): Likewise.
6314 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6315 (walk_stores): Likewise.
6316 (arc_address_cost): Make it static.
6317 (arc_verify_short): Likewise.
6318 (branch_dest): Likewise.
6319 (arc_attr_type): Likewise.
6320 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6321 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6322 (arc_final_prescan_insn): Remove inserting the nops due to
6323 hardware hazards. It is done in reorg step.
6324 (insn_length_variant_t): Remove.
6325 (insn_length_parameters_t): Likewise.
6326 (arc_insn_length_parameters): Likewise.
6327 (arc_get_insn_variants): Likewise.
6328 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6330 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6333 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6336 PR tree-optimization/84687
6337 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6339 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6341 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6343 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6344 Rename to ppc_speculation_barrier.
6345 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6346 __builtin_ppc_speculation_barrier.
6348 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6351 * combine.c (combine_simplify_rtx): Don't try to simplify if
6352 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6355 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6357 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6358 to 32 bytes when compiling for POWER9.
6360 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6363 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6364 regparm >= 3 with no arg reg available also for calls with
6365 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6368 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6370 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6372 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6375 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6377 2018-03-05 Richard Biener <rguenther@suse.de>
6379 PR tree-optimization/84486
6380 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6381 When inserting a __builtin_assume_aligned call set the LHS
6382 SSA name alignment info accordingly.
6384 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6386 PR tree-optimization/84114
6387 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6388 Avoid reassociation of FLOAT_MODE addition.
6390 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6392 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6393 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6394 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6395 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6396 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6397 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6398 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6400 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6401 __builtin_ia32_wbinvd): New builtins.
6402 (SPECIAL_ARGS2): New.
6403 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6404 (SPECIAL_ARGS2): New.
6405 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6406 (ix86_valid_target_attribute_inner_p): Ditto.
6407 (ix86_init_mmx_sse_builtins): Add special_args2.
6408 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6409 TARGET_WBNOINVD_P): New.
6410 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6411 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6412 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6413 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6414 * config/i386/pconfigintrin.h: New file.
6415 * config/i386/wbnoinvdintrin.h: Ditto.
6416 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6417 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6419 2018-03-05 Richard Biener <rguenther@suse.de>
6421 PR tree-optimization/84670
6422 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6424 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6425 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6426 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6427 doesn't grow if all successors (recursively) were visited at least
6430 2018-03-05 Richard Biener <rguenther@suse.de>
6432 PR tree-optimization/84650
6433 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6434 if executed in the loop pipeline.
6436 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6438 * doc/configfiles.texi (Configuration Files): Move info about
6439 conditionalizing $target-protos.h to...
6440 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6441 differs from $target-protos.h.
6443 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6444 Chung-Ju Wu <jasonwucj@gmail.com>
6446 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6447 * config/nds32/nds32-multiple.md (setmemsi): Define.
6448 * config/nds32/nds32-memory-manipulation.c
6449 (nds32_gen_dup_4_byte_to_word_value): New.
6450 (emit_setmem_word_loop): New.
6451 (emit_setmem_byte_loop): New.
6452 (nds32_expand_setmem_loop): New.
6453 (nds32_expand_setmem_loop_v3m): New.
6454 (nds32_expand_setmem_unroll): New.
6455 (nds32_expand_setmem): New.
6457 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6458 Chung-Ju Wu <jasonwucj@gmail.com>
6460 * config/nds32/nds32-memory-manipulation.c
6461 (nds32_emit_load_store): New.
6462 (nds32_emit_post_inc_load_store): New.
6463 (nds32_emit_mem_move): New.
6464 (nds32_emit_mem_move_block): New.
6465 (nds32_expand_movmemsi_loop_unknown_size): New.
6466 (nds32_expand_movmemsi_loop_known_size): New.
6467 (nds32_expand_movmemsi_loop): New.
6468 (nds32_expand_movmemsi_unroll): New.
6469 (nds32_expand_movmemqi): Rename ...
6470 (nds32_expand_movmemsi): ... to this.
6471 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6472 (movmemsi): ... to this.
6473 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6474 (nds32_expand_movmemsi): ... to this.
6476 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6477 Monk Chiang <sh.chiang04@gmail.com>
6478 Chung-Ju Wu <jasonwucj@gmail.com>
6480 * config/nds32/nds32-protos.h
6481 (nds32_expand_load_multiple): New arguments.
6482 (nds32_expand_store_multiple): Ditto.
6483 (nds32_valid_multiple_load_store): Rename ...
6484 (nds32_valid_multiple_load_store_p): ... to this.
6485 * config/nds32/nds32-memory-manipulation.c
6486 (nds32_expand_load_multiple): Refine implementation.
6487 (nds32_expand_store_multiple): Ditto.
6488 * config/nds32/nds32-multiple.md
6489 (load_multiple): Update nds32_expand_load_multiple interface.
6490 (store_multiple): Update nds32_expand_store_multiple interface.
6491 * config/nds32/nds32-predicates.c
6492 (nds32_valid_multiple_load_store): Rename ...
6493 (nds32_valid_multiple_load_store_p): ... to this and refine
6495 * config/nds32/predicates.md
6496 (nds32_load_multiple_and_update_address_operation): New predicate.
6497 (nds32_store_multiple_and_update_address_operation): New predicate.
6499 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6500 Chung-Ju Wu <jasonwucj@gmail.com>
6502 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6503 (combo): New attribute.
6504 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6506 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6508 * config/nds32/nds32.opt: Change -mcmodel= default value.
6510 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6511 Monk Chiang <sh.chiang04@gmail.com>
6512 Chung-Ju Wu <jasonwucj@gmail.com>
6514 * config/nds32/constants.md (unspec_element): New enum.
6515 * config/nds32/constraints.md (Umw): New constraint.
6516 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6517 * config/nds32/nds32-intrinsic.md: Likewise.
6518 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6519 (nds32_valid_smw_lwm_base_p): New.
6520 (nds32_output_smw_single_word): New.
6521 (nds32_output_lmw_single_word): New.
6522 (nds32_expand_unaligned_load): New.
6523 (nds32_expand_unaligned_store): New.
6524 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6525 (nds32_output_smw_single_word): Declare.
6526 (nds32_output_lmw_single_word): Declare.
6527 (nds32_expand_unaligned_load): Declare.
6528 (nds32_expand_unaligned_store): Declare.
6529 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6530 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6531 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6532 NDS32_BUILTIN_UASTORE_DW.
6533 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6536 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6537 Kito Cheng <kito.cheng@gmail.com>
6538 Chung-Ju Wu <jasonwucj@gmail.com>
6540 * config/nds32/nds32-intrinsic.c
6541 (nds32_expand_builtin_null_ftype_reg): Delete.
6542 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6543 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6544 (nds32_read_argument): New.
6545 (nds32_legitimize_target): Ditto.
6546 (nds32_legitimize_argument): Ditto.
6547 (nds32_check_constant_argument): Ditto.
6548 (nds32_expand_unop_builtin): Ditto.
6549 (nds32_expand_unopimm_builtin): Ditto.
6550 (nds32_expand_binop_builtin): Ditto.
6551 (nds32_builtin_decl_impl): Ditto.
6552 (builtin_description): Ditto.
6553 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6554 (nds32_init_builtins_impl): Ditto.
6555 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6556 (nds32_builtin_decl): New.
6557 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6558 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6560 2018-03-02 Jeff Law <law@redhat.com>
6562 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6563 (redundant_insn, fill_simple_delay_slots): Likewise.
6564 (fill_slots_from_thread): Likewise.
6565 * resource.c (mark_referenced_resources): Likewise.
6566 (mark_set_resources, find_dead_or_set_registers): Likewise.
6568 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6570 * substring-locations.h (format_warning_va): Formatting fix for
6572 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6574 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6575 * substring-locations.c: Include intl.h.
6576 (format_warning_va): Turned into small wrapper around
6577 format_warning_n_va, renamed to ...
6578 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6579 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6581 (format_warning_at_substring_n): New function.
6582 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6583 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6584 format_warning_at_substring with just a shorter name instead of
6585 const function pointer.
6586 (fmtwarn_n): New function.
6587 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6588 appropriate, get rid of all the fmtstr temporaries, move conditionals
6589 with G_() wrapped string literals directly into fmtwarn arguments,
6590 cast dir.len to (int), formatting fixes.
6592 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6594 * doc/invoke.texi: Remove "Cilk Plus" references.
6596 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6597 Richard Biener <rguenther@suse.de>
6600 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6601 for error or warning attributes if CALL_FROM_THUNK_P is set.
6604 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6607 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6608 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6611 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6612 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6614 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6615 instead of -1U in last predictors element's probability member.
6617 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6620 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6621 arguments if they are comparable.
6623 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6625 PR tree-optimization/84634
6626 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6627 masks and masked_loop_p with a single loop_masks, making sure it's
6628 null for bb vectorization.
6630 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6632 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6633 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6634 loop->force_vectorize to check whether there is no alias.
6636 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6639 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6641 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6643 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6644 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6645 instead of a loop around prev_real_insn.
6646 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6650 * config/i386/i386.c (ix86_print_operand): Use conditional
6651 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6654 2018-03-02 Richard Biener <rguenther@suse.de>
6656 PR tree-optimization/84427
6657 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6658 (bitmap_set_subtract_values): Rewrite to handle multiple
6661 (prune_clobbered_mems): Likewise.
6662 (phi_translate): Take edge instead of pred/phiblock.
6663 (phi_translate_1): Likewise.
6664 (phi_translate_set): Likewise. Insert all translated
6665 exprs for a value into the set, keeping possibly multiple
6666 expressions per value.
6667 (compute_antic_aux): Adjust for phi_translate changes.
6668 When intersecting union the expressions and prune those
6669 not in the final value set, keeping possibly multiple
6670 expressions per value. Do not use value-insertion
6671 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6672 all expressions. Add verification that the value-sets
6673 only shrink during iteration.
6674 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6675 (do_pre_regular_insertion): Likewise.
6676 (do_pre_partial_partial_insertion): Likewise.
6678 2018-03-02 Richard Biener <rguenther@suse.de>
6681 * config/darwin.c (saved_debug_info_level): New static global.
6682 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6683 (darwin_asm_lto_end): Restore debug info generation settings.
6685 2018-03-01 Martin Liska <mliska@suse.cz>
6688 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6691 2018-03-01 Richard Biener <rguenther@suse.de>
6694 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6695 type annotation with LTO when debug was disabled at compile-time.
6697 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6699 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6701 (mips_final_postscan_insn): Likewise.
6703 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6705 PR rtl-optimization/84528
6706 * alias.c (init_alias_target): Add commentary.
6707 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6708 a unique base value if the frame pointer is not eliminated
6709 to the stack pointer.
6711 2018-03-01 Tom de Vries <tom@codesourcery.com>
6713 PR rtl-optimization/83327
6714 * lra-int.h (hard_regs_spilled_into): Declare.
6715 * lra.c (hard_regs_spilled_into): Define.
6716 (init_reg_info): Init hard_regs_spilled_into.
6717 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6718 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6719 (process_bb_lives): Handle hard_regs_spilled_into.
6720 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6721 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6723 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6725 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6726 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6727 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6728 * config/rs6000/aix72.h: New file.
6730 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6732 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6733 instead of warning_at with conditional singular and plural messages
6737 * stor-layout.c (update_alignment_for_field): For
6738 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6739 && !DECL_PACKED (field), do the alignment update, just use
6740 only desired_align instead of MAX (type_align, desired_align)
6742 (place_field): Don't do known_align < desired_align handling
6743 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6744 is non-NULL, instead do it after rli->prev_field handling and
6745 only if not within a bitfield word. For DECL_PACKED (field)
6746 use type_align of BITS_PER_UNIT.
6748 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6750 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6751 superfluous parentheses and trailing spaces.
6753 2018-02-28 Richard Biener <rguenther@suse.de>
6755 PR tree-optimization/84584
6756 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6757 SCoPs with fake exit edge.
6759 2018-02-28 Martin Liska <mliska@suse.cz>
6762 * timevar.c (timer::print): Fix format to properly print 100%
6765 2018-02-28 Richard Biener <rguenther@suse.de>
6768 * genmatch.c (capture_info::walk_match): Do not mark
6769 captured expressions without operands as expr_p given
6770 they act more like predicates and should be subject to
6771 "lost tail" side-effect preserving.
6773 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6775 PR rtl-optimization/81611
6776 * auto-inc-dec.c (attempt_change): Move dead note from
6777 mem_insn if it's the next use of regno
6778 (find_address): Take address use of reg holding
6779 non-incremented value. Add parm to limit search to the named
6781 (merge_in_block): Attempt to use a mem insn that is the next
6782 use of the original regno.
6784 2018-02-27 Martin Sebor <msebor@redhat.com>
6787 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6788 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6790 2018-02-27 Martin Sebor <msebor@redhat.com>
6792 PR translation/84207
6793 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6794 n argument to unsigned HOST_WIDE_INT.
6795 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6796 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6797 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6798 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6800 2018-02-27 Richard Biener <rguenther@suse.de>
6802 PR tree-optimization/84512
6803 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6804 Do not use the estimate returned from record_stmt_cost for
6805 the scalar iteration cost but sum properly using add_stmt_cost.
6807 2018-02-27 Richard Biener <rguenther@suse.de>
6809 PR tree-optimization/84466
6810 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6811 Adjust last change to less strictly validate use operands.
6813 2018-02-27 Martin Liska <mliska@suse.cz>
6815 PR gcov-profile/84548
6816 * gcov.c (process_file): Allow partial overlap and consider it
6817 also as group functions.
6818 (output_lines): Properly calculate range of lines for a group.
6820 2018-02-27 Martin Liska <mliska@suse.cz>
6822 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6823 'ggc' suffixes. Change first column width.
6824 (timer::print): Fix formatting of the column.
6826 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6828 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6829 preserve inline entry blocks for the sake of debug inline
6830 entry point markers alone.
6831 (remove_unused_locals): Suggest in comments a better place to
6832 force the preservation of inline entry blocks that are
6833 otherwise unused, but do not preserve them.
6835 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6837 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6839 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6842 * config/i386/constraints.md (Bs): Replace
6843 ix86_indirect_branch_register with
6844 TARGET_INDIRECT_BRANCH_REGISTER.
6846 * config/i386/i386.md (indirect_jump): Likewise.
6847 (tablejump): Likewise.
6848 (*sibcall_memory): Likewise.
6849 (*sibcall_value_memory): Likewise.
6850 Peepholes of indirect call and jump via memory: Likewise.
6851 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6852 (*sibcall_value_GOT_32): Likewise.
6853 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6854 (GOT_memory_operand): Likewise.
6855 (call_insn_operand): Likewise.
6856 (sibcall_insn_operand): Likewise.
6857 (GOT32_symbol_operand): Likewise.
6858 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6860 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6862 PR rtl-optimization/83496
6863 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6864 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6865 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6866 redundant insn, if any.
6867 (relax_delay_slots): Likewise.
6868 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6870 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6872 PR tree-optimization/83965
6873 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6874 that grouped statements are part of a reduction chain. Return
6875 true if the statement is not marked as a reduction itself but
6877 (vect_recog_dot_prod_pattern): Don't check whether the statement
6878 is part of a group here.
6879 (vect_recog_sad_pattern): Likewise.
6880 (vect_recog_widen_sum_pattern): Likewise.
6882 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6885 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6887 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6890 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6892 (ix86_output_indirect_function_return): New prototype.
6893 (ix86_split_simple_return_pop_internal): Likewise.
6894 * config/i386/i386.c (indirect_return_via_cx): New.
6895 (indirect_return_via_cx_bnd): Likewise.
6896 (indirect_thunk_name): Handle return va CX_REG.
6897 (output_indirect_thunk_function): Create alias for
6898 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6899 (ix86_output_indirect_jmp): Remove the bool argument.
6900 (ix86_output_indirect_function_return): New function.
6901 (ix86_split_simple_return_pop_internal): Likewise.
6902 * config/i386/i386.md (*indirect_jump): Don't pass false
6903 to ix86_output_indirect_jmp.
6904 (*tablejump_1): Likewise.
6905 (simple_return_pop_internal): Change it to define_insn_and_split.
6906 Call ix86_split_simple_return_pop_internal to split it for
6908 (simple_return_indirect_internal): Call
6909 ix86_output_indirect_function_return instead of
6910 ix86_output_indirect_jmp.
6912 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6915 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6916 memset and value initialization afterwards.
6918 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6920 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6922 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6925 * common/config/aarch64/aarch64-common.c
6926 (aarch_option_optimization_table[]): Switch
6927 off fomit-frame-pointer
6929 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6930 Chung-Ju Wu <jasonwucj@gmail.com>
6932 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6934 (store_multiple): Ditto.
6936 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6938 * config.gcc: Add --with-cpu support for nds32 target.
6939 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6940 * config/nds32/nds32.opt: Add -mcpu= option.
6942 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6944 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6945 isel=yes): Warn for these deprecated options.
6947 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6949 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6950 ISA_2_5_MASKS_EMBEDDED.
6952 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6954 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6955 p->max as pointers rather than using iterative_hash_expr.
6957 2018-02-23 Carl Love <cel@us.ibm.com>
6959 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6960 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6962 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6963 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6964 P8V_BUILTIN_VEC_VUNSIGNED2.
6966 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6969 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6970 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6972 * lra-constraints.c (curr_insn_transform): Set up
6973 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6975 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6976 * lra-eliminations.c (spill_pseudos): Ditto.
6977 (process_insn_for_elimination): Ditto.
6978 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6979 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6980 LRA_NON_CLOBBERED_ALT.
6982 2018-02-22 Martin Sebor <msebor@redhat.com>
6984 PR tree-optimization/84480
6985 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6986 to maybe_diag_stxncpy_trunc. Call it.
6987 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6988 from gimple_fold_builtin_strcpy. Print inlining stack.
6989 (handle_builtin_stxncpy): Print inlining stack.
6990 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6992 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6995 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6996 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6997 and -fcheck-pointer-bounds are used together.
6998 (indirect_thunk_prefix): New enum.
6999 (indirect_thunk_need_prefix): New function.
7000 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7001 "_nt" instead of "_bnd" for NOTRACK prefix.
7002 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7003 (output_indirect_thunk_function): Likewise.
7005 (ix86_code_end): Update output_indirect_thunk_function calls.
7006 (ix86_output_indirect_branch_via_reg): Replace
7007 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7008 (ix86_output_indirect_branch_via_push): Likewise.
7009 (ix86_output_function_return): Likewise.
7010 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7011 incompatible with -fcf-protection=branch and
7012 -fcheck-pointer-bounds.
7014 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7017 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7018 Change gcc_assert call to output_operand_lossage.
7020 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7022 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7024 2018-02-22 DJ Delorie <dj@redhat.com>
7025 Sebastian Perta <sebastian.perta@renesas.com>
7026 Oleg Endo <olegendo@gcc.gnu.org>
7028 * config/rx/rx.c (rx_rtx_costs): New function.
7029 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7031 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7033 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7035 2018-02-22 Martin Liska <mliska@suse.cz>
7038 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7039 Add "native" as a possible value.
7041 2018-02-22 Martin Liska <mliska@suse.cz>
7044 * config/i386/i386.c (ix86_option_override_internal):
7045 Add "native" as a possible value for -march and -mtune.
7047 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7050 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7051 to all type variants.
7053 PR tree-optimization/84503
7054 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7055 width as info->bitpos + info->bitsize - start.
7056 (merged_store_group::merge_overlapping): Simplify width computation.
7057 (check_no_overlap): New function.
7058 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7059 start + width and last_order of the group, fail if check_no_overlap
7061 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7062 to group if check_no_overlap fails.
7064 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7066 * config/rs6000/altivec.md: Delete contraint arguments to
7067 define_expand, define_split, and define_peephole2, and in
7068 define_insn_and_split if always unused.
7069 * config/rs6000/darwin.md: Ditto.
7070 * config/rs6000/dfp.md: Ditto.
7071 * config/rs6000/rs6000.md: Ditto.
7072 * config/rs6000/sync.md: Ditto.
7073 * config/rs6000/vector.md: Ditto.
7074 * config/rs6000/vsx.md: Ditto.
7076 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7078 * config/rs6000/altivec.md: Write output control strings as braced
7079 blocks instead of double-quoted strings.
7080 * config/rs6000/darwin.md: Ditto.
7081 * config/rs6000/rs6000.md: Ditto.
7082 * config/rs6000/vector.md: Ditto.
7083 * config/rs6000/vsx.md: Ditto.
7085 2018-02-21 Jason Merrill <jason@redhat.com>
7087 PR c++/84314 - ICE with templates and fastcall attribute.
7088 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7090 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7092 * ipa-cp.c (determine_versionability): Fix comment typos.
7094 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7097 * ipa-cp.c (determine_versionability): Do not version functions caling
7100 2018-02-21 Martin Liska <mliska@suse.cz>
7103 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7104 Add "native" as a possible value.
7105 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7106 the macro when native cpu detection is available.
7108 2018-02-21 Martin Liska <mliska@suse.cz>
7111 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7112 Add "native" as a possible value.
7113 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7114 when native cpu detection is available.
7116 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7117 Martin Sebor <msebor@redhat.com>
7119 PR tree-optimization/84478
7120 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7122 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7123 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7124 support which is conservatively correct, for 2 only stay conservative
7125 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7126 argument to the 2 argument get_range_strlen, adjust 6 arg
7127 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7129 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7130 (gimple_fold_builtin_strlen): Pass true as last argument to
7133 2018-02-20 Martin Sebor <msebor@redhat.com>
7136 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7137 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7138 (builtin_memref::builtin_memref): Factor out parts into
7139 set_base_and_offset and call it.
7141 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7144 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7145 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7146 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7147 search at the associated MODE_INT.
7149 2018-02-20 Jeff Law <law@redhat.com>
7152 PR tree-optimization/81592
7154 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7155 for range data rather than using global data.
7156 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7157 range data rather than using global data.
7158 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7159 pass it to children as needed.
7160 (struct directive::fmtresult): Similarly.
7161 (struct directive::set_width): Similarly.
7162 (struct directive::set_precision): Similarly.
7163 (format_integer, format_directive, parse_directive): Similarly.
7164 (format_none): Accept unnamed vr_values parameter.
7165 (format_percent, format_floating, format_character): Similarly.
7166 (format_string, format_plain): Similarly.
7167 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7168 the EVRP range analyzer for range data rather than using global data.
7169 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7170 gimple-ssa-evrp-analyze.h
7171 (class sprintf_dom_walker): Add after_dom_children member function.
7172 Add evrp_range_analyzer member.
7173 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7174 range analyzer as needed.
7175 (sprintf_dom_walker::after_dom_children): New member function.
7176 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7178 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7179 (evrp_range_analyzer::pop_to_marker): Likewise.
7181 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7183 PR tree-optimization/84419
7184 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7185 with the required type if its current type is compatible but
7188 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7191 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7192 after vectorization.
7194 2018-02-20 Martin Liska <mliska@suse.cz>
7197 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7198 possible values if we don't have a hint.
7200 2018-02-20 Martin Liska <mliska@suse.cz>
7204 * final.c (shorten_branches): Build align_tab array with one
7206 * opts.c (finish_options): Add alignment option limit check.
7207 (MAX_CODE_ALIGN): Likewise.
7208 (MAX_CODE_ALIGN_VALUE): Likewise.
7209 * doc/invoke.texi: Document maximum allowed option value for
7210 all -falign-* options.
7212 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7215 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7216 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7217 * var-tracking.c (emit_note_insn_var_location): Remove all references
7218 to NOTE_INSN_CALL_ARG_LOCATION.
7219 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7220 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7221 Use copy_rtx_if_shared.
7222 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7223 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7224 (dwarf2out_var_location): Remove handling of
7225 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7227 * final.c (final_scan_insn): Remove all references to
7228 NOTE_INSN_CALL_ARG_LOCATION.
7229 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7230 before dumping final insns.
7231 * except.c (emit_note_eh_region_end): Remove all references to
7232 NOTE_INSN_CALL_ARG_LOCATION.
7233 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7234 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7235 * config/arc/arc.c (hwloop_optimize): Likewise.
7236 * config/arm/arm.c (create_fix_barrier): Likewise.
7237 * config/s390/s390.c (s390_chunkify_start): Likewise.
7238 * config/sh/sh.c (find_barrier): Likewise.
7239 * config/i386/i386.c (rest_of_insert_endbranch,
7240 ix86_seh_fixup_eh_fallthru): Likewise.
7241 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7242 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7243 * config/frv/frv.c (frv_function_prologue): Likewise.
7244 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7246 (note_outside_basic_block_p): Remove all references to
7247 NOTE_INSN_CALL_ARG_LOCATION.
7248 * gengtype.c (adjust_field_rtx_def): Likewise.
7249 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7251 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7252 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7255 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7258 PR tree-optimization/84452
7259 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7260 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7263 2018-02-19 Martin Liska <mliska@suse.cz>
7266 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7268 2018-02-19 Martin Liska <mliska@suse.cz>
7269 Richard Sandiford <richard.sandiford@linaro.org>
7271 PR tree-optimization/82491
7272 * gimple-fold.c (get_base_constructor): Make earlier bail out
7275 2018-02-19 Carl Love <cel@us.ibm.com>
7277 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7278 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7280 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7281 P8V_BUILTIN_VEC_NEG.
7283 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7285 * config/rl78/rl78.md (movdf): New define expand.
7287 2018-02-19 Martin Liska <mliska@suse.cz>
7290 * doc/invoke.texi: Fix typo.
7291 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7293 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7295 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7296 handle rs6000_single_float and rs6000_double_float specially for
7299 2018-02-16 Jeff Law <law@redhat.com>
7301 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7304 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7307 * ipa-inline.c (inline_small_functions): Fix a typo.
7309 2018-02-16 Nathan Sidwell <nathan@acm.org>
7311 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7313 2018-02-16 Carl Love <cel@us.ibm.com>
7315 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7316 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7317 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7318 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7319 expansion to P8V_BUILTIN_VEC_FLOAT2.
7321 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7323 PR rtl-optimization/70023
7324 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7325 src_regno into account.
7327 2018-02-16 Carl Love <cel@us.ibm.com>
7329 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7330 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7331 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7332 * config/rs6000/rs6000.c: Remove case statements for
7333 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7334 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7335 and P9V_BUILTIN_VEC_VINSERT4B.
7336 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7337 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7338 * config/rs6000/vsx.md:
7339 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7342 2018-02-16 Carl Love <cel@us.ibm.com>
7344 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7346 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7348 * config/rs6000/rs6000-c.c: Add the definitions for
7349 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7350 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7351 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7352 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7353 definition for insert4b and define insn *insert3b_internal.
7354 * doc/extend.texi: Add documentation for vec_extract4b.
7356 2018-02-16 Nathan Sidwell <nathan@acm.org>
7358 * doc/extend.texi (Backwards Compatibility): Mention friend
7359 injection. Note for-scope is deprecated.
7360 * doc/invoke.texi (-ffriend-injection): Deprecate.
7362 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7364 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7365 that moved to I2, also allow destinations that are a paradoxical
7366 subreg (instead of a normal reg).
7368 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7371 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7374 2018-02-16 Richard Biener <rguenther@suse.de>
7376 PR tree-optimization/84037
7377 PR tree-optimization/84016
7379 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7380 Adjust vec_construct for the fact we need additional higher latency
7381 128bit inserts for AVX256 and AVX512 vector builds.
7382 (ix86_add_stmt_cost): Scale vector construction cost for
7385 2018-02-16 Richard Biener <rguenther@suse.de>
7387 PR tree-optimization/84417
7388 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7389 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7390 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7392 2018-02-16 Martin Liska <mliska@suse.cz>
7395 * internal-fn.def (ASAN_CHECK): Set proper flags.
7396 (ASAN_MARK): Likewise.
7398 2018-02-16 Julia Koval <julia.koval@intel.com>
7400 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7401 from PTA_CANNONLAKE.
7403 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7406 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7407 Use ++iter rather than iter++ for std::list iterators.
7408 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7409 defer deleting them until all nodes in the forest are processed. Do
7410 free even leaf nodes. Change to_process into auto_vec.
7413 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7414 * vec.h (vec_default_construct): Use memset instead of placement new
7415 if BROKEN_VALUE_INITIALIZATION is defined.
7416 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7417 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7420 PR rtl-optimization/83723
7421 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7422 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7423 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7425 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7427 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7429 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7431 PR rtl-optimization/81443
7432 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7433 from inner REGs to paradoxical SUBREGs.
7435 2018-02-16 Richard Biener <rguenther@suse.de>
7437 PR tree-optimization/84399
7438 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7439 For operands we can analyze at their definition make sure we can
7440 analyze them at each use as well.
7442 2018-02-16 Richard Biener <rguenther@suse.de>
7444 PR tree-optimization/84190
7445 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7446 volatile accesses if the decl isn't volatile.
7448 2018-02-15 Jason Merrill <jason@redhat.com>
7450 PR c++/84314 - ICE with templates and fastcall attribute.
7451 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7452 TYPE_CANONICAL on an existing type.
7454 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7456 PR tree-optimization/84383
7457 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7458 dstoff nor call operand_equal_p if dstbase is NULL.
7460 PR tree-optimization/84334
7461 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7462 also a CONSTANT_CLASS_P, punt.
7464 2018-02-14 Jim Wilson <jimw@sifive.com>
7466 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7467 first SMALL_OPERAND check. New local min_second_step. Move assert
7468 to where locals are set. Add TARGET_RVC support.
7469 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7471 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7473 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7475 2018-02-14 Martin Sebor <msebor@redhat.com>
7477 PR tree-optimization/83698
7478 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7479 arrays constrain the offset range to their bounds.
7480 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7481 (builtin_access::overlap): Avoid setting the size of overlap if it's
7483 (maybe_diag_overlap): Also consider arrays when deciding what values
7484 of offsets to include in diagnostics.
7486 2018-02-14 Martin Sebor <msebor@redhat.com>
7489 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7490 that correspond to the kind of a declaration.
7492 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7495 * config/pa/pa.md: Load address of PIC label using the linkage table
7496 if the label is nonlocal.
7498 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7500 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7501 warning message if user requests -maltivec=be.
7502 * doc/invoke.texi: Document deprecation of -maltivec=be.
7504 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7507 * config/rs6000/rs6000-c.c: Update definitions for
7508 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7509 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7511 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7514 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7515 add _get_ssp intrinsics. Remove argument from
7516 __builtin_ia32_rdssp[d|q].
7517 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7518 * config/i386/i386-builtin.def: Remove argument from
7519 __builtin_ia32_rdssp[d|q].
7520 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7521 ix86_expand_special_args_builtin for _rdssp[d|q].
7522 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7523 Clear register before usage.
7524 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7525 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7527 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7529 PR tree-optimization/84357
7530 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7531 operand 1 of an ARRAY_REF too.
7533 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7536 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7537 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7539 (set_of_reg): New struct.
7540 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7541 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7542 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7544 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7545 Split into bitclr, bitset, bitinvert patterns if appropriate.
7546 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7547 use rx_fuse_in_memory_bitop.
7548 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7549 to named insn, correct maximum insn length.
7551 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7554 * machmode.def: Define a complex mode for PARTIAL_INT.
7555 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7557 * doc/rtl.texi: Document CSPImode.
7558 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7560 (msp430_hard_regno_nregs_with_padding): Likewise.
7562 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7565 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7567 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7569 PR rtl-optimization/84169
7570 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7571 we generated a parallel as new i3 and we split that to new i2 and i3
7572 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7573 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7574 those to i2, not i1. Partially rewrite this scan code.
7576 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7579 * stor-layout.c (place_field): For variable length fields, adjust
7580 offset_align afterwards not just based on the field's alignment,
7581 but also on the size.
7584 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7585 of exps and logs in the use_exp2 case.
7587 2018-02-13 Jeff Law <law@redhat.com>
7589 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7592 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7595 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7599 * common.opt (gas-loc-support, gas-locview-support): New.
7600 (ginline-points, ginternal-reset-location-views): New.
7601 * doc/invoke.texi: Document them. Use @itemx where intended.
7602 (gvariable-location-views): Adjust.
7603 * target.def (reset_location_view): New.
7604 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7605 (TARGET_RESET_LOCATION_VIEW): New.
7606 * doc/tm.texi: Rebuilt.
7607 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7608 (dwarf2out_default_as_locview_support): New.
7609 (output_asm_line_debug_info): Use option variables.
7610 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7611 (output_loc_list): Likewise.
7612 (add_high_low_attributes): Check option variables.
7613 Don't output entry view attribute in strict mode.
7614 (gen_inlined_subroutine_die): Check option variables.
7615 (dwarf2out_inline_entry): Likewise.
7616 (init_sections_and_labels): Likewise.
7617 (dwarf2out_early_finish): Likewise.
7618 (maybe_reset_location_view): New, from...
7619 (dwarf2out_var_location): ... here. Call it.
7620 * debug.h (dwarf2out_default_as_loc_support): Declare.
7621 (dwarf2out_default_as_locview_support): Declare.
7622 * hooks.c (hook_int_rtx_insn_0): New.
7623 * hooks.h (hook_int_rtx_insn_0): Declare.
7624 * toplev.c (process_options): Take -gas-loc-support and
7625 -gas-locview-support from dwarf2out. Enable
7626 -gvariable-location-views by default only with locview
7627 assembler support. Enable -ginternal-reset-location-views by
7628 default only if the target defines the corresponding hook.
7629 Enable -ginline-points by default if location views are
7630 enabled; force it disabled if statement frontiers are
7632 * tree-inline.c (expand_call_inline): Check option variables.
7633 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7635 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7637 PR tree-optimization/84321
7638 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7639 handling. Also check whether the anti-range contains any values
7640 that satisfy the mask; switch to a VR_RANGE if not.
7642 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7645 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7647 2018-02-13 Martin Jambor <mjambor@suse.cz>
7650 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7651 of call statements, also set location of a load to a temporary.
7653 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7655 * config/rl78/rl78.c (add_vector_labels): New function.
7656 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7657 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7658 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7659 which checks that no arguments are passed.
7660 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7661 * doc/extend.texi: Documentation for the new attribute.
7663 2018-02-13 Andreas Schwab <schwab@suse.de>
7665 * config/riscv/linux.h (CPP_SPEC): Define.
7667 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7670 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7671 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7672 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7673 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7674 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7675 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7676 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7677 temporarily for AES and PCLMUL builtins.
7679 PR tree-optimization/84339
7680 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7681 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7685 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7686 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7687 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7689 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7691 * omp-simd-clone.h: New file.
7692 * omp-simd-clone.c: Include omp-simd-clone.h.
7693 (expand_simd_clones): No longer static.
7694 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7695 cgraph.h and omp-simd-clone.h.
7696 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7697 (vect_recog_widen_shift_pattern): Formatting fix.
7698 (vect_pattern_recog_1): Don't check optab for calls.
7701 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7702 operands[2] into a REG before using gen_lowpart on it.
7704 2018-02-12 Jeff Law <law@redhat.com>
7707 * config/sh/sh.c (find_barrier): Consider a sibling call
7710 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7711 successfully back substituting a reg.
7713 2018-02-12 Richard Biener <rguenther@suse.de>
7715 PR tree-optimization/84037
7716 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7717 parameter, move visited init to caller.
7718 (vect_slp_analyze_operations): Separate cost from validity
7719 check, initialize visited once for all instances.
7720 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7722 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7724 (vect_model_promotion_demotion_cost): Likewise.
7725 (vectorizable_bswap): Guard cost modeling with !slp_node
7726 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7728 (vectorizable_call): Likewise.
7729 (vectorizable_conversion): Likewise.
7730 (vectorizable_assignment): Likewise.
7731 (vectorizable_shift): Likewise.
7732 (vectorizable_operation): Likewise.
7733 (vectorizable_store): Likewise.
7734 (vectorizable_load): Likewise.
7735 (vectorizable_condition): Likewise.
7736 (vectorizable_comparison): Likewise.
7738 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7741 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7742 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7743 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7745 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7748 * params.def (inline-min-speedup): Increase from 8 to 15.
7749 (max-inline-insns-auto): Decrease from 40 to 30.
7750 * ipa-split.c (consider_split): Add some buffer for function to
7751 be considered inlining candidate.
7752 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7755 2018-02-12 Richard Biener <rguenther@suse.de>
7757 PR tree-optimization/84037
7758 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7759 matched stmts if we cannot swap the non-matched ones.
7761 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7763 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7764 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7765 _mm_maskz_scalef_round_ss): New intrinsics.
7766 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7767 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7768 __builtin_ia32_scalefss_round): Remove.
7769 (__builtin_ia32_scalefsd_mask_round,
7770 __builtin_ia32_scalefss_mask_round): New intrinsics.
7771 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7772 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7773 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7774 "<round_constraint>")): Changed to ...
7775 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7776 "<round_scalar_constraint>")): ... this.
7777 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7778 %0, %1, %2<round_op3>}"): Changed to ...
7779 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7780 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7781 %2<round_scalar_mask_op3>}"): ... this.
7782 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7784 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7786 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7787 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7788 (_mm_maskz_sqrt_round_ss): New intrinsics.
7789 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7790 (__builtin_ia32_sqrtsd_mask_round)
7791 (__builtin_ia32_sqrtss_mask_round): New builtins.
7792 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7793 (__builtin_ia32_sqrtss_round): Remove.
7794 (__builtin_ia32_sqrtsd_mask_round)
7795 (__builtin_ia32_sqrtss_mask_round): New builtins.
7796 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7797 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7798 ((match_operand:VF_128 1 "vector_operand"
7799 "xBm,<round_constraint>")): Changed to ...
7800 ((match_operand:VF_128 1 "vector_operand"
7801 "xBm,<round_scalar_constraint>")): ... this.
7802 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7803 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7804 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7805 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7806 %<iptr>1<round_scalar_mask_op3>}): ... this.
7807 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7808 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7810 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7813 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7814 Cast vec_cmpeq result to correct type.
7815 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7816 Cast vec_cmpgt result to correct type.
7818 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7820 * final.c (final_scan_insn_1): Renamed from...
7821 (final_scan_insn): ... this. New wrapper, to recover
7822 seen from the outermost call in recursive ones.
7823 * config/sparc/sparc.c (output_return): Drop seen from call.
7824 (output_sibcall): Likewise.
7825 * config/visium/visium.c (output_branch): Likewise.
7827 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7829 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7832 2018-02-10 Alan Modra <amodra@gmail.com>
7835 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7836 Specify LR as an input.
7838 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7841 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7842 remove_member_access_dummy_vars): New functions.
7843 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7844 lower_omp_1, execute_lower_omp): Use them.
7846 PR rtl-optimization/84308
7847 * shrink-wrap.c (spread_components): Release todo vector.
7849 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7851 PR rtl-optimization/57193
7852 * ira-color.c (struct allocno_color_data): Add member
7853 conflict_allocno_hard_prefs.
7854 (update_conflict_allocno_hard_prefs): New.
7855 (bucket_allocno_compare_func): Add a preference based on
7856 conflict_allocno_hard_prefs.
7857 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7858 (color_allocnos): Remove a dead code. Initiate
7859 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7861 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7864 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7865 constraint from =wa to wa. Avoid a subreg on the output operand,
7866 instead use a pseudo and subreg it in a move.
7867 (p9_xxbrd_<mode>): Changed to ...
7868 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7869 (p9_xxbrd_v2df): New expander.
7870 (p9_xxbrw_<mode>): Changed to ...
7871 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7872 (p9_xxbrw_v4sf): New expander.
7874 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7876 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7878 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7881 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7882 multiply in 32-bit mode.
7883 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7884 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7887 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7889 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7890 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7891 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7892 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7894 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7897 * dwarf2out.c (is_trivial_indirect_ref): New function.
7898 (dwarf2out_late_global_decl): Do not generate a location
7899 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7900 and that are not defined in the current unit.
7902 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7904 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7905 instead of a libcall for UNORDERED.
7907 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7910 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7911 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7913 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7916 * config/s390/s390.c (s390_set_current_function): Invoke
7917 s390_indirect_branch_settings also if fndecl didn't change.
7919 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7921 * config/rs6000/rs6000.md (blockage): Set length to zero.
7923 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7925 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7927 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7930 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7931 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7935 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7936 PARALLEL incoming that failed vt_get_decl_and_offset check.
7939 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7940 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7942 (get_variable_section): For decls in named .bss* sections pass true as
7943 second argument to bss_initializer_p.
7945 2018-02-09 Marek Polacek <polacek@redhat.com>
7946 Jakub Jelinek <jakub@redhat.com>
7949 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7950 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7951 Sync some changes from cxx_fold_indirect_ref.
7953 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7955 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7957 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7958 (BLOCK_INLINE_ENTRY_LABEL): New.
7959 (dwarf2out_var_location): Disregard inline entry markers.
7960 (inline_entry_data): New struct.
7961 (inline_entry_data_hasher): New hashtable type.
7962 (inline_entry_data_hasher::hash): New.
7963 (inline_entry_data_hasher::equal): New.
7964 (inline_entry_data_table): New variable.
7965 (add_high_low_attributes): Add DW_AT_entry_pc and
7966 DW_AT_GNU_entry_view attributes if a pending entry is found
7967 in inline_entry_data_table. Add old entry_pc attribute only
7968 if debug nonbinding markers are disabled.
7969 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7970 markers are enabled.
7971 (block_within_block_p, dwarf2out_inline_entry): New.
7972 (dwarf2out_finish): Check that no entries remained in
7973 inline_entry_data_table.
7974 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7975 (final_scan_insn, notice_source_line): Likewise.
7976 (rest_of_clean_state): Skip inline entry markers.
7977 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7979 * gimple.c (gimple_build_debug_inline_entry): New.
7980 * gimple.h (enum gimple_debug_subcode): Add
7981 GIMPLE_DEBUG_INLINE_ENTRY.
7982 (gimple_build_debug_inline_entry): Declare.
7983 (gimple_debug_inline_entry_p): New.
7984 (gimple_debug_nonbind_marker_p): Adjust.
7985 * insn-notes.def (INLINE_ENTRY): New.
7986 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7987 inline entry marker notes.
7988 (print_insn): Likewise.
7989 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7990 (INSN_DEBUG_MARKER_KIND): Likewise.
7991 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7992 * tree-inline.c (expand_call_inline): Build and insert
7993 debug_inline_entry stmt.
7994 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7995 inline entry blocks early, if nonbind markers are enabled.
7996 (dump_scope_block): Dump fragment info.
7997 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7998 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7999 (gimple_build_debug_inline_entry): New.
8000 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8001 Enable/disable inline entry points too.
8002 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8003 (DEBUG_INSN): Describe inline entry markers.
8005 * common.opt (gvariable-location-views): New.
8006 (gvariable-location-views=incompat5): New.
8007 * config.in: Rebuilt.
8008 * configure: Rebuilt.
8009 * configure.ac: Test assembler for view support.
8010 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8011 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8012 * dwarf2out.c (var_loc_view): New typedef.
8013 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8014 (dwarf2out_locviews_in_attribute): New.
8015 (dwarf2out_locviews_in_loclist): New.
8016 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8017 (enum dw_line_info_opcode): Add LI_adv_address.
8018 (struct dw_line_info_table): Add view.
8019 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8020 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8021 (zero_view_p): New variable.
8022 (ZERO_VIEW_P): New macro.
8023 (output_asm_line_debug_info): New.
8024 (struct var_loc_node): Add view.
8025 (add_AT_view_list, AT_loc_list): New.
8026 (add_var_loc_to_decl): Add view param. Test it against last.
8027 (new_loc_list): Add view params. Record them.
8028 (AT_loc_list_ptr): Handle loc and view lists.
8029 (view_list_to_loc_list_val_node): New.
8030 (print_dw_val): Handle dw_val_class_view_list.
8031 (size_of_die): Likewise.
8032 (value_format): Likewise.
8033 (loc_list_has_views): New.
8034 (gen_llsym): Set vl_symbol too.
8035 (maybe_gen_llsym, skip_loc_list_entry): New.
8036 (dwarf2out_maybe_output_loclist_view_pair): New.
8037 (output_loc_list): Output view list or entries too.
8038 (output_view_list_offset): New.
8039 (output_die): Handle dw_val_class_view_list.
8040 (output_dwarf_version): New.
8041 (output_compilation_unit_header): Use it.
8042 (output_skeleton_debug_sections): Likewise.
8043 (output_rnglists, output_line_info): Likewise.
8044 (output_pubnames, output_aranges): Update version comments.
8045 (output_one_line_info_table): Output view numbers in asm comments.
8046 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8047 Call maybe_gen_llsym.
8048 (loc_list_from_tree_1): Adjust.
8049 (add_AT_location_description): Create view list attribute if
8050 needed, check it's absent otherwise.
8051 (convert_cfa_to_fb_loc_list): Adjust.
8052 (maybe_emit_file): Call output_asm_line_debug_info for test.
8053 (dwarf2out_var_location): Reset views as needed. Precompute
8054 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8055 attribute. Set view.
8056 (new_line_info_table): Reset next view.
8057 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8058 (dwarf2out_source_line): Likewise. Output view resets and labels to
8059 the assembler, or select appropriate line info opcodes.
8060 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8061 (optimize_string_length): Catch it. Adjust.
8062 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8063 dw_val_class_view_list, and remove it if no longer needed.
8064 (hash_loc_list): Hash view numbers.
8065 (loc_list_hasher::equal): Compare them.
8066 (optimize_location_lists): Check whether a view list symbol is
8067 needed, and whether the locview attribute is present, and
8068 whether they match. Remove the locview attribute if no longer
8070 (index_location_lists): Call skip_loc_list_entry for test.
8071 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8072 Use output_dwarf_version.
8073 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8074 (struct dw_val_node): Add val_view_list.
8075 * final.c (SEEN_NEXT_VIEW): New.
8076 (set_next_view_needed): New.
8077 (clear_next_view_needed): New.
8078 (maybe_output_next_view): New.
8079 (final_start_function): Rename to...
8080 (final_start_function_1): ... this. Take pointer to FIRST,
8081 add SEEN parameter. Emit param bindings in the initial view.
8082 (final_start_function): Reintroduce SEEN-less interface.
8083 (final): Rename to...
8084 (final_1): ... this. Take SEEN parameter. Output final pending
8085 next view at the end.
8086 (final): Reintroduce seen-less interface.
8087 (final_scan_insn): Output pending next view before switching
8088 sections or ending a block. Mark the next view as needed when
8089 outputting variable locations. Notify debug backend of section
8090 changes, and of location view changes.
8091 (rest_of_handle_final): Adjust.
8092 * toplev.c (process_options): Autodetect value for debug variable
8093 location views option. Warn on incompat5 without -gdwarf-5.
8094 * doc/invoke.texi (gvariable-location-views): New.
8095 (gvariable-location-views=incompat5): New.
8096 (gno-variable-location-views): New.
8098 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8100 PR tree-optimization/84136
8101 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8102 that the result of find_edge is non-NULL.
8104 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8107 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8108 storing integer register in SImode. Fix cost of 256 and 512
8109 byte aligned SSE register store.
8111 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8113 * config/i386/i386.c (ix86_multiplication_cost): Fix
8114 multiplication cost for TARGET_AVX512DQ.
8116 2018-02-08 Marek Polacek <polacek@redhat.com>
8118 PR tree-optimization/84238
8119 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8122 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8124 PR tree-optimization/84265
8125 * tree-vect-stmts.c (vectorizable_store): Don't treat
8126 VMAT_CONTIGUOUS accesses as grouped.
8127 (vectorizable_load): Likewise.
8129 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8131 PR tree-optimization/81635
8132 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8133 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8134 (test_round_for_mask): New functions.
8135 (wide_int_cc_tests): Call test_round_for_mask.
8136 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8137 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8138 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8139 range returned by get_range_info.
8141 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8144 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8145 * symtab.c: Include builtins.h
8146 (symtab_node::output_to_lto_symbol_table_p): Move here
8147 from lto-streamer-out.c:output_symbol_p.
8148 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8149 (output_symbol_p): Move all logic to symtab.c
8150 (produce_symtab): Update.
8152 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8154 * config/s390/s390-opts.h (enum indirect_branch): Define.
8155 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8156 (s390_indirect_branch_via_thunk)
8157 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8158 (enum s390_indirect_branch_type): Define.
8159 * config/s390/s390.c (struct s390_frame_layout, struct
8160 machine_function): Remove.
8161 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8162 (indirect_branch_table_label_no, indirect_branch_table_name):
8164 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8165 (enum s390_indirect_branch_option): Define.
8166 (s390_return_addr_from_memory): New function.
8167 (s390_handle_string_attribute): New function.
8168 (s390_attribute_table): Add new attribute handler.
8169 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8170 (s390_indirect_branch_via_thunk): New function.
8171 (s390_indirect_branch_via_inline_thunk): New function.
8172 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8173 sibling call optimization for non z10 compiles.
8174 (s390_emit_call): Force indirect branch target to be a single
8175 register. Add r1 clobber for non-z10 compiles.
8176 (s390_emit_epilogue): Emit return jump via return_use expander.
8177 (s390_reorg): Handle JUMP_INSNs as execute targets.
8178 (s390_option_override_internal): Perform validity checks for the
8179 new command line options.
8180 (s390_indirect_branch_attrvalue): New function.
8181 (s390_indirect_branch_settings): New function.
8182 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8183 (s390_output_indirect_thunk_function): New function.
8184 (s390_code_end): Implement target hook.
8185 (s390_case_values_threshold): Implement target hook.
8186 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8188 * config/s390/s390.h (struct s390_frame_layout)
8189 (struct machine_function): Move here from s390.c.
8190 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8191 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8192 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8193 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8194 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8195 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8196 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8197 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8198 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8199 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8200 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8201 (mnemonic attribute): Add values which aren't recognized
8203 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8204 pattern for branch conversion. Fix mnemonic attribute.
8205 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8206 indirect branch via thunk if requested.
8207 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8208 ("*indirect_jump"): Disable for branch conversion using out of
8210 ("indirect_jump_via_thunk<mode>_z10")
8211 ("indirect_jump_via_thunk<mode>")
8212 ("indirect_jump_via_inlinethunk<mode>_z10")
8213 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8214 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8215 ("casesi_jump_via_inlinethunk<mode>_z10")
8216 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8217 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8218 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8219 ("*indirect2_jump"): Disable for branch conversion.
8220 ("casesi_jump"): Turn into expander and expand patterns for branch
8222 ("return_use"): New expander.
8223 ("*return"): Emit return via thunk and rename it to ...
8224 ("*return<mode>"): ... this one.
8225 * config/s390/s390.opt: Add new options and and enum for the
8228 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8230 * lra-constraints.c (match_reload): Unconditionally use
8231 gen_lowpart_SUBREG, rather than selecting between that
8232 and equivalent gen_rtx_SUBREG code.
8234 2018-02-08 Richard Biener <rguenther@suse.de>
8236 PR tree-optimization/84233
8237 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8238 changed flag instead of boguously re-using phi_inserted.
8240 2018-02-08 Martin Jambor <mjambor@suse.cz>
8242 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8243 static local variables.
8245 2018-02-08 Richard Biener <rguenther@suse.de>
8247 PR tree-optimization/84278
8248 * tree-vect-stmts.c (vectorizable_store): When looking for
8249 smaller vector types to perform grouped strided loads/stores
8250 make sure the mode is supported by the target.
8251 (vectorizable_load): Likewise.
8253 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8255 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8256 Increase LDP/STP opportunities by adding adjacent callee-saves.
8258 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8260 PR rtl-optimization/84068
8261 PR rtl-optimization/83459
8262 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8264 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8266 PR tree-optimization/84224
8267 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8268 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8271 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8274 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8275 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8276 count, remove one USE.
8278 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8280 * doc/install.texi (Configuration): Document the
8281 --with-long-double-format={ibm,ieee} PowerPC configuration
8285 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8286 Convert from define_expand to be define_insn_and_split. Rework
8287 float/double/_Float128 conversions to QI/HI/SImode to work with
8288 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8289 conversions to QI/HImode types did a store and then a load to
8290 truncate the value. For conversions to VSX registers, don't split
8291 the insn, instead emit the code directly. Use the code iterator
8292 any_fix to combine signed and unsigned conversions.
8293 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8294 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8295 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8296 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8297 (fix_<mode>di2_hw): Likewise.
8298 (fixuns_<mode>di2_hw): Likewise.
8299 (fix_<mode>si2_hw): Likewise.
8300 (fixuns_<mode>si2_hw): Likewise.
8301 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8302 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8303 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8304 fix<uns>_trunc<SFDF:mode>si2_p8.
8305 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8307 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8308 (fix<uns>_<mode>_mem): Likewise.
8309 (fctiw<u>z_<mode>_mem): Likewise.
8310 (fix<uns>_<mode>_mem): Likewise.
8311 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8312 the register allocator from doing a direct move to the GPRs to do
8313 a store, and instead use the ISA 3.0 store byte/half-word from
8314 vector register instruction. For IEEE 128-bit floating point,
8315 also optimize stores of 32-bit ints.
8316 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8318 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8320 * genextract.c (push_pathstr_operand): New function to support
8322 (walk_rtx): Call push_pathstr_operand.
8323 (print_path): Support [a-zA-Z].
8325 2018-02-07 Richard Biener <rguenther@suse.de>
8327 PR tree-optimization/84037
8328 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8329 (cse_and_gimplify_to_preheader): Declare.
8330 (vect_get_place_in_interleaving_chain): Likewise.
8331 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8333 (_loop_vec_info::~_loop_vec_info): Delete it.
8334 (cse_and_gimplify_to_preheader): New function.
8335 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8336 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8337 (vectorizable_load): Likewise. For grouped stores always base
8338 the IV on the first element.
8339 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8340 condition before gimplifying.
8342 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8344 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8345 *DIV_EXPR and *MOD_EXPR.
8347 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8350 * config/i386/i386.c (ix86_option_override_internal): Mask out
8351 the CF_SET bit when checking -fcf-protection.
8353 2018-02-07 Tom de Vries <tom@codesourcery.com>
8356 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8359 2018-02-07 Richard Biener <rguenther@suse.de>
8361 PR tree-optimization/84204
8362 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8365 PR tree-optimization/84205
8366 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8367 special-case isl_ast_op_zdiv_r.
8369 PR tree-optimization/84223
8370 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8371 Only add conditions from within the region.
8372 (gather_bbs::after_dom_children): Adjust.
8374 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8377 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8378 * config/avr/avr.md: Only post-reload split REG-REG moves if
8379 either register is GENERAL_REG_P.
8381 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8383 PR tree-optimization/84235
8384 * tree-ssa-scopedtables.c
8385 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8386 if the subtraction is performed in floating point type where NaNs are
8387 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8388 build 1. Formatting fix.
8390 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8393 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8394 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8395 and skip it regardless of bb boundaries. Use CALL_P macro,
8396 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8397 unnecessarily, formatting fix.
8399 2018-02-06 Michael Collison <michael.collison@arm.com>
8401 * config/arm/thumb2.md:
8402 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8403 (*thumb_mov_notscc): Ditto.
8405 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8408 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8409 unsigned_fix, not "s".
8411 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8413 * configure.ac (gcc_fn_eh_frame_ro): New function.
8414 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8415 correct .eh_frame permissions.
8416 * configure: Regenerate.
8418 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8420 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8423 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8425 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8426 Display warning message for -mno-speculate-indirect-jumps.
8428 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8430 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8432 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8434 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8436 PR tree-optimization/84225
8437 * tree-eh.c (find_trapping_overflow): Only call
8438 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8440 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8443 * config/i386/i386.c: Reimplement the check of possible options
8444 -mibt/-mshstk conbination. Change error messages.
8445 * doc/invoke.texi: Fix a typo: remove extra '='.
8447 2018-02-06 Marek Polacek <polacek@redhat.com>
8449 PR tree-optimization/84228
8450 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8452 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8455 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8456 emitted arch directives.
8457 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8458 __ARM_FEATURE_COPROC before changing architectures.
8460 2018-02-06 Richard Biener <rguenther@suse.de>
8462 * config/i386/i386.c (print_reg): Fix typo.
8463 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8465 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8467 * configure: Regenerate.
8469 2018-02-05 Martin Sebor <msebor@redhat.com>
8471 PR tree-optimization/83369
8472 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8475 2018-02-05 Martin Liska <mliska@suse.cz>
8477 * doc/invoke.texi: Cherry-pick upstream r323995.
8479 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8481 * ira.c (ira_init_register_move_cost): Adjust comment.
8483 2018-02-05 Martin Liska <mliska@suse.cz>
8485 PR gcov-profile/84137
8486 * doc/gcov.texi: Fix typo in documentation.
8488 2018-02-05 Martin Liska <mliska@suse.cz>
8490 PR gcov-profile/83879
8491 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8492 using dlopen functionality.
8494 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8496 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8497 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8498 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8499 _mm_maskz_range_round_ss): New intrinsics.
8500 (__builtin_ia32_rangesd128_round)
8501 (__builtin_ia32_rangess128_round): Remove.
8502 (__builtin_ia32_rangesd128_mask_round,
8503 __builtin_ia32_rangess128_mask_round): New builtins.
8504 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8505 __builtin_ia32_rangess128_round): Remove.
8506 (__builtin_ia32_rangesd128_mask_round,
8507 __builtin_ia32_rangess128_mask_round): New builtins.
8508 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8509 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8510 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8511 "<round_saeonly_constraint>")): Changed to ...
8512 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8513 "<round_saeonly_scalar_constraint>")): ... this.
8514 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8515 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8516 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8517 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8518 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8520 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8522 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8524 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8525 Remove all values except native, 8540 and 8548.
8527 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8529 * config/i386/i386.c (ix86_output_function_return): Pass
8530 INVALID_REGNUM, instead of -1, as invalid register number to
8531 indirect_thunk_name and output_indirect_thunk.
8533 2018-02-02 Julia Koval <julia.koval@intel.com>
8535 * config.gcc: Add -march=icelake.
8536 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8537 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8538 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8539 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8540 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8541 (processor_target_table): Add icelake.
8542 (ix86_option_override_internal): Handle new PTAs.
8543 (get_builtin_code_for_version): Handle icelake.
8544 (M_INTEL_COREI7_ICELAKE): New.
8545 (fold_builtin_cpu): Handle icelake.
8546 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8547 * doc/invoke.texi: Add -march=icelake.
8549 2018-02-02 Julia Koval <julia.koval@intel.com>
8551 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8552 to wide_int_bitmask.
8553 * wide-int-bitmask.h: New.
8555 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8558 * config/i386/i386.md: Replace Pmode with word_mode in
8559 builtin_setjmp_setup and builtin_longjmp to support x32.
8561 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8565 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8567 (rs6000_supported_cpu_names): New static variable.
8568 (linux_cpu_translation_table): Likewise.
8569 (elf_platform) <cpu>: Define new static variable and use it.
8570 Translate kernel AT_PLATFORM name to canonical name if needed.
8571 Error if platform name is unknown.
8573 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8576 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8578 2018-02-01 Jeff Law <law@redhat.com>
8581 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8582 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8583 the scratch if RELEASE_VIA_POP is false.
8584 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8585 If we have to save a temporary register, decrement SIZE appropriately.
8586 Pass new arguments to release_scratch_register_on_entry.
8587 (ix86_adjust_stack_and_probe): Likewise.
8588 (ix86_emit_probe_stack_range): Pass new arguments to
8589 release_scratch_register_on_entry.
8591 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8593 PR rtl-optimization/84157
8594 * combine.c (change_zero_ext): Use REG_P predicate in
8595 front of HARD_REGISTER_P predicate.
8597 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8599 * config/avr/avr.c (avr_option_override): Move disabling of
8600 -fdelete-null-pointer-checks to...
8601 * common/config/avr/avr-common.c (avr_option_optimization_table):
8604 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8606 PR tree-optimization/81635
8607 * tree-data-ref.c (split_constant_offset_1): For types that
8608 wrap on overflow, try to use range info to prove that wrapping
8611 2018-02-01 Renlin Li <renlin.li@arm.com>
8614 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8616 (aarch64_register_move_cost): Likewise.
8617 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8619 (REG_CLASS_NAMES): Likewise.
8620 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8621 TAILCALL_ADDR_REGS. Remove IP registers.
8622 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8624 2018-02-01 Richard Biener <rguenther@suse.de>
8626 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8627 for specifying RPO order and allow NULL for that.
8628 * domwalk.c (dom_walker::dom_walker): Likewise.
8629 (dom_walker::walk): Handle NULL RPO order.
8630 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8632 (rewrite_update_dom_walker): Likewise.
8633 (mark_def_dom_walker): Likewise.
8635 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8637 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8638 (aarch64_maybe_expand_sve_subreg_move): Declare.
8639 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8640 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8642 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8643 that are semantically a reverse operation.
8644 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8645 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8646 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8648 (aarch64_can_change_mode_class): For big-endian, forbid changes
8649 between two SVE modes if they have different element sizes.
8651 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8653 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8654 the TImode handling for big-endian targets.
8656 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8658 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8659 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8661 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8662 Remove BSWAP handing for big-endian targets and use the form of
8663 LD1RQ appropariate for the mode.
8665 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8667 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8668 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8671 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8674 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8675 check for operands that need to go through aarch64_sve_reload_be.
8677 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8679 PR tree-optimization/81661
8680 PR tree-optimization/84117
8681 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8682 * tree-eh.c: Include gimplify.h.
8683 (find_trapping_overflow, replace_trapping_overflow,
8684 rewrite_to_non_trapping_overflow): New functions.
8685 * tree-vect-loop.c: Include tree-eh.h.
8686 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8687 * tree-data-ref.c: Include tree-eh.h.
8688 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8690 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8692 PR rtl-optimization/84123
8693 * combine.c (change_zero_ext): Check if hard register satisfies
8694 can_change_dest_mode before calling gen_lowpart_SUBREG.
8696 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8699 * ira.c (ira_init_register_move_cost): Remove assert.
8701 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8703 PR rtl-optimization/84071
8704 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8705 * doc/tm.texi: Regenerate.
8707 2018-01-31 Richard Biener <rguenther@suse.de>
8709 PR tree-optimization/84132
8710 * tree-data-ref.c (analyze_miv_subscript): Properly
8711 check whether evolution_function_is_affine_multivariate_p
8712 before calling gcd_of_steps_may_divide_p.
8714 2018-01-31 Julia Koval <julia.koval@intel.com>
8717 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8718 * config/i386/i386.md (rdpid_rex64) New.
8719 (rdpid): Make 32bit only.
8721 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8724 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8725 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8727 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8730 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8732 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8734 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8736 PR rtl-optimization/84071
8737 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8738 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8740 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8742 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8743 (arc_attribute_table): Add 'aux' attribute.
8744 (arc_in_small_data_p): Consider aux like variables.
8745 (arc_is_aux_reg_p): New function.
8746 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8747 (arc_get_aux_arg): New function.
8748 (prepare_move_operands): Handle aux-register access.
8749 (arc_handle_aux_attribute): New function.
8750 * doc/extend.texi (ARC Variable attributes): Add subsection.
8752 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8754 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8755 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8756 (arc_attribute_table): Add 'uncached' attribute.
8757 (arc_print_operand): Print '.di' flag for uncached memory
8759 (arc_in_small_data_p): Do not consider for small data the uncached
8761 (arc_is_uncached_mem_p): New function.
8762 * config/arc/predicates.md (compact_store_memory_operand): Check
8763 for uncached memory accesses.
8764 (nonvol_nonimm_operand): Likewise.
8765 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8767 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8770 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8771 falign-loops=): Add Optimization flag.
8773 2018-01-30 Jeff Law <law@redhat.com>
8776 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8777 INT_REGISTERS_SAVED. Check it prior to calling
8778 get_scratch_register_on_entry.
8779 (ix86_adjust_stack_and_probe): Similarly.
8780 (ix86_emit_probe_stack_range): Similarly.
8781 (ix86_expand_prologue): Corresponding changes.
8783 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8786 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8787 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8789 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8792 * lra-constraints.c (curr_insn_transform): Process AND in the
8795 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8797 PR rtl-optimization/83986
8798 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8799 dependence against last_pending_memory_flush in addition to
8802 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8804 PR tree-optimization/81611
8805 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8808 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8811 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8814 2018-01-30 Richard Biener <rguenther@suse.de>
8815 Jakub Jelinek <jakub@redhat.com>
8817 PR tree-optimization/84111
8818 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8819 inner loops added during recursion, as they don't have up-to-date
8822 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8825 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8826 (can_inline_edge_by_limits_p): ... here.
8827 (can_early_inline_edge_p, check_callers,
8828 update_caller_keys, update_callee_keys, recursive_inlining,
8829 add_new_edges_to_heap, speculation_useful_p,
8830 inline_small_functions,
8831 inline_small_functions, flatten_function,
8832 inline_to_all_callers_1): Update.
8834 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8836 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8839 2018-01-30 Richard Biener <rguenther@suse.de>
8841 PR tree-optimization/83008
8842 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8843 invariant and constant vector uses in stmts when they need
8846 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8849 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8850 * configure: Regenerate.
8852 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8854 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8856 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8857 Use gen_rtx_REG rather than gen_lowpart.
8859 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8861 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8862 rather than 0 when creating partial subregs.
8864 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8866 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8869 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8872 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8873 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8874 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8875 flags. This restores the settings used before the 2017-07-24.
8876 Turning off pre increment/decrement/modify allows IVOPTS to
8877 optimize DF/SF loops where the index is an int.
8879 2018-01-29 Richard Biener <rguenther@suse.de>
8880 Kelvin Nilsen <kelvin@gcc.gnu.org>
8883 * tree-vect-stmts.c (vectorizable_call): Don't call
8884 targetm.vectorize_builtin_md_vectorized_function if callee is
8887 2018-01-22 Carl Love <cel@us.ibm.com>
8889 * doc/extend.tex: Fix typo in second arg in
8890 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8892 2018-01-29 Richard Biener <rguenther@suse.de>
8894 PR tree-optimization/84086
8895 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8896 (flush_ssaname_freelist): When SSA names were released reset
8897 the SCEV hash table.
8899 2018-01-29 Richard Biener <rguenther@suse.de>
8901 PR tree-optimization/84057
8902 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8903 removed paths when removing edges.
8905 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8907 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8908 -mfunction-return=@var{choice}.
8910 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8913 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8915 (layout::print_source_line): Likewise.
8916 (test_get_line_width_without_trailing_whitespace): Add test cases.
8918 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8921 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8924 2018-01-26 Jim Wilson <jimw@sifive.com>
8926 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8928 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8931 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8933 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8934 and CMP + SUB-immediate -> SUBS.
8936 2018-01-26 Martin Sebor <msebor@redhat.com>
8938 PR tree-optimization/83896
8939 * tree-ssa-strlen.c (get_string_len): Rename...
8940 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8941 Avoid assuming length is constant.
8942 (handle_char_store): Use HOST_WIDE_INT for string length.
8944 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8947 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8948 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8950 2018-01-26 Richard Biener <rguenther@suse.de>
8952 PR rtl-optimization/84003
8953 * dse.c (record_store): Only record redundant stores when
8954 the earlier store aliases at least all accesses the later one does.
8956 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8958 PR rtl-optimization/83985
8959 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8960 REG_CFA_RESTORE insns.
8961 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8962 REG_CFA_RESTORE insns here.
8965 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8966 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8968 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8970 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8971 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8972 (arc_init): Likewise.
8973 (arc_override_options): Likewise.
8974 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8976 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8978 * config/arc/arc.h (TARGET_DBNZ): Define.
8979 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8980 properly set the tune attribute.
8981 (dbnz): Use TARGET_DBNZ guard.
8982 * config/arc/arc.opt (mtune): Add core3 option.
8984 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8986 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8987 recognize new pic like addresses.
8988 (arc_delegitimize_address): Clean up.
8990 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8992 * config/arc/arc-arches.def: Option mrf16 valid for all
8994 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8995 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8996 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8997 * config/arc/arc-tables.opt: Regenerate.
8998 * config/arc/arc.c (arc_conditional_register_usage): Handle
8999 reduced register file case.
9000 (arc_file_start): Set must have build attributes.
9001 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9003 * config/arc/arc.opt (mrf16): Add new option.
9004 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9005 * config/arc/genmultilib.awk: Handle new mrf16 option.
9006 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9007 * config/arc/t-multilib: Regenerate.
9008 * doc/invoke.texi (ARC Options): Document mrf16 option.
9010 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9012 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9013 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9014 (arc_attribute_table): Add 'secure_call' attribute.
9015 (arc_print_operand): Print secure call operand.
9016 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9018 (arc_is_secure_call_p): New function. * config/arc/arc.md
9019 (call_i): Add support for sjli instruction.
9020 (call_value_i): Likewise.
9021 * config/arc/constraints.md (Csc): New constraint.
9023 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9024 John Eric Martin <John.Martin@emmicro-us.com>
9026 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9027 * config/arc/arc.c (_arc_jli_section): New struct.
9028 (arc_jli_section): New type.
9029 (rc_jli_sections): New static variable.
9030 (arc_handle_jli_attribute): New function.
9031 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9032 (arc_file_end): New function.
9033 (TARGET_ASM_FILE_END): Define.
9034 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9035 (arc_add_jli_section): New function.
9036 (jli_call_scan): Likewise.
9037 (arc_reorg): Call jli_call_scan.
9038 (arc_output_addsi): Remove 'S' from printing asm operand.
9039 (arc_is_jli_call_p): New function.
9040 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9042 (movhi_insn): Likewise.
9043 (movsi_insn): Likewise.
9044 (movsi_set_cc_insn): Likewise.
9045 (loadqi_update): Likewise.
9046 (load_zeroextendqisi_update): Likewise.
9047 (load_signextendqisi_update): Likewise.
9048 (loadhi_update): Likewise.
9049 (load_zeroextendhisi_update): Likewise.
9050 (load_signextendhisi_update): Likewise.
9051 (loadsi_update): Likewise.
9052 (loadsf_update): Likewise.
9053 (movsicc_insn): Likewise.
9054 (bset_insn): Likewise.
9055 (bxor_insn): Likewise.
9056 (bclr_insn): Likewise.
9057 (bmsk_insn): Likewise.
9058 (bicsi3_insn): Likewise.
9059 (cmpsi_cc_c_insn): Likewise.
9060 (movsi_ne): Likewise.
9061 (movsi_cond_exec): Likewise.
9062 (clrsbsi2): Likewise.
9073 (call_i): Remove 'S' asm letter, add jli instruction.
9074 (call_value_i): Likewise.
9075 * config/arc/arc.op (mjli-always): New option.
9076 * config/arc/constraints.md (Cji): New constraint.
9077 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9079 (subsf3_fpx): Likewise.
9080 (mulsf3_fpx): Likewise.
9081 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9083 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9085 * doc/invoke.texi (ARC): Document mjli-always option.
9087 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9089 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9090 avoid addition with 0 and use incw and decw where possible.
9092 2018-01-26 Richard Biener <rguenther@suse.de>
9094 PR tree-optimization/81082
9095 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9096 association if it requires casting to unsigned.
9097 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9098 from fold_plusminus_mult_expr to catch important cases late when
9099 range info is available.
9101 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9103 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9104 * configure.ac (hidden_linkonce): New test.
9105 * configure: Regenerate.
9106 * config.in: Regenerate.
9108 2018-01-26 Julia Koval <julia.koval@intel.com>
9110 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9111 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9112 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9113 _mm_mask_bitshuffle_epi64_mask): Fix type.
9114 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9115 USI_FTYPE_V4DI_V4DI_USI): Remove.
9116 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9117 __builtin_ia32_vpshufbitqmb256_mask,
9118 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9119 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9120 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9122 2018-01-26 Alan Modra <amodra@gmail.com>
9125 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9126 UNSPEC_VBPERMQ. Sort other unspecs.
9128 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9130 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9132 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9135 * predict.c (drop_profile): Do not push/pop cfun; update also
9137 (handle_missing_profiles): Fix logic looking for zero profiles.
9139 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9142 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9143 on functions with #pragma omp declare simd or functions with simd
9145 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9146 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9147 Remove trailing \n from warning_at calls.
9149 2018-01-25 Tom de Vries <tom@codesourcery.com>
9152 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9153 for neutered workers.
9155 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9158 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9159 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9161 2018-01-24 Jeff Law <law@redhat.com>
9164 * i386.c (get_probe_interval): Move to earlier point.
9165 (ix86_compute_frame_layout): If -fstack-clash-protection and
9166 the frame is larger than the probe interval, then use pushes
9167 to save registers rather than reg->mem moves.
9168 (ix86_expand_prologue): Remove conditional for int_registers_saved
9171 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9174 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9175 min/max for never referenced object.
9177 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9180 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9182 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9183 attributes from DECL_ATTRIBUTES (decl) without affecting
9184 DECL_ATTRIBUTES (current_function_decl).
9185 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9186 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9188 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9190 PR tree-optimization/83979
9191 * fold-const.c (fold_comparison): Use constant_boolean_node
9192 instead of boolean_{true,false}_node.
9194 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9196 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9199 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9201 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9202 Simplify the clause that sets the length attribute.
9203 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9204 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9205 clause that sets the length attribute.
9206 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9208 2018-01-24 Tom de Vries <tom@codesourcery.com>
9211 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9212 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9213 Add strict parameter.
9214 (prevent_branch_around_nothing): Insert dummy insn between branch to
9215 label and label with no ptx insn inbetween.
9216 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9218 2018-01-24 Tom de Vries <tom@codesourcery.com>
9221 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9222 for neutered threads in warp.
9223 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9225 2018-01-24 Richard Biener <rguenther@suse.de>
9227 PR tree-optimization/83176
9228 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9231 2018-01-24 Richard Biener <rguenther@suse.de>
9233 PR tree-optimization/82819
9234 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9235 code generating pluses that are no-ops in the target precision.
9237 2018-01-24 Richard Biener <rguenther@suse.de>
9240 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9242 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9244 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9245 to merge probabilities.
9246 * predict.c (probably_never_executed): Also mark as cold functions
9247 with global 0 profile and guessed local profile.
9248 * profile-count.c (profile_probability::combine_with_count): New
9250 * profile-count.h (profile_probability::operator*,
9251 profile_probability::operator*=, profile_probability::operator/,
9252 profile_probability::operator/=): Reduce precision to adjusted
9253 and set value to guessed on contradictory divisions.
9254 (profile_probability::combine_with_freq): Remove.
9255 (profile_probability::combine_wiht_count): Declare.
9256 (profile_count::force_nonzero):: Set to adjusted.
9257 (profile_count::probability_in):: Set quality to adjusted.
9258 * tree-ssa-tail-merge.c (replace_block_by): Use
9261 2018-01-23 Andrew Waterman <andrew@sifive.com>
9262 Jim Wilson <jimw@sifive.com>
9264 * config/riscv/riscv.c (riscv_stack_boundary): New.
9265 (riscv_option_override): Set riscv_stack_boundary. Handle
9266 riscv_preferred_stack_boundary_arg.
9267 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9268 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9269 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9270 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9271 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9272 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9274 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9277 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9278 of struct ix86_frame.
9279 (ix86_expand_epilogue): Likewise. Add a local variable for
9280 the reg_save_offset field in struct ix86_frame.
9282 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9284 PR tree-optimization/82604
9285 * tree-loop-distribution.c (enum partition_kind): New enum item
9286 PKIND_PARTIAL_MEMSET.
9287 (partition_builtin_p): Support above new enum item.
9288 (generate_code_for_partition): Ditto.
9289 (compute_access_range): Differentiate cases that equality can be
9290 proven at all loops, the innermost loops or no loops.
9291 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9292 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9293 (finalize_partitions, distribute_loop): Don't fuse partition of
9294 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9295 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9298 2018-01-23 Martin Liska <mliska@suse.cz>
9300 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9301 order to ignore the predictor.
9302 (PRED_POLYMORPHIC_CALL): Likewise.
9303 (PRED_RECURSIVE_CALL): Likewise.
9305 2018-01-23 Martin Liska <mliska@suse.cz>
9307 * tree-profile.c (tree_profiling): Print function header to
9308 aware reader which function we are working on.
9309 * value-prof.c (gimple_find_values_to_profile): Do not print
9310 not interesting value histograms.
9312 2018-01-23 Martin Liska <mliska@suse.cz>
9314 * profile-count.h (enum profile_quality): Add
9315 profile_uninitialized as the first value. Do not number values
9316 as they are zero based.
9317 (profile_count::verify): Update sanity check.
9318 (profile_probability::verify): Likewise.
9320 2018-01-23 Nathan Sidwell <nathan@acm.org>
9322 * doc/invoke.texi (ffor-scope): Deprecate.
9324 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9326 PR tree-optimization/83510
9327 * domwalk.c (set_all_edges_as_executable): New function.
9328 (dom_walker::dom_walker): Convert bool param
9329 "skip_unreachable_blocks" to enum reachability. Move setup of
9330 edge flags to set_all_edges_as_executable and only do it when
9331 reachability is REACHABLE_BLOCKS.
9332 * domwalk.h (enum dom_walker::reachability): New enum.
9333 (dom_walker::dom_walker): Convert bool param
9334 "skip_unreachable_blocks" to enum reachability.
9335 (set_all_edges_as_executable): New decl.
9336 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9337 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9339 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9340 but converting true to REACHABLE_BLOCKS.
9341 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9343 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9344 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9345 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9347 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9348 if check_all_array_refs will be called.
9350 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9352 * tree.c (selftest::test_location_wrappers): Add more test
9355 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9357 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9358 (selftest::test_bit_in_range): Likewise.
9360 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9363 * doc/sourcebuild.texi (vect_float): Say that the selector
9364 only describes the situation when -funsafe-math-optimizations is on.
9365 (vect_float_strict): Document.
9367 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9369 PR tree-optimization/83965
9370 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9371 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9372 instead of checking only for a reduction.
9373 (vect_recog_widen_sum_pattern): Likewise.
9375 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9377 * predict.c (probably_never_executed): Only use precise profile info.
9378 (compute_function_frequency): Skip after inlining hack since we now
9379 have quality checking.
9381 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9383 * profile-count.h (profile_probability::very_unlikely,
9384 profile_probability::unlikely, profile_probability::even): Set
9385 precision to guessed.
9387 2018-01-23 Richard Biener <rguenther@suse.de>
9389 PR tree-optimization/83963
9390 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9391 Properly terminate dominator walk when crossing the exit edge not
9392 when visiting its source block.
9394 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9397 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9398 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9400 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9402 PR tree-optimization/83957
9403 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9404 semicolon after for body surrounded by braces.
9406 PR tree-optimization/83081
9407 * profile-count.h (profile_probability::split): New method.
9408 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9409 Use profile_probability::split.
9410 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9411 when splitting a single conditional jump into 2.
9413 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9415 PR tree-optimization/69452
9416 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9419 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9421 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9422 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9423 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9425 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9427 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9428 * config/rl78/rl78.md (movdi): New define_expand.
9429 * config/rl78/rl78.c (rl78_split_movdi): New function.
9431 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9434 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9436 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9437 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9438 128-bit to produce an UNSPEC move to get the double word with the
9439 signbit and then a shift directly to do signbit.
9440 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9441 implementation with a new version that just does either a direct
9442 move or a regular move. Move memory interface to separate insns.
9443 Move insns so they are next to the expander.
9444 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9445 with signbit move. Split big and little endian case.
9446 (signbit<mode>2_dm_mem_le): Likewise.
9447 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9448 (signbit<mode>2_dm2): Likewise.
9450 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9452 * config/rl78/rl78.md (anddi3): New define_expand.
9454 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9456 * config/rl78/rl78.md (umindi3): New define_expand.
9458 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9460 * config/rl78/rl78.md (smindi3): New define_expand.
9462 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9464 * config/rl78/rl78.md (smaxdi3): New define_expand.
9466 2018-01-22 Carl Love <cel@us.ibm.com>
9468 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9469 LVX_V1TI): Add macro expansion.
9470 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9471 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9472 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9473 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9474 Change check to determine if the instruction is a byte reversing
9475 entry. Fix typo in comment.
9476 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9477 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9478 Add def_builtin calls for new builtins.
9479 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9480 Add define_insn expansion.
9482 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9484 * config/rl78/rl78.md (umaxdi3): New define_expand.
9486 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9488 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9489 for non-QImode registers.
9491 2018-01-22 Richard Biener <rguenther@suse.de>
9493 PR tree-optimization/83963
9494 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9495 including the loop exit block.
9496 (scop_detection::merge_sese): Likewise.
9497 (scop_detection::add_scop): Do it here instead.
9499 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9501 * doc/sourcebuild.texi (arm_softfloat): Document.
9503 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9506 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9507 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9508 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9510 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9511 David Edelsohn <dje.gcc@gmail.com>
9514 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9515 Change "crset eq" to "crset 2".
9516 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9517 (*call_indirect_aix<mode>_nospec): Likewise.
9518 (*call_value_indirect_aix<mode>_nospec): Likewise.
9519 (*call_indirect_elfv2<mode>_nospec): Likewise.
9520 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9521 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9522 change assembly output from . to $.
9523 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9524 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9525 (*tablejump<mode>_internal1_nospec): Likewise.
9527 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9530 * config/sh/sh_optimize_sett_clrt.cc:
9531 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9533 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9535 PR tree-optimization/83940
9536 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9537 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9538 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9539 use it to pass back the definition type.
9540 (vect_check_store_rhs): Likewise rhs_dt_out.
9541 (vect_build_gather_load_calls): Add a mask_dt argument and use
9542 it instead of a call to vect_is_simple_use.
9543 (vectorizable_store): Update calls to vect_check_load_store_mask
9544 and vect_check_store_rhs. Use the dt returned by the latter instead
9545 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9546 instead of calls to vect_is_simple_use. Pass the scalar rather
9547 than the vector operand to vect_is_simple_use when handling
9548 second and subsequent copies of an rhs value.
9549 (vectorizable_load): Update calls to vect_check_load_store_mask
9550 and vect_build_gather_load_calls. Use the cached mask_dt and
9551 gs_info.offset_dt instead of calls to vect_is_simple_use.
9553 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9556 * tree-emutls.c: Include gimplify.h.
9557 (lower_emutls_2): New function.
9558 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9559 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9560 it before further processing.
9563 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9564 UINTVAL (trueop1) instead of INTVAL (op1).
9566 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9570 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9571 INCOMING_FRAME_SP_OFFSET if not defined.
9572 (scan_trace): Add ENTRY argument. If true and
9573 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9574 emit a note to adjust the CFA offset.
9575 (create_cfi_notes): Adjust scan_trace callers.
9576 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9577 INCOMING_FRAME_SP_OFFSET in the CIE.
9578 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9579 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9581 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9582 * doc/tm.texi: Regenerated.
9584 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9586 PR rtl-optimization/83147
9587 * lra-constraints.c (remove_inheritance_pseudos): Use
9588 lra_substitute_pseudo_within_insn.
9590 2018-01-19 Tom de Vries <tom@codesourcery.com>
9591 Cesar Philippidis <cesar@codesourcery.com>
9594 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9596 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9599 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9600 spaces for function labels.
9602 2018-01-19 Martin Liska <mliska@suse.cz>
9604 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9605 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9606 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9607 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9608 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9609 (PRED_CONST_RETURN): Change from 69 to 65.
9610 (PRED_NULL_RETURN): Change from 91 to 71.
9611 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9612 (PRED_LOOP_GUARD): Change from 66 to 73.
9614 2018-01-19 Martin Liska <mliska@suse.cz>
9616 * predict.c (predict_insn_def): Add new assert.
9617 (struct branch_predictor): Change type to signed integer.
9618 (test_prediction_value_range): Amend test to cover
9620 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9621 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9622 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9623 (PRED_LOOP_IV_COMPARE): Likewise.
9624 * predict.h (PROB_UNINITIALIZED): Define new constant.
9626 2018-01-19 Martin Liska <mliska@suse.cz>
9628 * predict.c (dump_prediction): Add new format for
9629 analyze_brprob.py script which is enabled with -details
9631 * profile-count.h (precise_p): New function.
9633 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9635 PR tree-optimization/83922
9636 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9637 there are no statements that need masking.
9638 (vect_active_double_reduction_p): New function.
9639 (vect_analyze_loop_operations): Use it when handling phis that
9640 are not in the loop header.
9642 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9644 PR tree-optimization/83914
9645 * tree-vect-loop.c (vectorizable_induction): Don't convert
9646 init_expr or apply the peeling adjustment for inductions
9647 that are nested within the vectorized loop.
9649 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9651 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9654 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9658 * function.h (gimplify_parameters): Add gimple_seq * argument.
9659 * function.c: Include gimple.h and options.h.
9660 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9661 for the added local temporaries if needed.
9662 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9663 if there are any parameter cleanups, wrap whole body into a
9664 try/finally with the cleanups.
9666 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9669 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9670 Use GET_MODE_CLASS for scalar floating point.
9672 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9676 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9677 Fix call of call_cgraph_insertion_hooks.
9679 2018-01-18 Martin Sebor <msebor@redhat.com>
9681 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9683 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9686 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9689 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9692 * common.opt: (-ffile-prefix-map): New option.
9693 * opts.c (common_handle_option): Defer it.
9694 * opts-global.c (handle_common_deferred_options): Handle it.
9695 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9696 * file-prefix-map.h: New file.
9697 (remap_debug_filename, add_debug_prefix_map): ...here.
9698 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9699 * final.c (debug_prefix_map, add_debug_prefix_map
9700 remap_debug_filename): Move to...
9701 * file-prefix-map.c: New file.
9702 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9703 generalize, get rid of alloca(), use strrchr() instead of strchr().
9704 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9705 Implement in terms of add_prefix_map().
9706 (remap_macro_filename, remap_debug_filename): Implement in term of
9708 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9709 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9710 * dbxout.c: Include file-prefix-map.h.
9711 * varasm.c: Likewise.
9712 * vmsdbgout.c: Likewise.
9713 * xcoffout.c: Likewise.
9714 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9715 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9716 * doc/invoke.texi (-ffile-prefix-map): Document.
9717 (-fdebug-prefix-map): Update description.
9719 2018-01-18 Martin Liska <mliska@suse.cz>
9721 * config/i386/i386.c (indirect_thunk_name): Document that also
9723 (output_indirect_thunk): Document why both instructions
9724 (pause and lfence) are generated.
9726 2018-01-18 Richard Biener <rguenther@suse.de>
9728 PR tree-optimization/83887
9729 * graphite-scop-detection.c
9730 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9731 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9732 (scop_detection::merge_sese): Re-implement with a flood-fill
9733 algorithm that properly finds a SESE region if it exists.
9735 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9738 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9739 pointer_diff optimizations use view_convert instead of convert.
9741 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9743 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9744 Generate different code for -mno-speculate-indirect-jumps.
9745 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9746 (*call_indirect_aix<mode>): Disable for
9747 -mno-speculate-indirect-jumps.
9748 (*call_indirect_aix<mode>_nospec): New define_insn.
9749 (*call_value_indirect_aix<mode>): Disable for
9750 -mno-speculate-indirect-jumps.
9751 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9752 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9753 -mno-speculate-indirect-jumps.
9754 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9756 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9758 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9759 long double type, set the flags for noting the default long double
9760 type, even if we don't pass or return a long double type.
9762 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9765 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9768 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9770 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9771 support for merge[hl].
9772 (fold_mergehl_helper): New helper function.
9773 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9774 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9775 (altivec_vmrglw_direct): Add xxmrglw insn.
9777 2018-01-17 Andrew Waterman <andrew@sifive.com>
9779 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9780 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9782 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9785 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9786 call the lto_location_cache before reading the
9787 DECL_SOURCE_LOCATION of the types.
9789 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9790 Richard Sandiford <richard.sandiford@linaro.org>
9792 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9793 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9794 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9795 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9796 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9798 * config/aarch64/constraints.md (aarch64_movti_operand):
9800 * config/aarch64/predicates.md (Uti): Add new constraint.
9802 2018-01-17 Carl Love <cel@us.ibm.com>
9804 * config/rs6000/vsx.md (define_expand xl_len_r,
9805 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9806 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9808 (define_expand, define_insn): Move the shift left from the
9809 define_insn to the define_expand for lxvl and stxvl instructions.
9810 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9811 and XL_LEN_R definitions to PURE.
9813 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9815 * config/i386/i386.c (indirect_thunk_name): Declare regno
9816 as unsigned int. Compare regno with INVALID_REGNUM.
9817 (output_indirect_thunk): Ditto.
9818 (output_indirect_thunk_function): Ditto.
9819 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9820 in the call to output_indirect_thunk_function.
9822 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9825 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9826 rather than the size of inner_type to determine the stack slot size
9827 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9829 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9832 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9835 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9837 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9838 endian Linux systems to optionally enable multilibs for selecting
9839 the long double type if the user configured an explicit type.
9840 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9841 have no long double multilibs if not defined.
9842 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9843 warn if the user used -mabi={ieee,ibm}longdouble and we built
9844 multilibs for long double.
9845 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9846 appropriate multilib option.
9847 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9849 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9850 for building long double multilibs.
9851 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9853 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9855 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9858 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9860 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9863 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9866 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9869 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9871 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9872 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9874 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9876 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9877 different rtl trees depending on TARGET_64BIT.
9878 (rs6000_gen_lvx): Likewise.
9880 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9882 * config/visium/visium.md (nop): Tweak comment.
9883 (hazard_nop): Likewise.
9885 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9887 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9888 -mspeculate-indirect-jumps.
9889 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9890 for -mno-speculate-indirect-jumps.
9891 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9892 (*call_value_indirect_elfv2<mode>): Disable for
9893 -mno-speculate-indirect-jumps.
9894 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9895 (indirect_jump): Emit different RTL for
9896 -mno-speculate-indirect-jumps.
9897 (*indirect_jump<mode>): Disable for
9898 -mno-speculate-indirect-jumps.
9899 (*indirect_jump<mode>_nospec): New define_insn.
9900 (tablejump): Emit different RTL for
9901 -mno-speculate-indirect-jumps.
9902 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9903 (tablejumpsi_nospec): New define_expand.
9904 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9905 (tablejumpdi_nospec): New define_expand.
9906 (*tablejump<mode>_internal1): Disable for
9907 -mno-speculate-indirect-jumps.
9908 (*tablejump<mode>_internal1_nospec): New define_insn.
9909 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9912 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9914 * caller-save.c (insert_save): Drop unnecessary parameter. All
9917 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9918 Richard Biener <rguenth@suse.de>
9921 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9922 return early, inline manually is_gimple_sizepos. Make sure if we
9923 call gimplify_expr we don't end up with a gimple constant.
9924 * tree.c (variably_modified_type_p): Don't return true for
9925 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9926 * gimplify.h (is_gimple_sizepos): Remove.
9928 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9930 PR tree-optimization/83857
9931 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9932 vectorizable_live_operation for pure SLP statements.
9933 (vectorizable_live_operation): Handle PHIs.
9935 2018-01-16 Richard Biener <rguenther@suse.de>
9937 PR tree-optimization/83867
9938 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9939 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9941 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9944 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9945 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9946 If off is not INTEGER_CST, issue a may not be aligned warning
9947 rather than isn't aligned. Use isn%'t rather than isn't.
9948 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9950 <case MULT_EXPR>: Improve the case when bottom and one of the
9951 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9952 operand, in that case check if the other operand is multiple of
9953 bottom divided by the INTEGER_CST operand.
9955 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9958 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9959 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9960 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9961 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9962 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9963 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9964 (pa_function_arg_size): New function.
9966 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9968 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9969 in a separate statement.
9971 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9973 PR tree-optimization/83847
9974 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9975 group gathers and scatters.
9977 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9979 PR rtl-optimization/86620
9980 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9982 PR rtl-optimization/83213
9983 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9984 to last if both are JUMP_INSNs.
9986 PR tree-optimization/83843
9987 * gimple-ssa-store-merging.c
9988 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9989 store_immediate_info for bswap/nop orig_stores.
9991 2018-01-15 Andrew Waterman <andrew@sifive.com>
9993 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9995 <UDIV>: Increase cost if !TARGET_DIV.
9997 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9999 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10000 (define_attr "cr_logical_3op"): New.
10001 (cceq_ior_compare): Adjust.
10002 (cceq_ior_compare_complement): Adjust.
10003 (*cceq_rev_compare): Adjust.
10004 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10005 (is_cracked_insn): Adjust.
10006 (insn_must_be_first_in_group): Adjust.
10007 * config/rs6000/40x.md: Adjust.
10008 * config/rs6000/440.md: Adjust.
10009 * config/rs6000/476.md: Adjust.
10010 * config/rs6000/601.md: Adjust.
10011 * config/rs6000/603.md: Adjust.
10012 * config/rs6000/6xx.md: Adjust.
10013 * config/rs6000/7450.md: Adjust.
10014 * config/rs6000/7xx.md: Adjust.
10015 * config/rs6000/8540.md: Adjust.
10016 * config/rs6000/cell.md: Adjust.
10017 * config/rs6000/e300c2c3.md: Adjust.
10018 * config/rs6000/e500mc.md: Adjust.
10019 * config/rs6000/e500mc64.md: Adjust.
10020 * config/rs6000/e5500.md: Adjust.
10021 * config/rs6000/e6500.md: Adjust.
10022 * config/rs6000/mpc.md: Adjust.
10023 * config/rs6000/power4.md: Adjust.
10024 * config/rs6000/power5.md: Adjust.
10025 * config/rs6000/power6.md: Adjust.
10026 * config/rs6000/power7.md: Adjust.
10027 * config/rs6000/power8.md: Adjust.
10028 * config/rs6000/power9.md: Adjust.
10029 * config/rs6000/rs64.md: Adjust.
10030 * config/rs6000/titan.md: Adjust.
10032 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10034 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10035 ix86_indirect_branch_register logic.
10037 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10039 * config/i386/constraints.md (Bs): Update
10040 ix86_indirect_branch_register check. Don't check
10041 ix86_indirect_branch_register with GOT_memory_operand.
10043 * config/i386/predicates.md (GOT_memory_operand): Don't check
10044 ix86_indirect_branch_register here.
10045 (GOT32_symbol_operand): Likewise.
10047 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10049 * config/i386/predicates.md (constant_call_address_operand):
10050 Rewrite ix86_indirect_branch_register logic.
10051 (sibcall_insn_operand): Likewise.
10053 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10055 * config/i386/constraints.md (Bs): Replace
10056 ix86_indirect_branch_thunk_register with
10057 ix86_indirect_branch_register.
10059 * config/i386/i386.md (indirect_jump): Likewise.
10060 (tablejump): Likewise.
10061 (*sibcall_memory): Likewise.
10062 (*sibcall_value_memory): Likewise.
10063 Peepholes of indirect call and jump via memory: Likewise.
10064 * config/i386/i386.opt: Likewise.
10065 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10066 (GOT_memory_operand): Likewise.
10067 (call_insn_operand): Likewise.
10068 (sibcall_insn_operand): Likewise.
10069 (GOT32_symbol_operand): Likewise.
10071 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10073 PR middle-end/83837
10074 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10075 type rather than type addr's type points to.
10076 (expand_omp_atomic_mutex): Likewise.
10077 (expand_omp_atomic): Likewise.
10079 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10082 * config/i386/i386.c (output_indirect_thunk_function): Use
10083 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10084 for __x86_return_thunk.
10086 2018-01-15 Richard Biener <rguenther@suse.de>
10088 PR middle-end/83850
10089 * expmed.c (extract_bit_field_1): Fix typo.
10091 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10094 * config/arm/iterators.md (VF): New mode iterator.
10095 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10096 Remove integer-related logic from pattern.
10097 (neon_vabd<mode>_3): Likewise.
10099 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10101 PR middle-end/82694
10102 * common.opt (fstrict-overflow): No longer an alias.
10103 (fwrapv-pointer): New option.
10104 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10105 also for pointer types based on flag_wrapv_pointer.
10106 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10107 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10108 opts->x_flag_wrapv got set.
10109 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10110 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10111 POINTER_TYPE_OVERFLOW_UNDEFINED.
10112 * match.pd: Likewise in address comparison pattern.
10113 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10115 2018-01-15 Richard Biener <rguenther@suse.de>
10118 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10119 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10120 Reset type names to their identifier if their TYPE_DECL doesn't
10121 have linkage (and thus is used for ODR and devirt).
10122 (save_debug_info_for_decl): Remove.
10123 (save_debug_info_for_type): Likewise.
10124 (add_tree_to_fld_list): Adjust.
10125 * tree-pretty-print.c (dump_generic_node): Make dumping of
10126 type names more robust.
10128 2018-01-15 Richard Biener <rguenther@suse.de>
10130 * BASE-VER: Bump to 8.0.1.
10132 2018-01-14 Martin Sebor <msebor@redhat.com>
10135 * builtins.c (check_access): Avoid warning when the no-warning bit
10138 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10140 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10141 * ira-color (allocno_hard_regs_compare): Likewise.
10143 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10146 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10147 Use .pushsection/.popsection.
10149 2018-01-14 Martin Sebor <msebor@redhat.com>
10152 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10154 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10156 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10157 entry from extra_headers.
10158 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10159 extra_headers, make the list bitwise identical to the i?86-*-* one.
10161 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10163 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10164 -mcmodel=large with -mindirect-branch=thunk,
10165 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10166 -mfunction-return=thunk-extern.
10167 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10168 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10169 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10171 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10173 * config/i386/i386.c (print_reg): Print the name of the full
10174 integer register without '%'.
10175 (ix86_print_operand): Handle 'V'.
10176 * doc/extend.texi: Document 'V' modifier.
10178 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10180 * config/i386/constraints.md (Bs): Disallow memory operand for
10181 -mindirect-branch-register.
10183 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10184 (GOT_memory_operand): Likewise.
10185 (call_insn_operand): Likewise.
10186 (sibcall_insn_operand): Likewise.
10187 (GOT32_symbol_operand): Likewise.
10188 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10189 for -mindirect-branch-register.
10190 (tablejump): Likewise.
10191 (*sibcall_memory): Likewise.
10192 (*sibcall_value_memory): Likewise.
10193 Disallow peepholes of indirect call and jump via memory for
10194 -mindirect-branch-register.
10195 (*call_pop): Replace m with Bw.
10196 (*call_value_pop): Likewise.
10197 (*sibcall_pop_memory): Replace m with Bs.
10198 * config/i386/i386.opt (mindirect-branch-register): New option.
10199 * doc/invoke.texi: Document -mindirect-branch-register option.
10201 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10203 * config/i386/i386-protos.h (ix86_output_function_return): New.
10204 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10205 set function_return_type.
10206 (indirect_thunk_name): Add ret_p to indicate thunk for function
10208 (output_indirect_thunk_function): Pass false to
10209 indirect_thunk_name.
10210 (ix86_output_indirect_branch_via_reg): Likewise.
10211 (ix86_output_indirect_branch_via_push): Likewise.
10212 (output_indirect_thunk_function): Create alias for function
10213 return thunk if regno < 0.
10214 (ix86_output_function_return): New function.
10215 (ix86_handle_fndecl_attribute): Handle function_return.
10216 (ix86_attribute_table): Add function_return.
10217 * config/i386/i386.h (machine_function): Add
10218 function_return_type.
10219 * config/i386/i386.md (simple_return_internal): Use
10220 ix86_output_function_return.
10221 (simple_return_internal_long): Likewise.
10222 * config/i386/i386.opt (mfunction-return=): New option.
10223 (indirect_branch): Mention -mfunction-return=.
10224 * doc/extend.texi: Document function_return function attribute.
10225 * doc/invoke.texi: Document -mfunction-return= option.
10227 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10229 * config/i386/i386-opts.h (indirect_branch): New.
10230 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10231 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10232 with local indirect jump when converting indirect call and jump.
10233 (ix86_set_indirect_branch_type): New.
10234 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10235 (indirectlabelno): New.
10236 (indirect_thunk_needed): Likewise.
10237 (indirect_thunk_bnd_needed): Likewise.
10238 (indirect_thunks_used): Likewise.
10239 (indirect_thunks_bnd_used): Likewise.
10240 (INDIRECT_LABEL): Likewise.
10241 (indirect_thunk_name): Likewise.
10242 (output_indirect_thunk): Likewise.
10243 (output_indirect_thunk_function): Likewise.
10244 (ix86_output_indirect_branch_via_reg): Likewise.
10245 (ix86_output_indirect_branch_via_push): Likewise.
10246 (ix86_output_indirect_branch): Likewise.
10247 (ix86_output_indirect_jmp): Likewise.
10248 (ix86_code_end): Call output_indirect_thunk_function if needed.
10249 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10251 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10252 (ix86_attribute_table): Add indirect_branch.
10253 * config/i386/i386.h (machine_function): Add indirect_branch_type
10254 and has_local_indirect_jump.
10255 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10257 (tablejump): Likewise.
10258 (*indirect_jump): Use ix86_output_indirect_jmp.
10259 (*tablejump_1): Likewise.
10260 (simple_return_indirect_internal): Likewise.
10261 * config/i386/i386.opt (mindirect-branch=): New option.
10262 (indirect_branch): New.
10265 (thunk-inline): Likewise.
10266 (thunk-extern): Likewise.
10267 * doc/extend.texi: Document indirect_branch function attribute.
10268 * doc/invoke.texi: Document -mindirect-branch= option.
10270 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10273 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10275 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10277 * ipa-inline.c (want_inline_small_function_p): Return false if
10278 inlining has already failed with CIF_FINAL_ERROR.
10279 (update_caller_keys): Call want_inline_small_function_p before
10281 (update_callee_keys): Likewise.
10283 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10285 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10287 (rs6000_quadword_masked_address_p): Likewise.
10288 (quad_aligned_load_p): Likewise.
10289 (quad_aligned_store_p): Likewise.
10290 (const_load_sequence_p): Add comment to describe the outer-most loop.
10291 (mimic_memory_attributes_and_flags): New function.
10292 (rs6000_gen_stvx): Likewise.
10293 (replace_swapped_aligned_store): Likewise.
10294 (rs6000_gen_lvx): Likewise.
10295 (replace_swapped_aligned_load): Likewise.
10296 (replace_swapped_load_constant): Capitalize argument name in
10297 comment describing this function.
10298 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10299 and stores that access quad-word aligned addresses and replace
10300 with stvx or lvx instructions when appropriate.
10301 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10302 New function prototype.
10303 (rs6000_quadword_masked_address_p): Likewise.
10304 (rs6000_gen_lvx): Likewise.
10305 (rs6000_gen_stvx): Likewise.
10306 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10307 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10308 when memory address is aligned.
10309 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10310 this split to select lvx instruction when memory address is aligned.
10311 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10312 instruction when memory address is aligned.
10313 (*vsx_le_perm_load_v16qi): Likewise.
10314 (four unnamed splitters): Modify to select the stvx instruction
10315 when memory is aligned.
10317 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10319 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10320 which appears in the queue multiple times.
10322 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10323 Alan Hayward <alan.hayward@arm.com>
10324 David Sherwood <david.sherwood@arm.com>
10326 * tree-vectorizer.h (vec_lower_bound): New structure.
10327 (_loop_vec_info): Add check_nonzero and lower_bounds.
10328 (LOOP_VINFO_CHECK_NONZERO): New macro.
10329 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10330 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10331 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10332 fields. Make seg_len the distance travelled, not including the
10334 (dr_direction_indicator): Declare.
10335 (dr_zero_step_indicator): Likewise.
10336 (dr_known_forward_stride_p): Likewise.
10337 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10339 (runtime_alias_check_p): Allow runtime alias checks with
10341 (operator ==): Compare access_size and align.
10342 (prune_runtime_alias_test_list): Rework for new distinction between
10343 the access_size and seg_len.
10344 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10346 (get_segment_min_max): New function.
10347 (create_intersect_range_checks): Use it.
10348 (dr_step_indicator): New function.
10349 (dr_direction_indicator): Likewise.
10350 (dr_zero_step_indicator): Likewise.
10351 (dr_known_forward_stride_p): Likewise.
10352 * tree-loop-distribution.c (data_ref_segment_size): Return
10353 DR_STEP * (niters - 1).
10354 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10356 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10357 (vect_preserves_scalar_order_p): New function, split out from...
10358 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10359 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10360 (vect_vfa_access_size): New function.
10361 (vect_vfa_align): Likewise.
10362 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10363 (dump_lower_bound): New function.
10364 (vect_check_lower_bound): Likewise.
10365 (vect_small_gap_p): Likewise.
10366 (vectorizable_with_step_bound_p): Likewise.
10367 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10368 depencies if the vectorization factor is 1. Convert the checks
10369 for nonzero steps into checks on the bounds of DR_STEP. Try using
10370 a bunds check for variable steps if the minimum required step is
10371 relatively small. Update calls to the dr_with_seg_len
10372 constructor and to vect_compile_time_alias.
10373 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10375 (vect_loop_versioning): Call it.
10376 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10378 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10380 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10381 Alan Hayward <alan.hayward@arm.com>
10382 David Sherwood <david.sherwood@arm.com>
10384 * doc/sourcebuild.texi (vect_scatter_store): Document.
10385 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10387 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10389 * genopinit.c (main): Add supports_vec_scatter_store and
10390 supports_vec_scatter_store_cached to target_optabs.
10391 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10392 IFN_MASK_SCATTER_STORE.
10393 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10395 * internal-fn.h (internal_store_fn_p): Declare.
10396 (internal_fn_stored_value_index): Likewise.
10397 * internal-fn.c (scatter_store_direct): New macro.
10398 (expand_scatter_store_optab_fn): New function.
10399 (direct_scatter_store_optab_supported_p): New macro.
10400 (internal_store_fn_p): New function.
10401 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10402 IFN_MASK_SCATTER_STORE.
10403 (internal_fn_mask_index): Likewise.
10404 (internal_fn_stored_value_index): New function.
10405 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10406 for scatter stores.
10407 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10408 * optabs-query.c (supports_vec_scatter_store_p): New function.
10409 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10410 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10411 true for scatter stores.
10412 (vect_gather_scatter_fn_p): Handle scatter stores too.
10413 (vect_check_gather_scatter): Consider using scatter stores if
10414 supports_vec_scatter_store_p.
10415 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10416 scatter stores too.
10417 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10418 internal_fn_stored_value_index.
10419 (check_load_store_masking): Handle scatter stores too.
10420 (vect_get_store_rhs): Make public.
10421 (vectorizable_call): Use internal_store_fn_p.
10422 (vectorizable_store): Handle scatter store internal functions.
10423 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10424 when deciding whether the end of the group has been reached.
10425 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10426 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10427 (mask_scatter_store<mode>): New insns.
10429 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10430 Alan Hayward <alan.hayward@arm.com>
10431 David Sherwood <david.sherwood@arm.com>
10433 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10434 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10435 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10437 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10438 Use vect_truncate_gather_scatter_offset if we can't treat the
10439 operation as a normal gather load or scatter store.
10440 (get_group_load_store_type): Take the gather_scatter_info
10441 as argument. Try using a gather load or scatter store for
10442 single-element groups.
10443 (get_load_store_type): Update calls to get_group_load_store_type
10444 and vect_use_strided_gather_scatters_p.
10446 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10447 Alan Hayward <alan.hayward@arm.com>
10448 David Sherwood <david.sherwood@arm.com>
10450 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10451 optional tree argument.
10452 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10454 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10455 but continue to use the current value as a fallback.
10456 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10457 to compare the updates.
10458 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10459 (get_load_store_type): Use it when handling a strided access.
10460 (vect_get_strided_load_store_ops): New function.
10461 (vect_get_data_ptr_increment): Likewise.
10462 (vectorizable_load): Handle strided gather loads. Always pass
10463 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10465 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10466 Alan Hayward <alan.hayward@arm.com>
10467 David Sherwood <david.sherwood@arm.com>
10469 * doc/md.texi (gather_load@var{m}): Document.
10470 (mask_gather_load@var{m}): Likewise.
10471 * genopinit.c (main): Add supports_vec_gather_load and
10472 supports_vec_gather_load_cached to target_optabs.
10473 * optabs-tree.c (init_tree_optimization_optabs): Use
10474 ggc_cleared_alloc to allocate target_optabs.
10475 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10476 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10478 * internal-fn.h (internal_load_fn_p): Declare.
10479 (internal_gather_scatter_fn_p): Likewise.
10480 (internal_fn_mask_index): Likewise.
10481 (internal_gather_scatter_fn_supported_p): Likewise.
10482 * internal-fn.c (gather_load_direct): New macro.
10483 (expand_gather_load_optab_fn): New function.
10484 (direct_gather_load_optab_supported_p): New macro.
10485 (direct_internal_fn_optab): New function.
10486 (internal_load_fn_p): Likewise.
10487 (internal_gather_scatter_fn_p): Likewise.
10488 (internal_fn_mask_index): Likewise.
10489 (internal_gather_scatter_fn_supported_p): Likewise.
10490 * optabs-query.c (supports_at_least_one_mode_p): New function.
10491 (supports_vec_gather_load_p): Likewise.
10492 * optabs-query.h (supports_vec_gather_load_p): Declare.
10493 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10494 and memory_type field.
10495 (NUM_PATTERNS): Bump to 15.
10496 * tree-vect-data-refs.c: Include internal-fn.h.
10497 (vect_gather_scatter_fn_p): New function.
10498 (vect_describe_gather_scatter_call): Likewise.
10499 (vect_check_gather_scatter): Try using internal functions for
10500 gather loads. Recognize existing calls to a gather load function.
10501 (vect_analyze_data_refs): Consider using gather loads if
10502 supports_vec_gather_load_p.
10503 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10504 (vect_get_gather_scatter_offset_type): Likewise.
10505 (vect_convert_mask_for_vectype): Likewise.
10506 (vect_add_conversion_to_patterm): Likewise.
10507 (vect_try_gather_scatter_pattern): Likewise.
10508 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10509 (vect_vect_recog_func_ptrs): Add it.
10510 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10511 internal_fn_mask_index and internal_gather_scatter_fn_p.
10512 (check_load_store_masking): Take the gather_scatter_info as an
10513 argument and handle gather loads.
10514 (vect_get_gather_scatter_ops): New function.
10515 (vectorizable_call): Check internal_load_fn_p.
10516 (vectorizable_load): Likewise. Handle gather load internal
10518 (vectorizable_store): Update call to check_load_store_masking.
10519 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10520 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10521 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10522 (aarch64_gather_scale_operand_d): New predicates.
10523 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10524 (mask_gather_load<mode>): New insns.
10526 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10527 Alan Hayward <alan.hayward@arm.com>
10528 David Sherwood <david.sherwood@arm.com>
10530 * optabs.def (fold_left_plus_optab): New optab.
10531 * doc/md.texi (fold_left_plus_@var{m}): Document.
10532 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10533 * internal-fn.c (fold_left_direct): Define.
10534 (expand_fold_left_optab_fn): Likewise.
10535 (direct_fold_left_optab_supported_p): Likewise.
10536 * fold-const-call.c (fold_const_fold_left): New function.
10537 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10538 * tree-parloops.c (valid_reduction_p): New function.
10539 (gather_scalar_reductions): Use it.
10540 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10541 (vect_finish_replace_stmt): Declare.
10542 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10543 (needs_fold_left_reduction_p): New function, split out from...
10544 (vect_is_simple_reduction): ...here. Accept reductions that
10545 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10546 (vect_force_simple_reduction): Also store the reduction type in
10547 the assignment's STMT_VINFO_REDUC_TYPE.
10548 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10549 (merge_with_identity): New function.
10550 (vect_expand_fold_left): Likewise.
10551 (vectorize_fold_left_reduction): Likewise.
10552 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10553 scalar phi in place for it. Check for target support and reject
10554 cases that would reassociate the operation. Defer the transform
10555 phase to vectorize_fold_left_reduction.
10556 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10557 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10558 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10560 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10562 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10563 call to ifc_temp_var.
10565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10566 Alan Hayward <alan.hayward@arm.com>
10567 David Sherwood <david.sherwood@arm.com>
10569 * target.def (legitimize_address_displacement): Take the original
10570 offset as a poly_int.
10571 * targhooks.h (default_legitimize_address_displacement): Update
10573 * targhooks.c (default_legitimize_address_displacement): Likewise.
10574 * doc/tm.texi: Regenerate.
10575 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10576 as an argument, moving assert of ad->disp == ad->disp_term to...
10577 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10578 Try calling targetm.legitimize_address_displacement before expanding
10579 the address rather than afterwards, and adjust for the new interface.
10580 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10581 Match the new hook interface. Handle SVE addresses.
10582 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10583 new hook interface.
10585 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10587 * Makefile.in (OBJS): Add early-remat.o.
10588 * target.def (select_early_remat_modes): New hook.
10589 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10590 * doc/tm.texi: Regenerate.
10591 * targhooks.h (default_select_early_remat_modes): Declare.
10592 * targhooks.c (default_select_early_remat_modes): New function.
10593 * timevar.def (TV_EARLY_REMAT): New timevar.
10594 * passes.def (pass_early_remat): New pass.
10595 * tree-pass.h (make_pass_early_remat): Declare.
10596 * early-remat.c: New file.
10597 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10599 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10601 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10602 Alan Hayward <alan.hayward@arm.com>
10603 David Sherwood <david.sherwood@arm.com>
10605 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10606 vfm1 with a bound_epilog parameter.
10607 (vect_do_peeling): Update calls accordingly, and move the prologue
10608 call earlier in the function. Treat the base bound_epilog as 0 for
10609 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10610 this base when peeling for gaps.
10611 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10612 with fully-masked loops.
10613 (vect_estimate_min_profitable_iters): Handle the single peeled
10614 iteration in that case.
10616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10617 Alan Hayward <alan.hayward@arm.com>
10618 David Sherwood <david.sherwood@arm.com>
10620 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10621 single-element interleaving even if the size is not a power of 2.
10622 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10623 accesses for single-element interleaving if the group size is
10626 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10627 Alan Hayward <alan.hayward@arm.com>
10628 David Sherwood <david.sherwood@arm.com>
10630 * doc/md.texi (fold_extract_last_@var{m}): Document.
10631 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10632 * optabs.def (fold_extract_last_optab): New optab.
10633 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10634 * internal-fn.c (fold_extract_direct): New macro.
10635 (expand_fold_extract_optab_fn): Likewise.
10636 (direct_fold_extract_optab_supported_p): Likewise.
10637 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10638 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10639 EXTRACT_LAST_REDUCTION.
10640 (get_initial_def_for_reduction): Do not create an initial vector
10641 for EXTRACT_LAST_REDUCTION reductions.
10642 (vectorizable_reduction): Leave the scalar phi in place for
10643 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10644 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10645 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10646 transform phase to vectorizable_condition.
10647 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10649 (vect_finish_stmt_generation): ...here.
10650 (vect_finish_replace_stmt): New function.
10651 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10652 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10654 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10656 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10657 Alan Hayward <alan.hayward@arm.com>
10658 David Sherwood <david.sherwood@arm.com>
10660 * doc/md.texi (extract_last_@var{m}): Document.
10661 * optabs.def (extract_last_optab): New optab.
10662 * internal-fn.def (EXTRACT_LAST): New internal function.
10663 * internal-fn.c (cond_unary_direct): New macro.
10664 (expand_cond_unary_optab_fn): Likewise.
10665 (direct_cond_unary_optab_supported_p): Likewise.
10666 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10667 loops using EXTRACT_LAST.
10668 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10669 (extract_last_<mode>): ...this optab.
10670 (vec_extract<mode><Vel>): Update accordingly.
10672 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10673 Alan Hayward <alan.hayward@arm.com>
10674 David Sherwood <david.sherwood@arm.com>
10676 * target.def (empty_mask_is_expensive): New hook.
10677 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10678 * doc/tm.texi: Regenerate.
10679 * targhooks.h (default_empty_mask_is_expensive): Declare.
10680 * targhooks.c (default_empty_mask_is_expensive): New function.
10681 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10682 if the target says that empty masks are expensive.
10683 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10685 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10687 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10688 Alan Hayward <alan.hayward@arm.com>
10689 David Sherwood <david.sherwood@arm.com>
10691 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10692 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10693 (vect_use_loop_mask_for_alignment_p): New function.
10694 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10695 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10696 niters_skip argument. Make sure that the first niters_skip elements
10697 of the first iteration are inactive.
10698 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10699 Update call to vect_set_loop_masks_directly.
10700 (get_misalign_in_elems): New function, split out from...
10701 (vect_gen_prolog_loop_niters): ...here.
10702 (vect_update_init_of_dr): Take a code argument that specifies whether
10703 the adjustment should be added or subtracted.
10704 (vect_update_init_of_drs): Likewise.
10705 (vect_prepare_for_masked_peels): New function.
10706 (vect_do_peeling): Skip prologue peeling if we're using a mask
10707 instead. Update call to vect_update_inits_of_drs.
10708 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10710 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10711 alignment. Do not include the number of peeled iterations in
10712 the minimum threshold in that case.
10713 (vectorizable_induction): Adjust the start value down by
10714 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10715 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10716 Take the number of skipped iterations into account when calculating
10718 * tree-vect-stmts.c (vect_gen_while_not): New function.
10720 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10721 Alan Hayward <alan.hayward@arm.com>
10722 David Sherwood <david.sherwood@arm.com>
10724 * doc/sourcebuild.texi (vect_fully_masked): Document.
10725 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10726 default value to 0.
10727 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10729 (vect_analyze_loop_2): ...here. Don't check the vectorization
10730 factor against the number of loop iterations if the loop is
10733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10734 Alan Hayward <alan.hayward@arm.com>
10735 David Sherwood <david.sherwood@arm.com>
10737 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10738 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10739 (dump_groups): Update accordingly.
10740 (iv_use::mem_type): New member variable.
10741 (address_p): New function.
10742 (record_use): Add a mem_type argument and initialize the new
10744 (record_group_use): Add a mem_type argument. Use address_p.
10745 Remove obsolete null checks of base_object. Update call to record_use.
10746 (find_interesting_uses_op): Update call to record_group_use.
10747 (find_interesting_uses_cond): Likewise.
10748 (find_interesting_uses_address): Likewise.
10749 (get_mem_type_for_internal_fn): New function.
10750 (find_address_like_use): Likewise.
10751 (find_interesting_uses_stmt): Try find_address_like_use before
10752 calling find_interesting_uses_op.
10753 (addr_offset_valid_p): Use the iv mem_type field as the type
10754 of the addressed memory.
10755 (add_autoinc_candidates): Likewise.
10756 (get_address_cost): Likewise.
10757 (split_small_address_groups_p): Use address_p.
10758 (split_address_groups): Likewise.
10759 (add_iv_candidate_for_use): Likewise.
10760 (autoinc_possible_for_pair): Likewise.
10761 (rewrite_groups): Likewise.
10762 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10763 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10764 (get_alias_ptr_type_for_ptr_address): New function.
10765 (rewrite_use_address): Rewrite address uses in calls that were
10766 identified by find_address_like_use.
10768 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10769 Alan Hayward <alan.hayward@arm.com>
10770 David Sherwood <david.sherwood@arm.com>
10772 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10774 * gimple-expr.h (is_gimple_addressable: Likewise.
10775 * gimple-expr.c (is_gimple_address): Likewise.
10776 * internal-fn.c (expand_call_mem_ref): New function.
10777 (expand_mask_load_optab_fn): Use it.
10778 (expand_mask_store_optab_fn): Likewise.
10780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10781 Alan Hayward <alan.hayward@arm.com>
10782 David Sherwood <david.sherwood@arm.com>
10784 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10785 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10786 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10787 (cond_umax@var{mode}): Document.
10788 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10789 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10790 (cond_umin_optab, cond_umax_optab): New optabs.
10791 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10792 (COND_IOR, COND_XOR): New internal functions.
10793 * internal-fn.h (get_conditional_internal_fn): Declare.
10794 * internal-fn.c (cond_binary_direct): New macro.
10795 (expand_cond_binary_optab_fn): Likewise.
10796 (direct_cond_binary_optab_supported_p): Likewise.
10797 (get_conditional_internal_fn): New function.
10798 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10799 Cope with reduction statements that are vectorized as calls rather
10801 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10802 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10803 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10804 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10805 (UNSPEC_COND_EOR): New unspecs.
10806 (optab): Add mappings for them.
10807 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10808 (sve_int_op, sve_fp_op): New int attributes.
10810 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10811 Alan Hayward <alan.hayward@arm.com>
10812 David Sherwood <david.sherwood@arm.com>
10814 * optabs.def (while_ult_optab): New optab.
10815 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10816 * internal-fn.def (WHILE_ULT): New internal function.
10817 * internal-fn.h (direct_internal_fn_supported_p): New override
10818 that takes two types as argument.
10819 * internal-fn.c (while_direct): New macro.
10820 (expand_while_optab_fn): New function.
10821 (convert_optab_supported_p): Likewise.
10822 (direct_while_optab_supported_p): New macro.
10823 * wide-int.h (wi::udiv_ceil): New function.
10824 * tree-vectorizer.h (rgroup_masks): New structure.
10825 (vec_loop_masks): New typedef.
10826 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10827 and fully_masked_p.
10828 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10829 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10830 (vect_max_vf): New function.
10831 (slpeel_make_loop_iterate_ntimes): Delete.
10832 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10833 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10834 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10835 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10836 internal-fn.h, stor-layout.h and optabs-query.h.
10837 (vect_set_loop_mask): New function.
10838 (add_preheader_seq): Likewise.
10839 (add_header_seq): Likewise.
10840 (interleave_supported_p): Likewise.
10841 (vect_maybe_permute_loop_masks): Likewise.
10842 (vect_set_loop_masks_directly): Likewise.
10843 (vect_set_loop_condition_masked): Likewise.
10844 (vect_set_loop_condition_unmasked): New function, split out from
10845 slpeel_make_loop_iterate_ntimes.
10846 (slpeel_make_loop_iterate_ntimes): Rename to..
10847 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10848 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10849 (vect_do_peeling): Update call accordingly.
10850 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10852 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10853 mask_compare_type, can_fully_mask_p and fully_masked_p.
10854 (release_vec_loop_masks): New function.
10855 (_loop_vec_info): Use it to free the loop masks.
10856 (can_produce_all_loop_masks_p): New function.
10857 (vect_get_max_nscalars_per_iter): Likewise.
10858 (vect_verify_full_masking): Likewise.
10859 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10860 retries, and free the mask rgroups before retrying. Check loop-wide
10861 reasons for disallowing fully-masked loops. Make the final decision
10862 about whether use a fully-masked loop or not.
10863 (vect_estimate_min_profitable_iters): Do not assume that peeling
10864 for the number of iterations will be needed for fully-masked loops.
10865 (vectorizable_reduction): Disable fully-masked loops.
10866 (vectorizable_live_operation): Likewise.
10867 (vect_halve_mask_nunits): New function.
10868 (vect_double_mask_nunits): Likewise.
10869 (vect_record_loop_mask): Likewise.
10870 (vect_get_loop_mask): Likewise.
10871 (vect_transform_loop): Handle the case in which the final loop
10872 iteration might handle a partial vector. Call vect_set_loop_condition
10873 instead of slpeel_make_loop_iterate_ntimes.
10874 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10875 (check_load_store_masking): New function.
10876 (prepare_load_store_mask): Likewise.
10877 (vectorizable_store): Handle fully-masked loops.
10878 (vectorizable_load): Likewise.
10879 (supportable_widening_operation): Use vect_halve_mask_nunits for
10881 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10882 (vect_gen_while): New function.
10883 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10884 (aarch64_uqdec<mode>): New insn.
10886 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10887 Alan Hayward <alan.hayward@arm.com>
10888 David Sherwood <david.sherwood@arm.com>
10890 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10891 (reduc_xor_scal_optab): New optabs.
10892 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10893 (reduc_xor_scal_@var{m}): Document.
10894 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10895 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10896 internal functions.
10897 * fold-const-call.c (fold_const_call): Handle them.
10898 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10899 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10900 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10901 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10902 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10903 (UNSPEC_XORV): New unspecs.
10904 (optab): Add entries for them.
10905 (BITWISEV): New int iterator.
10906 (bit_reduc_op): New int attributes.
10908 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10909 Alan Hayward <alan.hayward@arm.com>
10910 David Sherwood <david.sherwood@arm.com>
10912 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10913 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10914 * optabs.def (vec_shl_insert_optab): New optab.
10915 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10916 (duplicate_and_interleave): Likewise.
10917 * tree-vect-loop.c: Include internal-fn.h.
10918 (neutral_op_for_slp_reduction): New function, split out from
10919 get_initial_defs_for_reduction.
10920 (get_initial_def_for_reduction): Handle option 2 for variable-length
10921 vectors by loading the neutral value into a vector and then shifting
10922 the initial value into element 0.
10923 (get_initial_defs_for_reduction): Replace the code argument with
10924 the neutral value calculated by neutral_op_for_slp_reduction.
10925 Use gimple_build_vector for constant-length vectors.
10926 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10927 but the first group_size elements have a neutral value.
10928 Use duplicate_and_interleave otherwise.
10929 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10930 Update call to get_initial_defs_for_reduction. Handle SLP
10931 reductions for variable-length vectors by creating one vector
10932 result for each scalar result, with the elements associated
10933 with other scalar results stubbed out with the neutral value.
10934 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10935 Require IFN_VEC_SHL_INSERT for double reductions on
10936 variable-length vectors, or SLP reductions that have
10937 a neutral value. Require can_duplicate_and_interleave_p
10938 support for variable-length unchained SLP reductions if there
10939 is no neutral value, such as for MIN/MAX reductions. Also require
10940 the number of vector elements to be a multiple of the number of
10941 SLP statements when doing variable-length unchained SLP reductions.
10942 Update call to vect_create_epilog_for_reduction.
10943 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10944 and remove initial values.
10945 (duplicate_and_interleave): Make public.
10946 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10947 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10949 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10950 Alan Hayward <alan.hayward@arm.com>
10951 David Sherwood <david.sherwood@arm.com>
10953 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10954 (can_duplicate_and_interleave_p): New function.
10955 (vect_get_and_check_slp_defs): Take the vector of statements
10956 rather than just the current one. Remove excess parentheses.
10957 Restriction rejectinon of vect_constant_def and vect_external_def
10958 for variable-length vectors to boolean types, or types for which
10959 can_duplicate_and_interleave_p is false.
10960 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10961 (duplicate_and_interleave): New function.
10962 (vect_get_constant_vectors): Use gimple_build_vector for
10963 constant-length vectors and suitable variable-length constant
10964 vectors. Use duplicate_and_interleave for other variable-length
10965 vectors. Don't defer the update when inserting new statements.
10967 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10968 Alan Hayward <alan.hayward@arm.com>
10969 David Sherwood <david.sherwood@arm.com>
10971 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10972 min_profitable_iters doesn't go negative.
10974 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10975 Alan Hayward <alan.hayward@arm.com>
10976 David Sherwood <david.sherwood@arm.com>
10978 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10979 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10980 * optabs.def (vec_mask_load_lanes_optab): New optab.
10981 (vec_mask_store_lanes_optab): Likewise.
10982 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10983 (MASK_STORE_LANES): Likewise.
10984 * internal-fn.c (mask_load_lanes_direct): New macro.
10985 (mask_store_lanes_direct): Likewise.
10986 (expand_mask_load_optab_fn): Handle masked operations.
10987 (expand_mask_load_lanes_optab_fn): New macro.
10988 (expand_mask_store_optab_fn): Handle masked operations.
10989 (expand_mask_store_lanes_optab_fn): New macro.
10990 (direct_mask_load_lanes_optab_supported_p): Likewise.
10991 (direct_mask_store_lanes_optab_supported_p): Likewise.
10992 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10994 (vect_load_lanes_supported): Likewise.
10995 * tree-vect-data-refs.c (strip_conversion): New function.
10996 (can_group_stmts_p): Likewise.
10997 (vect_analyze_data_ref_accesses): Use it instead of checking
10998 for a pair of assignments.
10999 (vect_store_lanes_supported): Take a masked_p parameter.
11000 (vect_load_lanes_supported): Likewise.
11001 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11002 vect_store_lanes_supported and vect_load_lanes_supported.
11003 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11004 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11005 parameter. Don't allow gaps for masked accesses.
11006 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11007 and vect_load_lanes_supported.
11008 (get_load_store_type): Take a masked_p parameter and update
11009 call to get_group_load_store_type.
11010 (vectorizable_store): Update call to get_load_store_type.
11011 Handle IFN_MASK_STORE_LANES.
11012 (vectorizable_load): Update call to get_load_store_type.
11013 Handle IFN_MASK_LOAD_LANES.
11015 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11016 Alan Hayward <alan.hayward@arm.com>
11017 David Sherwood <david.sherwood@arm.com>
11019 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11021 * config/aarch64/aarch64-protos.h
11022 (aarch64_sve_struct_memory_operand_p): Declare.
11023 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11024 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11025 (VPRED, vpred): Handle SVE structure modes.
11026 * config/aarch64/constraints.md (Utx): New constraint.
11027 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11028 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11029 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11030 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11031 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11032 structure modes. Split into pieces after RA.
11033 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11034 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11036 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11037 SVE structure modes.
11038 (aarch64_classify_address): Likewise.
11039 (sizetochar): Move earlier in file.
11040 (aarch64_print_operand): Handle SVE register lists.
11041 (aarch64_array_mode): New function.
11042 (aarch64_sve_struct_memory_operand_p): Likewise.
11043 (TARGET_ARRAY_MODE): Redefine.
11045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11046 Alan Hayward <alan.hayward@arm.com>
11047 David Sherwood <david.sherwood@arm.com>
11049 * target.def (array_mode): New target hook.
11050 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11051 * doc/tm.texi: Regenerate.
11052 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11053 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11054 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11055 targetm.array_mode.
11056 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11059 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11060 Alan Hayward <alan.hayward@arm.com>
11061 David Sherwood <david.sherwood@arm.com>
11063 * fold-const.c (fold_binary_loc): Check the argument types
11064 rather than the result type when testing for a vector operation.
11066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11068 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11069 * doc/tm.texi: Regenerate.
11071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11072 Alan Hayward <alan.hayward@arm.com>
11073 David Sherwood <david.sherwood@arm.com>
11075 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11076 (sve): Document new AArch64 extension.
11077 * doc/md.texi (w): Extend the description of the AArch64
11078 constraint to include SVE vectors.
11079 (Upl, Upa): Document new AArch64 predicate constraints.
11080 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11082 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11083 (msve-vector-bits=): New option.
11084 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11085 SVE when these are disabled.
11086 (sve): New extension.
11087 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11088 modes. Adjust their number of units based on aarch64_sve_vg.
11089 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11090 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11091 aarch64_addr_query_type.
11092 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11093 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11094 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11095 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11096 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11097 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11098 (aarch64_simd_imm_zero_p): Delete.
11099 (aarch64_check_zero_based_sve_index_immediate): Declare.
11100 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11101 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11102 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11103 (aarch64_sve_float_mul_immediate_p): Likewise.
11104 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11105 rather than an rtx.
11106 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11107 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11108 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11109 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11110 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11111 (aarch64_regmode_natural_size): Likewise.
11112 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11113 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11115 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11116 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11117 for VG and the SVE predicate registers.
11118 (V_ALIASES): Add a "z"-prefixed alias.
11119 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11120 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11121 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11122 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11123 (REG_CLASS_NAMES): Add entries for them.
11124 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11125 and the predicate registers.
11126 (aarch64_sve_vg): Declare.
11127 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11128 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11129 (REGMODE_NATURAL_SIZE): Define.
11130 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11132 * config/aarch64/aarch64.c: Include cfgrtl.h.
11133 (simd_immediate_info): Add a constructor for series vectors,
11134 and an associated step field.
11135 (aarch64_sve_vg): New variable.
11136 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11137 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11138 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11139 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11140 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11141 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11142 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11143 (aarch64_get_mask_mode): New functions.
11144 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11145 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11146 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11147 predicate modes and predicate registers. Explicitly restrict
11148 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11149 to store a vector mode if it is recognized by
11150 aarch64_classify_vector_mode.
11151 (aarch64_regmode_natural_size): New function.
11152 (aarch64_hard_regno_caller_save_mode): Return the original mode
11154 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11155 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11156 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11157 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11159 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11160 does not overlap dest if the function is frame-related. Handle
11162 (aarch64_split_add_offset): New function.
11163 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11164 them aarch64_add_offset.
11165 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11166 and update call to aarch64_sub_sp.
11167 (aarch64_add_cfa_expression): New function.
11168 (aarch64_expand_prologue): Pass extra temporary registers to the
11169 functions above. Handle the case in which we need to emit new
11170 DW_CFA_expressions for registers that were originally saved
11171 relative to the stack pointer, but now have to be expressed
11172 relative to the frame pointer.
11173 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11175 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11176 IP0 and IP1 values for SVE frames.
11177 (aarch64_expand_vec_series): New function.
11178 (aarch64_expand_sve_widened_duplicate): Likewise.
11179 (aarch64_expand_sve_const_vector): Likewise.
11180 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11181 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11182 into the register, rather than emitting a SET directly.
11183 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11184 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11185 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11186 (offset_9bit_signed_scaled_p): New functions.
11187 (aarch64_replicate_bitmask_imm): New function.
11188 (aarch64_bitmask_imm): Use it.
11189 (aarch64_cannot_force_const_mem): Reject expressions involving
11190 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11191 (aarch64_classify_index): Handle SVE indices, by requiring
11192 a plain register index with a scale that matches the element size.
11193 (aarch64_classify_address): Handle SVE addresses. Assert that
11194 the mode of the address is VOIDmode or an integer mode.
11195 Update call to aarch64_classify_symbol.
11196 (aarch64_classify_symbolic_expression): Update call to
11197 aarch64_classify_symbol.
11198 (aarch64_const_vec_all_in_range_p): New function.
11199 (aarch64_print_vector_float_operand): Likewise.
11200 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11201 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11202 and the FP immediates 1.0 and 0.5.
11203 (aarch64_print_address_internal): Handle SVE addresses.
11204 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11205 (aarch64_regno_regclass): Handle predicate registers.
11206 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11208 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11209 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11210 (aarch64_convert_sve_vector_bits): New function.
11211 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11212 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11213 rather than an rtx.
11214 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11215 Handle SVE vector and predicate modes. Accept VL-based constants
11216 that need only one temporary register, and VL offsets that require
11217 no temporary registers.
11218 (aarch64_conditional_register_usage): Mark the predicate registers
11219 as fixed if SVE isn't available.
11220 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11221 Return true for SVE vector and predicate modes.
11222 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11223 rather than an unsigned int. Handle SVE modes.
11224 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11226 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11228 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11229 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11230 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11231 (aarch64_sve_float_mul_immediate_p): New functions.
11232 (aarch64_sve_valid_immediate): New function.
11233 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11234 Explicitly reject structure modes. Check for INDEX constants.
11235 Handle PTRUE and PFALSE constants.
11236 (aarch64_check_zero_based_sve_index_immediate): New function.
11237 (aarch64_simd_imm_zero_p): Delete.
11238 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11239 vector modes. Accept constants in the range of CNT[BHWD].
11240 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11241 ask for an Advanced SIMD mode.
11242 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11243 (aarch64_simd_vector_alignment): Handle SVE predicates.
11244 (aarch64_vectorize_preferred_vector_alignment): New function.
11245 (aarch64_simd_vector_alignment_reachable): Use it instead of
11247 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11248 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11250 (MAX_VECT_LEN): Delete.
11251 (expand_vec_perm_d): Add a vec_flags field.
11252 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11253 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11254 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11256 (aarch64_evpc_rev): Rename to...
11257 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11258 (aarch64_evpc_rev_global): New function.
11259 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11260 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11262 (aarch64_evpc_sve_tbl): New function.
11263 (aarch64_expand_vec_perm_const_1): Update after rename of
11264 aarch64_evpc_rev. Handle SVE permutes too, trying
11265 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11266 than aarch64_evpc_tbl.
11267 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11268 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11269 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11270 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11271 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11272 (aarch64_expand_sve_vcond): New functions.
11273 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11274 of aarch64_vector_mode_p.
11275 (aarch64_dwarf_poly_indeterminate_value): New function.
11276 (aarch64_compute_pressure_classes): Likewise.
11277 (aarch64_can_change_mode_class): Likewise.
11278 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11279 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11280 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11281 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11282 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11283 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11284 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11285 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11287 (Dn, Dl, Dr): Accept const as well as const_vector.
11288 (Dz): Likewise. Compare against CONST0_RTX.
11289 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11290 of "vector" where appropriate.
11291 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11292 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11293 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11294 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11295 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11296 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11297 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11298 (v_int_equiv): Extend to SVE modes.
11299 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11301 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11302 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11303 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11304 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11305 (SVE_COND_FP_CMP): New int iterators.
11306 (perm_hilo): Handle the new unpack unspecs.
11307 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11309 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11310 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11311 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11312 (aarch64_equality_operator, aarch64_constant_vector_operand)
11313 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11314 (aarch64_sve_nonimmediate_operand): Likewise.
11315 (aarch64_sve_general_operand): Likewise.
11316 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11317 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11318 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11319 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11320 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11321 (aarch64_sve_float_arith_immediate): Likewise.
11322 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11323 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11324 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11325 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11326 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11327 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11328 (aarch64_sve_float_arith_operand): Likewise.
11329 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11330 (aarch64_sve_float_mul_operand): Likewise.
11331 (aarch64_sve_vec_perm_operand): Likewise.
11332 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11333 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11334 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11335 as well as const_vector.
11336 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11337 in file. Use CONST0_RTX and CONSTM1_RTX.
11338 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11339 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11340 Use aarch64_simd_imm_zero.
11341 * config/aarch64/aarch64-sve.md: New file.
11342 * config/aarch64/aarch64.md: Include it.
11343 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11344 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11345 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11346 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11347 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11348 (sve): New attribute.
11349 (enabled): Disable instructions with the sve attribute unless
11351 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11352 aarch64_expand_mov_immediate.
11353 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11354 CNT[BHSD] immediates.
11355 (movti): Split CONST_POLY_INT moves into two halves.
11356 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11357 Split additions that need a temporary here if the destination
11358 is the stack pointer.
11359 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11360 (*add<mode>3_poly_1): New instruction.
11361 (set_clobber_cc): New expander.
11363 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11365 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11366 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11367 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11368 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11369 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11370 Change innermode from fixed_mode_size to machine_mode.
11371 (simplify_subreg): Update call accordingly. Handle a constant-sized
11372 subreg of a variable-length CONST_VECTOR.
11374 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11375 Alan Hayward <alan.hayward@arm.com>
11376 David Sherwood <david.sherwood@arm.com>
11378 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11379 (add_offset_to_base): New function, split out from...
11380 (create_mem_ref): ...here. When handling a scale other than 1,
11381 check first whether the address is valid without the offset.
11382 Add it into the base if so, leaving the index and scale as-is.
11384 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11387 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11388 fold_for_warn before checking if arg2 is INTEGER_CST.
11390 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11392 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11393 (store_multiple_operation): Delete.
11394 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11395 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11396 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11397 guarded by TARGET_STRING.
11398 (rs6000_output_load_multiple): Delete.
11399 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11400 OPTION_MASK_STRING / TARGET_STRING handling.
11401 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11402 (const rs6000_opt_masks) <"string">: Change mask to 0.
11403 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11404 (MASK_STRING): Delete.
11405 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11407 (load_multiple): Delete.
11414 (store_multiple): Delete.
11421 (movmemsi_8reg): Delete.
11422 (corresponding unnamed define_insn): Delete.
11423 (movmemsi_6reg): Delete.
11424 (corresponding unnamed define_insn): Delete.
11425 (movmemsi_4reg): Delete.
11426 (corresponding unnamed define_insn): Delete.
11427 (movmemsi_2reg): Delete.
11428 (corresponding unnamed define_insn): Delete.
11429 (movmemsi_1reg): Delete.
11430 (corresponding unnamed define_insn): Delete.
11431 * config/rs6000/rs6000.opt (mno-string): New.
11432 (mstring): Replace by deprecation warning stub.
11433 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11435 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11437 * regrename.c (regrename_do_replace): If replacing the same
11438 reg multiple times, try to reuse last created gen_raw_REG.
11441 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11442 main to workaround a bug in GDB.
11444 2018-01-12 Tom de Vries <tom@codesourcery.com>
11447 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11449 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11451 PR rtl-optimization/80481
11452 * ira-color.c (get_cap_member): New function.
11453 (allocnos_conflict_by_live_ranges_p): Use it.
11454 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11455 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11457 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11460 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11461 (*saddl_se_1): Ditto.
11462 (*ssubsi_1): Ditto.
11463 (*ssubl_se_1): Ditto.
11465 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11467 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11468 rather than wi::to_widest for DR_INITs.
11469 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11470 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11471 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11473 (vect_analyze_group_access_1): Note that here.
11475 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11477 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11478 polynomial type sizes.
11480 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11482 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11483 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11484 (gimple_add_tmp_var): Likewise.
11486 2018-01-12 Martin Liska <mliska@suse.cz>
11488 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11489 (gimple_alloc_sizes): Likewise.
11490 (dump_gimple_statistics): Use PRIu64 in printf format.
11491 * gimple.h: Change uint64_t to int.
11493 2018-01-12 Martin Liska <mliska@suse.cz>
11495 * tree-core.h: Use uint64_t instead of int.
11496 * tree.c (tree_node_counts): Likewise.
11497 (tree_node_sizes): Likewise.
11498 (dump_tree_statistics): Use PRIu64 in printf format.
11500 2018-01-12 Martin Liska <mliska@suse.cz>
11502 * Makefile.in: As qsort_chk is implemented in vec.c, add
11503 vec.o to linkage of gencfn-macros.
11504 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11505 passing the info to record_node_allocation_statistics.
11506 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11508 * ggc-common.c (struct ggc_usage): Add operator== and use
11509 it in operator< and compare function.
11510 * mem-stats.h (struct mem_usage): Likewise.
11511 * vec.c (struct vec_usage): Remove operator< and compare
11512 function. Can be simply inherited.
11514 2018-01-12 Martin Jambor <mjambor@suse.cz>
11517 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11518 * tree-ssa-math-opts.c: Include domwalk.h.
11519 (convert_mult_to_fma_1): New function.
11520 (fma_transformation_info): New type.
11521 (fma_deferring_state): Likewise.
11522 (cancel_fma_deferring): New function.
11523 (result_of_phi): Likewise.
11524 (last_fma_candidate_feeds_initial_phi): Likewise.
11525 (convert_mult_to_fma): Added deferring logic, split actual
11526 transformation to convert_mult_to_fma_1.
11527 (math_opts_dom_walker): New type.
11528 (math_opts_dom_walker::after_dom_children): New method, body moved
11529 here from pass_optimize_widening_mul::execute, added deferring logic
11531 (pass_optimize_widening_mul::execute): Moved most of code to
11532 math_opts_dom_walker::after_dom_children.
11533 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11534 * config/i386/i386.c (ix86_option_override_internal): Added
11535 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11537 2018-01-12 Richard Biener <rguenther@suse.de>
11540 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11541 inline instance vars.
11543 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11546 * config/rx/rx.c (rx_is_restricted_memory_address):
11547 Handle SUBREG case.
11549 2018-01-12 Richard Biener <rguenther@suse.de>
11551 PR tree-optimization/80846
11552 * target.def (split_reduction): New target hook.
11553 * targhooks.c (default_split_reduction): New function.
11554 * targhooks.h (default_split_reduction): Declare.
11555 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11556 target requests first reduce vectors by combining low and high
11558 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11559 (get_vectype_for_scalar_type_and_size): Export.
11560 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11561 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11562 * doc/tm.texi: Regenerate.
11563 * config/i386/i386.c (ix86_split_reduction): Implement
11564 TARGET_VECTORIZE_SPLIT_REDUCTION.
11566 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11569 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11570 in PIC mode except for TARGET_VXWORKS_RTP.
11571 * config/sparc/sparc.c: Include cfgrtl.h.
11572 (TARGET_INIT_PIC_REG): Define.
11573 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11574 (sparc_pic_register_p): New predicate.
11575 (sparc_legitimate_address_p): Use it.
11576 (sparc_legitimize_pic_address): Likewise.
11577 (sparc_delegitimize_address): Likewise.
11578 (sparc_mode_dependent_address_p): Likewise.
11579 (gen_load_pcrel_sym): Remove 4th parameter.
11580 (load_got_register): Adjust call to above. Remove obsolete stuff.
11581 (sparc_expand_prologue): Do not call load_got_register here.
11582 (sparc_flat_expand_prologue): Likewise.
11583 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11584 (sparc_use_pseudo_pic_reg): New function.
11585 (sparc_init_pic_reg): Likewise.
11586 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11587 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11589 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11591 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11592 Add item for branch_cost.
11594 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11596 PR rtl-optimization/83565
11597 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11598 not extend the result to a larger mode for rotate operations.
11599 (num_sign_bit_copies1): Likewise.
11601 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11604 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11606 Use values-Xc.o for -pedantic.
11607 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11609 2018-01-12 Martin Liska <mliska@suse.cz>
11612 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11614 (possible_polymorphic_call_targets): Use it.
11615 (ipa_devirt): Likewise.
11617 2018-01-12 Martin Liska <mliska@suse.cz>
11619 * profile-count.h (enum profile_quality): Use 0 as invalid
11620 enum value of profile_quality.
11622 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11624 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11625 -mext-string options.
11627 2018-01-12 Richard Biener <rguenther@suse.de>
11629 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11630 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11631 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11633 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11635 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11637 * configure.ac (--with-long-double-format): Add support for the
11638 configuration option to change the default long double format on
11640 * config.gcc (powerpc*-linux*-*): Likewise.
11641 * configure: Regenerate.
11642 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11643 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11644 used without modification.
11646 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11648 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11649 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11650 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11651 MISC_BUILTIN_SPEC_BARRIER.
11652 (rs6000_init_builtins): Likewise.
11653 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11655 (speculation_barrier): New define_insn.
11656 * doc/extend.texi: Document __builtin_speculation_barrier.
11658 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11661 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11662 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11663 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11665 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11666 integral modes instead of "ss" and "sd".
11667 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11668 vectors with 32-bit and 64-bit elements.
11669 (vecdupssescalarmodesuffix): New mode attribute.
11670 (vec_dup<mode>): Use it.
11672 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11675 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11676 frame if argument is passed on stack.
11678 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11681 * ree.c (combine_reaching_defs): Optimize also
11682 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11683 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11685 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11687 PR middle-end/83189
11688 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11690 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11692 PR middle-end/83718
11693 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11694 after they are computed.
11696 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11698 PR tree-optimization/83695
11699 * gimple-loop-linterchange.cc
11700 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11701 reset cached scev information after interchange.
11702 (pass_linterchange::execute): Remove call to scev_reset_htab.
11704 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11706 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11707 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11708 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11709 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11710 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11711 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11712 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11713 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11714 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11715 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11716 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11717 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11718 (V_lane_reg): Likewise.
11719 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11721 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11722 (vfmal_lane_low<mode>_intrinsic,
11723 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11724 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11725 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11726 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11727 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11728 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11730 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11732 * config/arm/arm-cpus.in (fp16fml): New feature.
11733 (ALL_SIMD): Add fp16fml.
11734 (armv8.2-a): Add fp16fml as an option.
11735 (armv8.3-a): Likewise.
11736 (armv8.4-a): Add fp16fml as part of fp16.
11737 * config/arm/arm.h (TARGET_FP16FML): Define.
11738 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11740 * config/arm/arm-modes.def (V2HF): Define.
11741 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11742 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11743 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11744 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11745 vfmsl_low, vfmsl_high): New set of builtins.
11746 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11747 (vfml_op): New code attribute.
11748 (VFMLHALVES): New int iterator.
11749 (VFML, VFMLSEL): New mode attributes.
11750 (V_reg): Define mapping for V2HF.
11751 (V_hi, V_lo): New mode attributes.
11752 (VF_constraint): Likewise.
11753 (vfml_half, vfml_half_selector): New int attributes.
11754 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11756 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11757 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11759 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11760 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11761 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11762 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11764 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11765 Document new effective target and option set.
11767 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11769 * config/arm/arm-cpus.in (armv8_4): New feature.
11770 (ARMv8_4a): New fgroup.
11771 (armv8.4-a): New arch.
11772 * config/arm/arm-tables.opt: Regenerate.
11773 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11774 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11775 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11776 Add matching rules for -march=armv8.4-a and extensions.
11777 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11779 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11782 * config/rx/rx.md (BW): New mode attribute.
11783 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11785 2018-01-11 Richard Biener <rguenther@suse.de>
11787 PR tree-optimization/83435
11788 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11789 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11790 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11792 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11793 Alan Hayward <alan.hayward@arm.com>
11794 David Sherwood <david.sherwood@arm.com>
11796 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11798 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11799 (aarch64_print_address_internal): Use it to check for a zero offset.
11801 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11802 Alan Hayward <alan.hayward@arm.com>
11803 David Sherwood <david.sherwood@arm.com>
11805 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11806 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11807 Return a poly_int64 rather than a HOST_WIDE_INT.
11808 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11809 rather than a HOST_WIDE_INT.
11810 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11811 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11812 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11813 final_offset from HOST_WIDE_INT to poly_int64.
11814 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11815 to_constant when getting the number of units in an Advanced SIMD
11817 (aarch64_builtin_vectorized_function): Check for a constant number
11819 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11821 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11822 attribute instead of GET_MODE_NUNITS.
11823 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11824 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11825 GET_MODE_SIZE for fixed-size registers.
11826 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11827 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11828 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11829 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11830 (aarch64_print_operand, aarch64_print_address_internal)
11831 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11832 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11833 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11834 Handle polynomial GET_MODE_SIZE.
11835 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11836 wider than SImode without modification.
11837 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11838 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11839 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11840 passing and returning SVE modes.
11841 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11842 rather than GEN_INT.
11843 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11844 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11845 (aarch64_allocate_and_probe_stack_space): Likewise.
11846 (aarch64_layout_frame): Cope with polynomial offsets.
11847 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11848 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11849 polynomial offsets.
11850 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11851 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11852 poly_int64 rather than a HOST_WIDE_INT.
11853 (aarch64_get_separate_components, aarch64_process_components)
11854 (aarch64_expand_prologue, aarch64_expand_epilogue)
11855 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11856 (aarch64_anchor_offset): New function, split out from...
11857 (aarch64_legitimize_address): ...here.
11858 (aarch64_builtin_vectorization_cost): Handle polynomial
11859 TYPE_VECTOR_SUBPARTS.
11860 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11862 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11863 number of elements from the PARALLEL rather than the mode.
11864 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11865 rather than GET_MODE_BITSIZE.
11866 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11867 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11868 (aarch64_expand_vec_perm_const_1): Handle polynomial
11869 d->perm.length () and d->perm elements.
11870 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11871 Apply to_constant to d->perm elements.
11872 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11873 polynomial CONST_VECTOR_NUNITS.
11874 (aarch64_move_pointer): Take amount as a poly_int64 rather
11876 (aarch64_progress_pointer): Avoid temporary variable.
11877 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11878 the mode attribute instead of GET_MODE.
11880 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11881 Alan Hayward <alan.hayward@arm.com>
11882 David Sherwood <david.sherwood@arm.com>
11884 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11885 x exists before using it.
11886 (aarch64_add_constant_internal): Rename to...
11887 (aarch64_add_offset_1): ...this. Replace regnum with separate
11888 src and dest rtxes. Handle the case in which they're different,
11889 including when the offset is zero. Replace scratchreg with an rtx.
11890 Use 2 additions if there is no spare register into which we can
11891 move a 16-bit constant.
11892 (aarch64_add_constant): Delete.
11893 (aarch64_add_offset): Replace reg with separate src and dest
11894 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11895 Use aarch64_add_offset_1.
11896 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11897 an rtx rather than an int. Take the delta as a poly_int64
11898 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11899 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11900 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11901 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11902 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11903 and aarch64_add_sp.
11904 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11905 aarch64_add_constant.
11907 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11909 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11910 Use scalar_float_mode.
11912 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11914 * config/aarch64/aarch64-simd.md
11915 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11916 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11917 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11918 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11919 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11920 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11921 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11922 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11923 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11924 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11926 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11929 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11930 targ_options->x_arm_arch_string is non NULL.
11932 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11934 * config/aarch64/aarch64.h
11935 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11937 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11940 * expmed.c (emit_store_flag_force): Swap if const op0
11941 and change VOIDmode to mode of op0.
11943 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11945 PR rtl-optimization/83761
11946 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11947 than bytes to mode_for_size.
11949 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11951 PR middle-end/83189
11952 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11953 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11956 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11958 PR middle-end/83575
11959 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11960 when in layout mode.
11961 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11962 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11965 2018-01-10 Michael Collison <michael.collison@arm.com>
11967 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11968 * config/aarch64/aarch64-option-extension.def: Add
11969 AARCH64_OPT_EXTENSION of 'fp16fml'.
11970 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11971 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11972 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11973 * config/aarch64/constraints.md (Ui7): New constraint.
11974 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11975 (VFMLA_SEL_W): Ditto.
11978 (VFMLA16_LOW): New int iterator.
11979 (VFMLA16_HIGH): Ditto.
11980 (UNSPEC_FMLAL): New unspec.
11981 (UNSPEC_FMLSL): Ditto.
11982 (UNSPEC_FMLAL2): Ditto.
11983 (UNSPEC_FMLSL2): Ditto.
11984 (f16mac): New code attribute.
11985 * config/aarch64/aarch64-simd-builtins.def
11986 (aarch64_fmlal_lowv2sf): Ditto.
11987 (aarch64_fmlsl_lowv2sf): Ditto.
11988 (aarch64_fmlalq_lowv4sf): Ditto.
11989 (aarch64_fmlslq_lowv4sf): Ditto.
11990 (aarch64_fmlal_highv2sf): Ditto.
11991 (aarch64_fmlsl_highv2sf): Ditto.
11992 (aarch64_fmlalq_highv4sf): Ditto.
11993 (aarch64_fmlslq_highv4sf): Ditto.
11994 (aarch64_fmlal_lane_lowv2sf): Ditto.
11995 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11996 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11997 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11998 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11999 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12000 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12001 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12002 (aarch64_fmlal_lane_highv2sf): Ditto.
12003 (aarch64_fmlsl_lane_highv2sf): Ditto.
12004 (aarch64_fmlal_laneq_highv2sf): Ditto.
12005 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12006 (aarch64_fmlalq_lane_highv4sf): Ditto.
12007 (aarch64_fmlsl_lane_highv4sf): Ditto.
12008 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12009 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12010 * config/aarch64/aarch64-simd.md:
12011 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12012 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12013 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12014 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12015 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12016 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12017 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12018 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12019 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12020 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12021 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12022 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12023 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12024 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12025 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12026 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12027 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12028 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12029 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12030 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12031 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12032 (vfmlsl_low_u32): Ditto.
12033 (vfmlalq_low_u32): Ditto.
12034 (vfmlslq_low_u32): Ditto.
12035 (vfmlal_high_u32): Ditto.
12036 (vfmlsl_high_u32): Ditto.
12037 (vfmlalq_high_u32): Ditto.
12038 (vfmlslq_high_u32): Ditto.
12039 (vfmlal_lane_low_u32): Ditto.
12040 (vfmlsl_lane_low_u32): Ditto.
12041 (vfmlal_laneq_low_u32): Ditto.
12042 (vfmlsl_laneq_low_u32): Ditto.
12043 (vfmlalq_lane_low_u32): Ditto.
12044 (vfmlslq_lane_low_u32): Ditto.
12045 (vfmlalq_laneq_low_u32): Ditto.
12046 (vfmlslq_laneq_low_u32): Ditto.
12047 (vfmlal_lane_high_u32): Ditto.
12048 (vfmlsl_lane_high_u32): Ditto.
12049 (vfmlal_laneq_high_u32): Ditto.
12050 (vfmlsl_laneq_high_u32): Ditto.
12051 (vfmlalq_lane_high_u32): Ditto.
12052 (vfmlslq_lane_high_u32): Ditto.
12053 (vfmlalq_laneq_high_u32): Ditto.
12054 (vfmlslq_laneq_high_u32): Ditto.
12055 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12056 (AARCH64_FL_FOR_ARCH8_4): New.
12057 (AARCH64_ISA_F16FML): New ISA flag.
12058 (TARGET_F16FML): New feature flag for fp16fml.
12059 (doc/invoke.texi): Document new fp16fml option.
12061 2018-01-10 Michael Collison <michael.collison@arm.com>
12063 * config/aarch64/aarch64-builtins.c:
12064 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12065 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12066 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12067 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12068 (AARCH64_ISA_SHA3): New ISA flag.
12069 (TARGET_SHA3): New feature flag for sha3.
12070 * config/aarch64/iterators.md (sha512_op): New int attribute.
12071 (CRYPTO_SHA512): New int iterator.
12072 (UNSPEC_SHA512H): New unspec.
12073 (UNSPEC_SHA512H2): Ditto.
12074 (UNSPEC_SHA512SU0): Ditto.
12075 (UNSPEC_SHA512SU1): Ditto.
12076 * config/aarch64/aarch64-simd-builtins.def
12077 (aarch64_crypto_sha512hqv2di): New builtin.
12078 (aarch64_crypto_sha512h2qv2di): Ditto.
12079 (aarch64_crypto_sha512su0qv2di): Ditto.
12080 (aarch64_crypto_sha512su1qv2di): Ditto.
12081 (aarch64_eor3qv8hi): Ditto.
12082 (aarch64_rax1qv2di): Ditto.
12083 (aarch64_xarqv2di): Ditto.
12084 (aarch64_bcaxqv8hi): Ditto.
12085 * config/aarch64/aarch64-simd.md:
12086 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12087 (aarch64_crypto_sha512su0qv2di): Ditto.
12088 (aarch64_crypto_sha512su1qv2di): Ditto.
12089 (aarch64_eor3qv8hi): Ditto.
12090 (aarch64_rax1qv2di): Ditto.
12091 (aarch64_xarqv2di): Ditto.
12092 (aarch64_bcaxqv8hi): Ditto.
12093 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12094 (vsha512h2q_u64): Ditto.
12095 (vsha512su0q_u64): Ditto.
12096 (vsha512su1q_u64): Ditto.
12097 (veor3q_u16): Ditto.
12098 (vrax1q_u64): Ditto.
12099 (vxarq_u64): Ditto.
12100 (vbcaxq_u16): Ditto.
12101 * config/arm/types.md (crypto_sha512): New type attribute.
12102 (crypto_sha3): Ditto.
12103 (doc/invoke.texi): Document new sha3 option.
12105 2018-01-10 Michael Collison <michael.collison@arm.com>
12107 * config/aarch64/aarch64-builtins.c:
12108 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12109 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12110 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12111 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12112 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12113 (AARCH64_ISA_SM4): New ISA flag.
12114 (TARGET_SM4): New feature flag for sm4.
12115 * config/aarch64/aarch64-simd-builtins.def
12116 (aarch64_sm3ss1qv4si): Ditto.
12117 (aarch64_sm3tt1aq4si): Ditto.
12118 (aarch64_sm3tt1bq4si): Ditto.
12119 (aarch64_sm3tt2aq4si): Ditto.
12120 (aarch64_sm3tt2bq4si): Ditto.
12121 (aarch64_sm3partw1qv4si): Ditto.
12122 (aarch64_sm3partw2qv4si): Ditto.
12123 (aarch64_sm4eqv4si): Ditto.
12124 (aarch64_sm4ekeyqv4si): Ditto.
12125 * config/aarch64/aarch64-simd.md:
12126 (aarch64_sm3ss1qv4si): Ditto.
12127 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12128 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12129 (aarch64_sm4eqv4si): Ditto.
12130 (aarch64_sm4ekeyqv4si): Ditto.
12131 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12132 (sm3part_op): Ditto.
12133 (CRYPTO_SM3TT): Ditto.
12134 (CRYPTO_SM3PART): Ditto.
12135 (UNSPEC_SM3SS1): New unspec.
12136 (UNSPEC_SM3TT1A): Ditto.
12137 (UNSPEC_SM3TT1B): Ditto.
12138 (UNSPEC_SM3TT2A): Ditto.
12139 (UNSPEC_SM3TT2B): Ditto.
12140 (UNSPEC_SM3PARTW1): Ditto.
12141 (UNSPEC_SM3PARTW2): Ditto.
12142 (UNSPEC_SM4E): Ditto.
12143 (UNSPEC_SM4EKEY): Ditto.
12144 * config/aarch64/constraints.md (Ui2): New constraint.
12145 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12146 * config/arm/types.md (crypto_sm3): New type attribute.
12147 (crypto_sm4): Ditto.
12148 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12149 (vsm3tt1aq_u32): Ditto.
12150 (vsm3tt1bq_u32): Ditto.
12151 (vsm3tt2aq_u32): Ditto.
12152 (vsm3tt2bq_u32): Ditto.
12153 (vsm3partw1q_u32): Ditto.
12154 (vsm3partw2q_u32): Ditto.
12155 (vsm4eq_u32): Ditto.
12156 (vsm4ekeyq_u32): Ditto.
12157 (doc/invoke.texi): Document new sm4 option.
12159 2018-01-10 Michael Collison <michael.collison@arm.com>
12161 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12162 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12163 (AARCH64_FL_FOR_ARCH8_4): New.
12164 (AARCH64_FL_V8_4): New flag.
12165 (doc/invoke.texi): Document new armv8.4-a option.
12167 2018-01-10 Michael Collison <michael.collison@arm.com>
12169 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12170 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12171 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12172 * config/aarch64/aarch64-option-extension.def: Add
12173 AARCH64_OPT_EXTENSION of 'sha2'.
12174 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12175 (crypto): Disable sha2 and aes if crypto disabled.
12176 (crypto): Enable aes and sha2 if enabled.
12177 (simd): Disable sha2 and aes if simd disabled.
12178 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12180 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12181 (TARGET_SHA2): New feature flag for sha2.
12182 (TARGET_AES): New feature flag for aes.
12183 * config/aarch64/aarch64-simd.md:
12184 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12185 conditional on TARGET_AES.
12186 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12187 (aarch64_crypto_sha1hsi): Make pattern conditional
12189 (aarch64_crypto_sha1hv4si): Ditto.
12190 (aarch64_be_crypto_sha1hv4si): Ditto.
12191 (aarch64_crypto_sha1su1v4si): Ditto.
12192 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12193 (aarch64_crypto_sha1su0v4si): Ditto.
12194 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12195 (aarch64_crypto_sha256su0v4si): Ditto.
12196 (aarch64_crypto_sha256su1v4si): Ditto.
12197 (doc/invoke.texi): Document new aes and sha2 options.
12199 2018-01-10 Martin Sebor <msebor@redhat.com>
12201 PR tree-optimization/83781
12202 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12205 2018-01-11 Martin Sebor <msebor@gmail.com>
12206 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12208 PR tree-optimization/83501
12209 PR tree-optimization/81703
12211 * tree-ssa-strlen.c (get_string_cst): Rename...
12212 (get_string_len): ...to this. Handle global constants.
12213 (handle_char_store): Adjust.
12215 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12216 Jim Wilson <jimw@sifive.com>
12218 * config/riscv/riscv-protos.h (riscv_output_return): New.
12219 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12220 (riscv_attribute_table, riscv_output_return),
12221 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12222 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12223 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12225 (riscv_expand_prologue): Add early return for naked function.
12226 (riscv_expand_epilogue): Likewise.
12227 (riscv_function_ok_for_sibcall): Return false for naked function.
12228 (riscv_set_current_function): New.
12229 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12230 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12231 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12232 * doc/extend.texi (RISC-V Function Attributes): New.
12234 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12236 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12237 check for 128-bit long double before checking TCmode.
12238 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12239 128-bit long doubles before checking TFmode or TCmode.
12240 (FLOAT128_IBM_P): Likewise.
12242 2018-01-10 Martin Sebor <msebor@redhat.com>
12244 PR tree-optimization/83671
12245 * builtins.c (c_strlen): Unconditionally return zero for the empty
12247 Use -Warray-bounds for warnings.
12248 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12249 for non-constant array indices with COMPONENT_REF, arrays of
12250 arrays, and pointers to arrays.
12251 (gimple_fold_builtin_strlen): Determine and set length range for
12252 non-constant character arrays.
12254 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12256 PR middle-end/81897
12257 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12260 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12262 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12264 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12267 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12268 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12269 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12270 indexed_or_indirect_operand predicate.
12271 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12272 (*vsx_le_perm_load_v8hi): Likewise.
12273 (*vsx_le_perm_load_v16qi): Likewise.
12274 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12275 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12276 (*vsx_le_perm_store_v8hi): Likewise.
12277 (*vsx_le_perm_store_v16qi): Likewise.
12278 (eight unnamed splitters): Likewise.
12280 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12282 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12283 * config/rs6000/emmintrin.h: Likewise.
12284 * config/rs6000/mmintrin.h: Likewise.
12285 * config/rs6000/xmmintrin.h: Likewise.
12287 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12290 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12292 * tree.c (tree_nop_conversion): Return true for location wrapper
12294 (maybe_wrap_with_location): New function.
12295 (selftest::check_strip_nops): New function.
12296 (selftest::test_location_wrappers): New function.
12297 (selftest::tree_c_tests): Call it.
12298 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12299 (maybe_wrap_with_location): New decl.
12300 (EXPR_LOCATION_WRAPPER_P): New macro.
12301 (location_wrapper_p): New inline function.
12302 (tree_strip_any_location_wrapper): New inline function.
12304 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12307 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12308 stack_realign_offset for the largest alignment of stack slot
12310 (ix86_find_max_used_stack_alignment): New function.
12311 (ix86_finalize_stack_frame_flags): Use it. Set
12312 max_used_stack_alignment if we don't realign stack.
12313 * config/i386/i386.h (machine_function): Add
12314 max_used_stack_alignment.
12316 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12318 * config/arm/arm.opt (-mbranch-cost): New option.
12319 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12322 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12325 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12326 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12328 2018-01-10 Richard Biener <rguenther@suse.de>
12331 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12332 early out so it also covers the case where we have a non-NULL
12335 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12337 PR tree-optimization/83753
12338 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12339 for non-strided grouped accesses if the number of elements is 1.
12341 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12344 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12345 * i386.h (TARGET_USE_GATHER): Define.
12346 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12348 2018-01-10 Martin Liska <mliska@suse.cz>
12351 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12352 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12354 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12355 CLEANUP_NO_PARTITIONING is not set.
12357 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12359 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12360 for vectors, as a partial revert of r254296.
12361 * rtl.h (const_vec_p): Delete.
12362 (const_vec_duplicate_p): Don't test for vector CONSTs.
12363 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12364 * expmed.c (make_tree): Likewise.
12367 * common.md (E, F): Use CONSTANT_P instead of checking for
12369 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12370 checking for CONST_VECTOR.
12372 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12374 PR middle-end/83575
12375 * predict.c (force_edge_cold): Handle in more sane way edges
12376 with no prediction.
12378 2018-01-09 Carl Love <cel@us.ibm.com>
12380 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12382 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12383 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12384 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12385 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12386 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12387 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12388 * config/rs6000/rs6000-protos.h: Add extern defition for
12389 rs6000_generate_float2_double_code.
12390 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12392 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12393 (float2_v2df): Add define_expand.
12395 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12398 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12399 op_mode in the force_to_mode call.
12401 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12403 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12404 instead of checking each element individually.
12405 (aarch64_evpc_uzp): Likewise.
12406 (aarch64_evpc_zip): Likewise.
12407 (aarch64_evpc_ext): Likewise.
12408 (aarch64_evpc_rev): Likewise.
12409 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12410 instead of checking each element individually. Return true without
12412 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12413 whether all selected elements come from the same input, instead of
12414 checking each element individually. Remove calls to gen_rtx_REG,
12415 start_sequence and end_sequence and instead assert that no rtl is
12418 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12420 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12421 order of HIGH and CONST checks.
12423 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12425 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12426 if the destination isn't an SSA_NAME.
12428 2018-01-09 Richard Biener <rguenther@suse.de>
12430 PR tree-optimization/83668
12431 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12433 (canonicalize_loop_form): ... here, renamed from ...
12434 (canonicalize_loop_closed_ssa_form): ... this and amended to
12435 swap successor edges for loop exit blocks to make us use
12436 the RPO order we need for initial schedule generation.
12438 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12440 PR tree-optimization/64811
12441 * match.pd: When optimizing comparisons with Inf, avoid
12442 introducing or losing exceptions from comparisons with NaN.
12444 2018-01-09 Martin Liska <mliska@suse.cz>
12447 * asan.c (shadow_mem_size): Add gcc_assert.
12449 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12451 Don't save registers in main().
12454 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12455 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12456 * config/avr/avr.c (avr_set_current_function): Don't error if
12457 naked, OS_task or OS_main are specified at the same time.
12458 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12460 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12462 * common/config/avr/avr-common.c (avr_option_optimization_table):
12463 Switch on -mmain-is-OS_task for optimizing compilations.
12465 2018-01-09 Richard Biener <rguenther@suse.de>
12467 PR tree-optimization/83572
12468 * graphite.c: Include cfganal.h.
12469 (graphite_transform_loops): Connect infinite loops to exit
12470 and remove fake edges at the end.
12472 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12474 * ipa-inline.c (edge_badness): Revert accidental checkin.
12476 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12479 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12480 symbols; not inline clones.
12482 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12485 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12486 hard registers. Formatting fixes.
12488 PR preprocessor/83722
12489 * gcc.c (try_generate_repro): Pass
12490 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12491 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12494 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12495 Kito Cheng <kito.cheng@gmail.com>
12497 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12498 (riscv_leaf_function_p): Delete.
12499 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12501 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12503 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12505 (do_ifelse): New function.
12506 (do_isel): New function.
12507 (do_sub3): New function.
12508 (do_add3): New function.
12509 (do_load_mask_compare): New function.
12510 (do_overlap_load_compare): New function.
12511 (expand_compare_loop): New function.
12512 (expand_block_compare): Call expand_compare_loop() when appropriate.
12513 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12514 option description.
12515 (-mblock-compare-inline-loop-limit): New option.
12517 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12520 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12521 Reverse order of second and third operands in first alternative.
12522 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12523 of first and second elements in UNSPEC_VPERMR vector.
12524 (altivec_expand_vec_perm_le): Likewise.
12526 2018-01-08 Jeff Law <law@redhat.com>
12528 PR rtl-optimizatin/81308
12529 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12530 (process_switch): If group_case_labels makes a change, then set
12532 (pass_convert_switch::execute): If a switch is converted, then
12533 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12535 PR rtl-optimization/81308
12536 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12539 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12541 PR target/83663 - Revert r255946
12542 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12543 generation for cases where splatting a value is not useful.
12544 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12545 across a vec_duplicate and a paradoxical subreg forming a vector
12546 mode to a vec_concat.
12548 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12550 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12551 -march=armv8.3-a variants.
12552 * config/arm/t-multilib: Likewise.
12553 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12555 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12557 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12559 (cceq_ior_compare_complement): Give it a name so I can use it, and
12560 change boolean_or_operator predicate to boolean_operator so it can
12561 be used to generate a crand.
12562 (eqne): New code iterator.
12563 (bd/bd_neg): New code_attrs.
12564 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12565 a single define_insn.
12566 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12567 decrement (bdnzt/bdnzf/bdzt/bdzf).
12568 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12569 with the new names of the branch decrement patterns, and added the
12570 names of the branch decrement conditional patterns.
12572 2018-01-08 Richard Biener <rguenther@suse.de>
12574 PR tree-optimization/83563
12575 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12578 2018-01-08 Richard Biener <rguenther@suse.de>
12580 PR middle-end/83713
12581 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12583 2018-01-08 Richard Biener <rguenther@suse.de>
12585 PR tree-optimization/83685
12586 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12587 references to abnormals.
12589 2018-01-08 Richard Biener <rguenther@suse.de>
12592 * dwarf2out.c (output_indirect_strings): Handle empty
12593 skeleton_debug_str_hash.
12594 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12596 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12598 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12599 (emit_store_direct): Likewise.
12600 (arc_trampoline_adjust_address): Likewise.
12601 (arc_asm_trampoline_template): New function.
12602 (arc_initialize_trampoline): Use asm_trampoline_template.
12603 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12604 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12605 * config/arc/arc.md (flush_icache): Delete pattern.
12607 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12609 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12610 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12613 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12616 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12617 by not USED_FOR_TARGET.
12618 (make_pass_resolve_sw_modes): Likewise.
12620 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12622 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12625 2018-01-08 Richard Biener <rguenther@suse.de>
12627 PR middle-end/83580
12628 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12630 2018-01-08 Richard Biener <rguenther@suse.de>
12632 PR middle-end/83517
12633 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12635 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12637 PR middle-end/81897
12638 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12639 basic blocks with a small number of successors.
12640 (convert_control_dep_chain_into_preds): Improve handling of
12642 (dump_predicates): Split apart into...
12643 (dump_pred_chain): ...here...
12644 (dump_pred_info): ...and here.
12645 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12646 (can_chain_union_be_invalidated_p): Improve check for invalidation
12648 (uninit_uses_cannot_happen): Avoid unnecessary if
12649 convert_control_dep_chain_into_preds yielded nothing.
12651 2018-01-06 Martin Sebor <msebor@redhat.com>
12653 PR tree-optimization/83640
12654 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12655 subtracting negative offset from size.
12656 (builtin_access::overlap): Adjust offset bounds of the access to fall
12657 within the size of the object if possible.
12659 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12661 PR rtl-optimization/83699
12662 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12663 extract_bit_field_as_subreg to cases in which the extracted
12664 value is also a vector.
12666 * lra-constraints.c (process_alt_operands): Test for the equivalence
12667 substitutions when detecting a possible reload cycle.
12669 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12672 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12673 by default if flag_selective_schedling{,2}. Formatting fixes.
12675 PR rtl-optimization/83682
12676 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12677 if it has non-VECTOR_MODE element mode.
12678 (vec_duplicate_p): Likewise.
12680 PR middle-end/83694
12681 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12682 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12684 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12687 * config/i386/i386-builtin.def
12688 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12689 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12690 Require also OPTION_MASK_ISA_AVX512F in addition to
12691 OPTION_MASK_ISA_GFNI.
12692 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12693 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12694 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12695 to OPTION_MASK_ISA_GFNI.
12696 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12697 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12698 OPTION_MASK_ISA_AVX512BW.
12699 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12700 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12701 addition to OPTION_MASK_ISA_GFNI.
12702 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12703 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12704 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12705 to OPTION_MASK_ISA_GFNI.
12706 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12707 a requirement for all ISAs rather than any of them with a few
12709 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12711 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12712 bitmasks to be enabled with 3 exceptions, instead of requiring any
12713 enabled ISA with lots of exceptions.
12714 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12715 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12716 Change avx512bw in isa attribute to avx512f.
12717 * config/i386/sgxintrin.h: Add license boilerplate.
12718 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12719 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12720 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12721 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12723 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12724 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12725 temporarily sse2 rather than sse if not enabled already.
12728 * config/i386/sse.md (VI248_VLBW): Rename to ...
12729 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12730 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12731 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12732 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12733 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12734 mode iterator instead of VI248_VLBW.
12736 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12738 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12739 (record_modified): Skip clobbers; add debug output.
12740 (param_change_prob): Use sreal frequencies.
12742 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12744 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12745 punt for user-aligned variables.
12747 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12749 * tree-chrec.c (chrec_contains_symbols): Return true for
12752 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12755 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12756 of (x|y) == x for BICS pattern.
12758 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12760 PR tree-optimization/83605
12761 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12762 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12765 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12767 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12768 * config/epiphany/rtems.h: New file.
12770 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12771 Uros Bizjak <ubizjak@gmail.com>
12774 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12775 QIreg_operand instead of register_operand predicate.
12776 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12777 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12778 comments instead of -fmitigate[-_]rop.
12780 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12783 * cgraphunit.c (symbol_table::compile): Switch to text_section
12784 before calling assembly_start debug hook.
12785 * run-rtl-passes.c (run_rtl_passes): Likewise.
12788 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12790 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12791 range_int_cst_p rather than !symbolic_range_p before calling
12792 extract_range_from_multiplicative_op_1.
12794 2018-01-04 Jeff Law <law@redhat.com>
12796 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12797 redundant test in assertion.
12799 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12801 * doc/rtl.texi: Document machine_mode wrapper classes.
12803 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12805 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12806 using tree_to_uhwi.
12808 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12810 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12811 the VEC_PERM_EXPR fold to fail.
12813 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12816 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12817 to switched_sections.
12819 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12822 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12823 test for d.testing.
12825 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12828 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12829 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12831 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12834 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12835 is BLKmode and bitpos not zero or mode change is needed.
12837 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12840 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12843 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12846 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12847 instead of MULT rtx. Update all corresponding splitters.
12848 (*saddl_se): Ditto.
12849 (*ssub<modesuffix>): Ditto.
12850 (*ssubl_se): Ditto.
12851 (*cmp_sadd_di): Update split patterns.
12852 (*cmp_sadd_si): Ditto.
12853 (*cmp_sadd_sidi): Ditto.
12854 (*cmp_ssub_di): Ditto.
12855 (*cmp_ssub_si): Ditto.
12856 (*cmp_ssub_sidi): Ditto.
12857 * config/alpha/predicates.md (const23_operand): New predicate.
12858 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12859 Look for ASHIFT, not MULT inner operand.
12860 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12862 2018-01-04 Martin Liska <mliska@suse.cz>
12864 PR gcov-profile/83669
12865 * gcov.c (output_intermediate_file): Add version to intermediate
12867 * doc/gcov.texi: Document new field 'version' in intermediate
12868 file format. Fix location of '-k' option of gcov command.
12870 2018-01-04 Martin Liska <mliska@suse.cz>
12873 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12875 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12877 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12879 2018-01-03 Martin Sebor <msebor@redhat.com>
12881 PR tree-optimization/83655
12882 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12883 checking calls with invalid arguments.
12885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12887 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12888 (vectorizable_mask_load_store): Delete.
12889 (vectorizable_call): Return false for masked loads and stores.
12890 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12891 instead of gimple_assign_rhs1.
12892 (vectorizable_load): Handle IFN_MASK_LOAD.
12893 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12897 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12899 (vectorizable_mask_load_store): ...here.
12900 (vectorizable_load): ...and here.
12902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12904 * tree-vect-stmts.c (vect_build_all_ones_mask)
12905 (vect_build_zero_merge_argument): New functions, split out from...
12906 (vectorizable_load): ...here.
12908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12910 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12912 (vectorizable_mask_load_store): ...here.
12913 (vectorizable_store): ...and here.
12915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12917 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12919 (vectorizable_mask_load_store): ...here.
12921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12923 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12924 (vect_model_store_cost): Take a vec_load_store_type instead of a
12926 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12927 (vect_model_store_cost): Take a vec_load_store_type instead of a
12929 (vectorizable_mask_load_store): Update accordingly.
12930 (vectorizable_store): Likewise.
12931 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12935 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12936 IFN_MASK_LOAD calls here rather than...
12937 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12940 Alan Hayward <alan.hayward@arm.com>
12941 David Sherwood <david.sherwood@arm.com>
12943 * expmed.c (extract_bit_field_1): For vector extracts,
12944 fall back to extract_bit_field_as_subreg if vec_extract
12947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12948 Alan Hayward <alan.hayward@arm.com>
12949 David Sherwood <david.sherwood@arm.com>
12951 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12952 they are variable or constant sized.
12953 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12954 slots for constant-sized data.
12956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12957 Alan Hayward <alan.hayward@arm.com>
12958 David Sherwood <david.sherwood@arm.com>
12960 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12961 handling COND_EXPRs with boolean comparisons, try to find a better
12962 basis for the mask type than the boolean itself.
12964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12966 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12967 is calculated and how it can be overridden.
12968 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12969 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12971 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12975 Alan Hayward <alan.hayward@arm.com>
12976 David Sherwood <david.sherwood@arm.com>
12978 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12979 Remove the mode argument.
12980 (aarch64_simd_valid_immediate): Remove the mode and inverse
12982 * config/aarch64/iterators.md (bitsize): New iterator.
12983 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12984 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12985 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12986 aarch64_simd_valid_immediate.
12987 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12988 (aarch64_reg_or_bic_imm): Likewise.
12989 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12990 with an insn_type enum and msl with a modifier_type enum.
12991 Replace element_width with a scalar_mode. Change the shift
12992 to unsigned int. Add constructors for scalar_float_mode and
12993 scalar_int_mode elements.
12994 (aarch64_vect_float_const_representable_p): Delete.
12995 (aarch64_can_const_movi_rtx_p)
12996 (aarch64_simd_scalar_immediate_valid_for_move)
12997 (aarch64_simd_make_constant): Update call to
12998 aarch64_simd_valid_immediate.
12999 (aarch64_advsimd_valid_immediate_hs): New function.
13000 (aarch64_advsimd_valid_immediate): Likewise.
13001 (aarch64_simd_valid_immediate): Remove mode and inverse
13002 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13003 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13004 and aarch64_float_const_representable_p on the result.
13005 (aarch64_output_simd_mov_immediate): Remove mode argument.
13006 Update call to aarch64_simd_valid_immediate and use of
13007 simd_immediate_info.
13008 (aarch64_output_scalar_simd_mov_immediate): Update call
13011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13012 Alan Hayward <alan.hayward@arm.com>
13013 David Sherwood <david.sherwood@arm.com>
13015 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13016 (mode_nunits): Likewise CONST_MODE_NUNITS.
13017 * machmode.def (ADJUST_NUNITS): Document.
13018 * genmodes.c (mode_data::need_nunits_adj): New field.
13019 (blank_mode): Update accordingly.
13020 (adj_nunits): New variable.
13021 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13023 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13024 listed in adj_nunits.
13025 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13026 listed in adj_nunits. Don't emit case statements for such modes.
13027 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13028 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13029 nothing if adj_nunits is nonnull.
13030 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13031 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13032 (emit_mode_fbit): Update use of print_maybe_const_decl.
13033 (emit_move_size): Likewise. Treat the array as non-const
13035 (emit_mode_adjustments): Handle adj_nunits.
13037 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13039 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13040 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13041 (VECTOR_MODES): Use it.
13042 (make_vector_modes): Take the prefix as an argument.
13044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13045 Alan Hayward <alan.hayward@arm.com>
13046 David Sherwood <david.sherwood@arm.com>
13048 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13049 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13050 for MODE_VECTOR_BOOL.
13051 * machmode.def (VECTOR_BOOL_MODE): Document.
13052 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13053 (make_vector_bool_mode): New function.
13054 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13056 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13057 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13059 * stor-layout.c (int_mode_for_mode): Likewise.
13060 * tree.c (build_vector_type_for_mode): Likewise.
13061 * varasm.c (output_constant_pool_2): Likewise.
13062 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13063 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13064 for MODE_VECTOR_BOOL.
13065 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13066 of mode class checks.
13067 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13068 instead of a list of mode class checks.
13069 (expand_vector_scalar_condition): Likewise.
13070 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13073 Alan Hayward <alan.hayward@arm.com>
13074 David Sherwood <david.sherwood@arm.com>
13076 * machmode.h (mode_size): Change from unsigned short to
13078 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13079 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13080 or if measurement_type is not polynomial.
13081 (fixed_size_mode::includes_p): Check for constant-sized modes.
13082 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13083 return a poly_uint16 rather than an unsigned short.
13084 (emit_mode_size): Change the type of mode_size from unsigned short
13085 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13086 (emit_mode_adjustments): Cope with polynomial vector sizes.
13087 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13089 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13091 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13092 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13093 * caller-save.c (setup_save_areas): Likewise.
13094 (replace_reg_with_saved_mem): Likewise.
13095 * calls.c (emit_library_call_value_1): Likewise.
13096 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13097 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13098 (gen_lowpart_for_combine): Likewise.
13099 * convert.c (convert_to_integer_1): Likewise.
13100 * cse.c (equiv_constant, cse_insn): Likewise.
13101 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13102 (cselib_subst_to_values): Likewise.
13103 * dce.c (word_dce_process_block): Likewise.
13104 * df-problems.c (df_word_lr_mark_ref): Likewise.
13105 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13106 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13107 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13108 (rtl_for_decl_location): Likewise.
13109 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13110 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13111 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13112 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13113 (expand_expr_real_1): Likewise.
13114 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13115 (pad_below): Likewise.
13116 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13117 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13118 * ira.c (get_subreg_tracking_sizes): Likewise.
13119 * ira-build.c (ira_create_allocno_objects): Likewise.
13120 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13121 (ira_sort_regnos_for_alter_reg): Likewise.
13122 * ira-costs.c (record_operand_costs): Likewise.
13123 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13124 (resolve_simple_move): Likewise.
13125 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13126 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13127 (lra_constraints): Likewise.
13128 (CONST_POOL_OK_P): Reject variable-sized modes.
13129 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13130 (add_pseudo_to_slot, lra_spill): Likewise.
13131 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13132 * optabs-query.c (get_best_extraction_insn): Likewise.
13133 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13134 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13135 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13136 * recog.c (offsettable_address_addr_space_p): Likewise.
13137 * regcprop.c (maybe_mode_change): Likewise.
13138 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13139 * regrename.c (build_def_use): Likewise.
13140 * regstat.c (dump_reg_info): Likewise.
13141 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13142 (find_reloads, find_reloads_subreg_address): Likewise.
13143 * reload1.c (eliminate_regs_1): Likewise.
13144 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13145 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13146 (simplify_binary_operation_1, simplify_subreg): Likewise.
13147 * targhooks.c (default_function_arg_padding): Likewise.
13148 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13149 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13150 (verify_gimple_assign_ternary): Likewise.
13151 * tree-inline.c (estimate_move_cost): Likewise.
13152 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13153 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13154 (get_address_cost_ainc): Likewise.
13155 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13156 (vect_supportable_dr_alignment): Likewise.
13157 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13158 (vectorizable_reduction): Likewise.
13159 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13160 (vectorizable_operation, vectorizable_load): Likewise.
13161 * tree.c (build_same_sized_truth_vector_type): Likewise.
13162 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13163 * var-tracking.c (emit_note_insn_var_location): Likewise.
13164 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13165 (ADDR_VEC_ALIGN): Likewise.
13167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13168 Alan Hayward <alan.hayward@arm.com>
13169 David Sherwood <david.sherwood@arm.com>
13171 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13173 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13174 or if measurement_type is polynomial.
13175 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13176 * combine.c (make_extraction): Likewise.
13177 * dse.c (find_shift_sequence): Likewise.
13178 * dwarf2out.c (mem_loc_descriptor): Likewise.
13179 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13180 (extract_bit_field, extract_low_bits): Likewise.
13181 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13182 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13183 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13184 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13185 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13186 * reload.c (find_reloads): Likewise.
13187 * reload1.c (alter_reg): Likewise.
13188 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13189 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13190 * tree-if-conv.c (predicate_mem_writes): Likewise.
13191 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13192 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13193 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13194 * valtrack.c (dead_debug_insert_temp): Likewise.
13195 * varasm.c (mergeable_constant_section): Likewise.
13196 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13199 Alan Hayward <alan.hayward@arm.com>
13200 David Sherwood <david.sherwood@arm.com>
13202 * expr.c (expand_assignment): Cope with polynomial mode sizes
13203 when assigning to a CONCAT.
13205 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13206 Alan Hayward <alan.hayward@arm.com>
13207 David Sherwood <david.sherwood@arm.com>
13209 * machmode.h (mode_precision): Change from unsigned short to
13211 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13213 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13214 or if measurement_type is not polynomial.
13215 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13216 in which the mode is already known to be a scalar_int_mode.
13217 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13218 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13220 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13221 for GET_MODE_PRECISION.
13222 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13223 for GET_MODE_PRECISION.
13224 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13226 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13227 (expand_field_assignment, make_extraction): Likewise.
13228 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13229 (get_last_value): Likewise.
13230 * convert.c (convert_to_integer_1): Likewise.
13231 * cse.c (cse_insn): Likewise.
13232 * expr.c (expand_expr_real_1): Likewise.
13233 * lra-constraints.c (simplify_operand_subreg): Likewise.
13234 * optabs-query.c (can_atomic_load_p): Likewise.
13235 * optabs.c (expand_atomic_load): Likewise.
13236 (expand_atomic_store): Likewise.
13237 * ree.c (combine_reaching_defs): Likewise.
13238 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13239 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13240 * tree.h (type_has_mode_precision_p): Likewise.
13241 * ubsan.c (instrument_si_overflow): Likewise.
13243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13244 Alan Hayward <alan.hayward@arm.com>
13245 David Sherwood <david.sherwood@arm.com>
13247 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13248 polynomial numbers of units.
13249 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13250 (valid_vector_subparts_p): New function.
13251 (build_vector_type): Remove temporary shim and take the number
13252 of units as a poly_uint64 rather than an int.
13253 (build_opaque_vector_type): Take the number of units as a
13254 poly_uint64 rather than an int.
13255 * tree.c (build_vector_from_ctor): Handle polynomial
13256 TYPE_VECTOR_SUBPARTS.
13257 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13258 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13259 (build_vector_from_val): If the number of units is variable,
13260 use build_vec_duplicate_cst for constant operands and
13261 VEC_DUPLICATE_EXPR otherwise.
13262 (make_vector_type): Remove temporary is_constant ().
13263 (build_vector_type, build_opaque_vector_type): Take the number of
13264 units as a poly_uint64 rather than an int.
13265 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13267 * cfgexpand.c (expand_debug_expr): Likewise.
13268 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13269 (store_constructor, expand_expr_real_1): Likewise.
13270 (const_scalar_mask_from_tree): Likewise.
13271 * fold-const-call.c (fold_const_reduction): Likewise.
13272 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13273 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13274 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13275 (fold_relational_const): Likewise.
13276 (native_interpret_vector): Likewise. Change the size from an
13277 int to an unsigned int.
13278 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13279 TYPE_VECTOR_SUBPARTS.
13280 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13281 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13282 duplicating a non-constant operand into a variable-length vector.
13283 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13284 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13285 * ipa-icf.c (sem_variable::equals): Likewise.
13286 * match.pd: Likewise.
13287 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13288 * print-tree.c (print_node): Likewise.
13289 * stor-layout.c (layout_type): Likewise.
13290 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13291 * tree-cfg.c (verify_gimple_comparison): Likewise.
13292 (verify_gimple_assign_binary): Likewise.
13293 (verify_gimple_assign_ternary): Likewise.
13294 (verify_gimple_assign_single): Likewise.
13295 * tree-pretty-print.c (dump_generic_node): Likewise.
13296 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13297 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13298 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13299 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13300 (vect_shift_permute_load_chain): Likewise.
13301 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13302 (expand_vector_condition, optimize_vector_constructor): Likewise.
13303 (lower_vec_perm, get_compute_type): Likewise.
13304 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13305 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13306 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13307 (vect_recog_mask_conversion_pattern): Likewise.
13308 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13309 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13310 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13311 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13312 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13313 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13314 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13315 (supportable_widening_operation): Likewise.
13316 (supportable_narrowing_operation): Likewise.
13317 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13319 * varasm.c (output_constant): Likewise.
13321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13322 Alan Hayward <alan.hayward@arm.com>
13323 David Sherwood <david.sherwood@arm.com>
13325 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13326 so that both the length == 3 and length != 3 cases set up their
13327 own permute vectors. Add comments explaining why we know the
13328 number of elements is constant.
13329 (vect_permute_load_chain): Likewise.
13331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13332 Alan Hayward <alan.hayward@arm.com>
13333 David Sherwood <david.sherwood@arm.com>
13335 * machmode.h (mode_nunits): Change from unsigned char to
13337 (ONLY_FIXED_SIZE_MODES): New macro.
13338 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13339 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13340 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13342 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13343 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13344 or if measurement_type is not polynomial.
13345 * genmodes.c (ZERO_COEFFS): New macro.
13346 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13348 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13349 Use ZERO_COEFFS when emitting initializers.
13350 * data-streamer.h (bp_pack_poly_value): New function.
13351 (bp_unpack_poly_value): Likewise.
13352 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13353 for GET_MODE_NUNITS.
13354 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13355 for GET_MODE_NUNITS.
13356 * tree.c (make_vector_type): Remove temporary shim and make
13357 the real function take the number of units as a poly_uint64
13358 rather than an int.
13359 (build_vector_type_for_mode): Handle polynomial nunits.
13360 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13361 * emit-rtl.c (const_vec_series_p_1): Likewise.
13362 (gen_rtx_CONST_VECTOR): Likewise.
13363 * fold-const.c (test_vec_duplicate_folding): Likewise.
13364 * genrecog.c (validate_pattern): Likewise.
13365 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13366 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13367 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13368 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13369 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13370 * rtlanal.c (subreg_get_info): Likewise.
13371 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13372 (vect_grouped_load_supported): Likewise.
13373 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13374 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13375 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13376 (simplify_const_unary_operation, simplify_binary_operation_1)
13377 (simplify_const_binary_operation, simplify_ternary_operation)
13378 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13379 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13380 instead of CONST_VECTOR_NUNITS.
13381 * varasm.c (output_constant_pool_2): Likewise.
13382 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13383 explicit-encoded elements in the XVEC for variable-length vectors.
13385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13387 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13390 Alan Hayward <alan.hayward@arm.com>
13391 David Sherwood <david.sherwood@arm.com>
13393 * coretypes.h (fixed_size_mode): Declare.
13394 (fixed_size_mode_pod): New typedef.
13395 * builtins.h (target_builtins::x_apply_args_mode)
13396 (target_builtins::x_apply_result_mode): Change type to
13397 fixed_size_mode_pod.
13398 * builtins.c (apply_args_size, apply_result_size, result_vector)
13399 (expand_builtin_apply_args_1, expand_builtin_apply)
13400 (expand_builtin_return): Update accordingly.
13402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13404 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13405 * cselib.c (cselib_hash_rtx): Likewise.
13406 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13407 CONST_VECTOR encoding.
13409 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13410 Jeff Law <law@redhat.com>
13413 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13414 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13415 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13416 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13419 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13420 explicitly probe *sp in a noreturn function if there were any callee
13421 register saves or frame pointer is needed.
13423 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13426 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13427 BLKmode for ternary, binary or unary expressions.
13430 * var-tracking.c (delete_vta_debug_insn): New inline function.
13431 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13432 insns from get_insns () to NULL instead of each bb separately.
13433 Use delete_vta_debug_insn. No longer static.
13434 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13435 delete_vta_debug_insns callers.
13436 * rtl.h (delete_vta_debug_insns): Declare.
13437 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13438 instead of variable_tracking_main.
13440 2018-01-03 Martin Sebor <msebor@redhat.com>
13442 PR tree-optimization/83603
13443 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13444 arguments past the endof the argument list in functions declared
13445 without a prototype.
13446 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13447 Avoid checking when arguments are null.
13449 2018-01-03 Martin Sebor <msebor@redhat.com>
13452 * doc/extend.texi (attribute const): Fix a typo.
13453 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13454 issuing -Wsuggest-attribute for void functions.
13456 2018-01-03 Martin Sebor <msebor@redhat.com>
13458 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13459 offset_int::from instead of wide_int::to_shwi.
13460 (maybe_diag_overlap): Remove assertion.
13461 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13462 * gimple-ssa-sprintf.c (format_directive): Same.
13463 (parse_directive): Same.
13464 (sprintf_dom_walker::compute_format_length): Same.
13465 (try_substitute_return_value): Same.
13467 2018-01-03 Jeff Law <law@redhat.com>
13469 PR middle-end/83654
13470 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13471 non-constant residual for zero at runtime and avoid probing in
13472 that case. Reorganize code for trailing problem to mirror handling
13475 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13477 PR tree-optimization/83501
13478 * tree-ssa-strlen.c (get_string_cst): New.
13479 (handle_char_store): Call get_string_cst.
13481 2018-01-03 Martin Liska <mliska@suse.cz>
13483 PR tree-optimization/83593
13484 * tree-ssa-strlen.c: Include tree-cfg.h.
13485 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13486 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13487 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13489 (strlen_dom_walker::before_dom_children): Call
13490 gimple_purge_dead_eh_edges. Dump tranformation with details
13492 (strlen_dom_walker::before_dom_children): Update call by adding
13493 new argument cleanup_eh.
13494 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13496 2018-01-03 Martin Liska <mliska@suse.cz>
13499 * cif-code.def (VARIADIC_THUNK): New enum value.
13500 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13503 2018-01-03 Jan Beulich <jbeulich@suse.com>
13505 * sse.md (mov<mode>_internal): Tighten condition for when to use
13506 vmovdqu<ssescalarsize> for TI and OI modes.
13508 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13510 Update copyright years.
13512 2018-01-03 Martin Liska <mliska@suse.cz>
13515 * ipa-visibility.c (function_and_variable_visibility): Skip
13516 functions with noipa attribure.
13518 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13520 * gcc.c (process_command): Update copyright notice dates.
13521 * gcov-dump.c (print_version): Ditto.
13522 * gcov.c (print_version): Ditto.
13523 * gcov-tool.c (print_version): Ditto.
13524 * gengtype.c (create_file): Ditto.
13525 * doc/cpp.texi: Bump @copying's copyright year.
13526 * doc/cppinternals.texi: Ditto.
13527 * doc/gcc.texi: Ditto.
13528 * doc/gccint.texi: Ditto.
13529 * doc/gcov.texi: Ditto.
13530 * doc/install.texi: Ditto.
13531 * doc/invoke.texi: Ditto.
13533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13535 * vector-builder.h (vector_builder::m_full_nelts): Change from
13536 unsigned int to poly_uint64.
13537 (vector_builder::full_nelts): Update prototype accordingly.
13538 (vector_builder::new_vector): Likewise.
13539 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13540 (vector_builder::operator ==): Likewise.
13541 (vector_builder::finalize): Likewise.
13542 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13543 Take the number of elements as a poly_uint64 rather than an
13545 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13546 from unsigned int to poly_uint64.
13547 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13548 (vec_perm_indices::new_vector): Likewise.
13549 (vec_perm_indices::length): Likewise.
13550 (vec_perm_indices::nelts_per_input): Likewise.
13551 (vec_perm_indices::input_nelts): Likewise.
13552 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13553 number of elements per input as a poly_uint64 rather than an
13554 unsigned int. Use the original encoding for variable-length
13555 vectors, rather than clamping each individual element.
13556 For the second and subsequent elements in each pattern,
13557 clamp the step and base before clamping their sum.
13558 (vec_perm_indices::series_p): Handle polynomial element counts.
13559 (vec_perm_indices::all_in_range_p): Likewise.
13560 (vec_perm_indices_to_tree): Likewise.
13561 (vec_perm_indices_to_rtx): Likewise.
13562 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13563 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13564 (tree_vector_builder::new_binary_operation): Handle polynomial
13565 element counts. Return false if we need to know the number
13566 of elements at compile time.
13567 * fold-const.c (fold_vec_perm): Punt if the number of elements
13568 isn't known at compile time.
13570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13572 * vec-perm-indices.h (vec_perm_builder): Change element type
13573 from HOST_WIDE_INT to poly_int64.
13574 (vec_perm_indices::element_type): Update accordingly.
13575 (vec_perm_indices::clamp): Handle polynomial element_types.
13576 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13577 (vec_perm_indices::all_in_range_p): Likewise.
13578 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13580 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13581 polynomial vec_perm_indices element types.
13582 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13583 * fold-const.c (fold_vec_perm): Likewise.
13584 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13585 * tree-vect-generic.c (lower_vec_perm): Likewise.
13586 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13587 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13588 element type to HOST_WIDE_INT.
13590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13591 Alan Hayward <alan.hayward@arm.com>
13592 David Sherwood <david.sherwood@arm.com>
13594 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13595 rather than an int. Use plus_constant.
13596 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13597 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13600 Alan Hayward <alan.hayward@arm.com>
13601 David Sherwood <david.sherwood@arm.com>
13603 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13604 a HOST_WIDE_INT to a poly_int64.
13606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13607 Alan Hayward <alan.hayward@arm.com>
13608 David Sherwood <david.sherwood@arm.com>
13610 * calls.c (load_register_parameters): Cope with polynomial
13611 mode sizes. Require a constant size for BLKmode parameters
13612 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13613 forces a parameter to be padded at the lsb end in order to
13614 fill a complete number of words, require the parameter size
13615 to be ordered wrt UNITS_PER_WORD.
13617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13618 Alan Hayward <alan.hayward@arm.com>
13619 David Sherwood <david.sherwood@arm.com>
13621 * reload1.c (spill_stack_slot_width): Change element type
13622 from unsigned int to poly_uint64_pod.
13623 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
13630 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
13637 operands whose sizes aren't ordered.
13638 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13645 that the mode size is in the set {1, 2, 4, 8, 16}.
13647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13648 Alan Hayward <alan.hayward@arm.com>
13649 David Sherwood <david.sherwood@arm.com>
13651 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13652 Use plus_constant instead of gen_rtx_PLUS.
13654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13655 Alan Hayward <alan.hayward@arm.com>
13656 David Sherwood <david.sherwood@arm.com>
13658 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13659 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13660 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13661 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13662 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13663 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13664 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13665 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13666 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13667 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13669 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13670 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13671 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13672 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13673 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13674 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13675 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13676 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13677 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13678 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13680 * expr.c (emit_move_resolve_push): Treat the input and result
13681 of PUSH_ROUNDING as a poly_int64.
13682 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13683 (emit_push_insn): Likewise.
13684 * lra-eliminations.c (mark_not_eliminable): Likewise.
13685 * recog.c (push_operand): Likewise.
13686 * reload1.c (elimination_effects): Likewise.
13687 * rtlanal.c (nonzero_bits1): Likewise.
13688 * calls.c (store_one_arg): Likewise. Require the padding to be
13689 known at compile time.
13691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13692 Alan Hayward <alan.hayward@arm.com>
13693 David Sherwood <david.sherwood@arm.com>
13695 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13696 Use plus_constant instead of gen_rtx_PLUS.
13698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13699 Alan Hayward <alan.hayward@arm.com>
13700 David Sherwood <david.sherwood@arm.com>
13702 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13703 rather than an int.
13705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13706 Alan Hayward <alan.hayward@arm.com>
13707 David Sherwood <david.sherwood@arm.com>
13709 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13710 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13711 via stack temporaries. Treat the mode size as polynomial too.
13713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13714 Alan Hayward <alan.hayward@arm.com>
13715 David Sherwood <david.sherwood@arm.com>
13717 * expr.c (expand_expr_real_2): When handling conversions involving
13718 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13719 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13720 as a poly_uint64 too.
13722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13723 Alan Hayward <alan.hayward@arm.com>
13724 David Sherwood <david.sherwood@arm.com>
13726 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
13733 REGMODE_NATURAL_SIZE.
13734 * expmed.c (store_bit_field_1): Likewise.
13735 * expr.c (store_constructor): Likewise.
13736 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13737 and polynomial REGMODE_NATURAL_SIZE.
13738 (gen_lowpart_common): Likewise.
13739 * reginfo.c (record_subregs_of_mode): Likewise.
13740 * rtlanal.c (read_modify_subreg_p): Likewise.
13742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13743 Alan Hayward <alan.hayward@arm.com>
13744 David Sherwood <david.sherwood@arm.com>
13746 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13747 numbers of elements.
13749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13750 Alan Hayward <alan.hayward@arm.com>
13751 David Sherwood <david.sherwood@arm.com>
13753 * match.pd: Cope with polynomial numbers of vector elements.
13755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13756 Alan Hayward <alan.hayward@arm.com>
13757 David Sherwood <david.sherwood@arm.com>
13759 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13760 in a POINTER_PLUS_EXPR.
13762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13763 Alan Hayward <alan.hayward@arm.com>
13764 David Sherwood <david.sherwood@arm.com>
13766 * omp-simd-clone.c (simd_clone_subparts): New function.
13767 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13768 (ipa_simd_modify_function_body): Likewise.
13770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13771 Alan Hayward <alan.hayward@arm.com>
13772 David Sherwood <david.sherwood@arm.com>
13774 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13775 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13776 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13777 (expand_vector_condition, vector_element): Likewise.
13778 (subparts_gt): New function.
13779 (get_compute_type): Use subparts_gt.
13780 (count_type_subparts): Delete.
13781 (expand_vector_operations_1): Use subparts_gt instead of
13782 count_type_subparts.
13784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13785 Alan Hayward <alan.hayward@arm.com>
13786 David Sherwood <david.sherwood@arm.com>
13788 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13789 (vect_compile_time_alias): ...this new function. Do the calculation
13790 on poly_ints rather than trees.
13791 (vect_prune_runtime_alias_test_list): Update call accordingly.
13793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13794 Alan Hayward <alan.hayward@arm.com>
13795 David Sherwood <david.sherwood@arm.com>
13797 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13799 (vect_schedule_slp_instance): Likewise.
13801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13802 Alan Hayward <alan.hayward@arm.com>
13803 David Sherwood <david.sherwood@arm.com>
13805 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13806 constant and extern definitions for variable-length vectors.
13807 (vect_get_constant_vectors): Note that the number of units
13808 is known to be constant.
13810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13811 Alan Hayward <alan.hayward@arm.com>
13812 David Sherwood <david.sherwood@arm.com>
13814 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13815 of units as polynomial. Choose between WIDE and NARROW based
13818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13819 Alan Hayward <alan.hayward@arm.com>
13820 David Sherwood <david.sherwood@arm.com>
13822 * tree-vect-stmts.c (simd_clone_subparts): New function.
13823 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13826 Alan Hayward <alan.hayward@arm.com>
13827 David Sherwood <david.sherwood@arm.com>
13829 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13830 vectors as polynomial. Use build_index_vector for
13831 IFN_GOMP_SIMD_LANE.
13833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13834 Alan Hayward <alan.hayward@arm.com>
13835 David Sherwood <david.sherwood@arm.com>
13837 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13838 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13839 for variable-length vectors.
13840 (vectorizable_mask_load_store): Treat the number of units as
13841 polynomial, asserting that it is constant if the condition has
13842 already been enforced.
13843 (vectorizable_store, vectorizable_load): Likewise.
13845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13846 Alan Hayward <alan.hayward@arm.com>
13847 David Sherwood <david.sherwood@arm.com>
13849 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13850 of units as polynomial. Punt if we can't tell at compile time
13851 which vector contains the final result.
13853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13854 Alan Hayward <alan.hayward@arm.com>
13855 David Sherwood <david.sherwood@arm.com>
13857 * tree-vect-loop.c (vectorizable_induction): Treat the number
13858 of units as polynomial. Punt on SLP inductions. Use an integer
13859 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13860 cast of such a series for variable-length floating-point
13863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13864 Alan Hayward <alan.hayward@arm.com>
13865 David Sherwood <david.sherwood@arm.com>
13867 * tree.h (build_index_vector): Declare.
13868 * tree.c (build_index_vector): New function.
13869 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13870 of units as polynomial, forcibly converting it to a constant if
13871 vectorizable_reduction has already enforced the condition.
13872 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13873 to create a {1,2,3,...} vector.
13874 (vectorizable_reduction): Treat the number of units as polynomial.
13875 Choose vectype_in based on the largest scalar element size rather
13876 than the smallest number of units. Enforce the restrictions
13879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13880 Alan Hayward <alan.hayward@arm.com>
13881 David Sherwood <david.sherwood@arm.com>
13883 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13884 number of units as polynomial.
13886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13887 Alan Hayward <alan.hayward@arm.com>
13888 David Sherwood <david.sherwood@arm.com>
13890 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13891 * target.def (autovectorize_vector_sizes): Return the vector sizes
13892 by pointer, using vector_sizes rather than a bitmask.
13893 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13894 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13895 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13897 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13898 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13899 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13900 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13901 * omp-general.c (omp_max_vf): Likewise.
13902 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13903 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13904 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13905 * tree-vect-slp.c (vect_slp_bb): Likewise.
13906 * doc/tm.texi: Regenerate.
13907 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13909 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13910 the vector size as a poly_uint64 rather than an unsigned int.
13911 (current_vector_size): Change from an unsigned int to a poly_uint64.
13912 (get_vectype_for_scalar_type): Update accordingly.
13913 * tree.h (build_truth_vector_type): Take the size and number of
13914 units as a poly_uint64 rather than an unsigned int.
13915 (build_vector_type): Add a temporary overload that takes
13916 the number of units as a poly_uint64 rather than an unsigned int.
13917 * tree.c (make_vector_type): Likewise.
13918 (build_truth_vector_type): Take the number of units as a poly_uint64
13919 rather than an unsigned int.
13921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13922 Alan Hayward <alan.hayward@arm.com>
13923 David Sherwood <david.sherwood@arm.com>
13925 * target.def (get_mask_mode): Take the number of units and length
13926 as poly_uint64s rather than unsigned ints.
13927 * targhooks.h (default_get_mask_mode): Update accordingly.
13928 * targhooks.c (default_get_mask_mode): Likewise.
13929 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13930 * doc/tm.texi: Regenerate.
13932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13933 Alan Hayward <alan.hayward@arm.com>
13934 David Sherwood <david.sherwood@arm.com>
13936 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13937 * omp-general.c (omp_max_vf): Likewise.
13938 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13939 (expand_omp_simd): Handle polynomial safelen.
13940 * omp-low.c (omplow_simd_context): Add a default constructor.
13941 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13942 (lower_rec_simd_input_clauses): Update accordingly.
13943 (lower_rec_input_clauses): Likewise.
13945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13946 Alan Hayward <alan.hayward@arm.com>
13947 David Sherwood <david.sherwood@arm.com>
13949 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13950 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13951 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13952 (vect_analyze_slp_cost): Likewise.
13953 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13954 (vect_model_load_cost): Likewise.
13956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13957 Alan Hayward <alan.hayward@arm.com>
13958 David Sherwood <david.sherwood@arm.com>
13960 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13961 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13962 from an unsigned int * to a poly_uint64_pod *.
13963 (calculate_unrolling_factor): New function.
13964 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13967 Alan Hayward <alan.hayward@arm.com>
13968 David Sherwood <david.sherwood@arm.com>
13970 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13971 from an unsigned int to a poly_uint64.
13972 (_loop_vec_info::slp_unrolling_factor): Likewise.
13973 (_loop_vec_info::vectorization_factor): Change from an int
13975 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13976 (vect_get_num_vectors): New function.
13977 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13978 (vect_get_num_copies): Use vect_get_num_vectors.
13979 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13980 to an unsigned int *.
13981 (vect_analyze_data_refs): Change min_vf from an int * to a
13983 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13984 than an unsigned HOST_WIDE_INT.
13985 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13986 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13987 to an unsigned int *.
13988 (vect_analyze_data_ref_dependences): Likewise.
13989 (vect_compute_data_ref_alignment): Handle polynomial vf.
13990 (vect_enhance_data_refs_alignment): Likewise.
13991 (vect_prune_runtime_alias_test_list): Likewise.
13992 (vect_shift_permute_load_chain): Likewise.
13993 (vect_supportable_dr_alignment): Likewise.
13994 (dependence_distance_ge_vf): Take the vectorization factor as a
13995 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13996 (vect_analyze_data_refs): Change min_vf from an int * to a
13998 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13999 vfm1 as a poly_uint64 rather than an int. Make the same change
14000 for the returned bound_scalar.
14001 (vect_gen_vector_loop_niters): Handle polynomial vf.
14002 (vect_do_peeling): Likewise. Update call to
14003 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14004 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14006 * tree-vect-loop.c (vect_determine_vectorization_factor)
14007 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14008 (vect_get_known_peeling_cost): Likewise.
14009 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14010 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14011 (vect_transform_loop): Likewise. Use the lowest possible VF when
14012 updating the upper bounds of the loop.
14013 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14014 rather than an int.
14015 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14016 polynomial unroll factors.
14017 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14018 (vect_make_slp_decision): Likewise.
14019 (vect_supported_load_permutation_p): Likewise, and polynomial
14021 (vect_analyze_slp_cost): Handle polynomial vf.
14022 (vect_slp_analyze_node_operations): Likewise.
14023 (vect_slp_analyze_bb_1): Likewise.
14024 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14025 than an unsigned HOST_WIDE_INT.
14026 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14027 (vectorizable_load): Handle polynomial vf.
14028 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14030 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14033 Alan Hayward <alan.hayward@arm.com>
14034 David Sherwood <david.sherwood@arm.com>
14036 * match.pd: Handle bit operations involving three constants
14037 and try to fold one pair.
14039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14041 * tree-vect-loop-manip.c: Include gimple-fold.h.
14042 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14043 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14044 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14045 Add a path that uses a step of VF instead of 1, but disable it
14047 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14048 and niters_no_overflow parameters. Update calls to
14049 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14050 Create a new SSA name if the latter choses to use a ste other
14051 than zero, and return it via niters_vector_mult_vf_var.
14052 * tree-vect-loop.c (vect_transform_loop): Update calls to
14053 vect_do_peeling, vect_gen_vector_loop_niters and
14054 slpeel_make_loop_iterate_ntimes.
14055 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14056 (vect_gen_vector_loop_niters): Update declarations after above changes.
14058 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14060 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14061 128-bit round to integer instructions.
14062 (ceil<mode>2): Likewise.
14063 (btrunc<mode>2): Likewise.
14064 (round<mode>2): Likewise.
14066 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14068 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14069 unaligned VSX load/store on P8/P9.
14070 (expand_block_clear): Allow the use of unaligned VSX
14071 load/store on P8/P9.
14073 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14075 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14077 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14078 swap associated with both a load and a store.
14080 2018-01-02 Andrew Waterman <andrew@sifive.com>
14082 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14083 * config/riscv/riscv.md (clear_cache): Use it.
14085 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14087 * web.c: Remove out-of-date comment.
14089 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14091 * expr.c (fixup_args_size_notes): Check that any existing
14092 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14093 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14094 (emit_single_push_insn): ...here.
14096 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14098 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14099 (const_vector_encoded_nelts): New function.
14100 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14101 (const_vector_int_elt, const_vector_elt): Declare.
14102 * emit-rtl.c (const_vector_int_elt_1): New function.
14103 (const_vector_elt): Likewise.
14104 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14105 of CONST_VECTOR_ELT.
14107 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14109 * expr.c: Include rtx-vector-builder.h.
14110 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14111 directly on the tree encoding.
14112 (const_vector_from_tree): Likewise.
14113 * optabs.c: Include rtx-vector-builder.h.
14114 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14115 sequence of "u" values.
14116 * vec-perm-indices.c: Include rtx-vector-builder.h.
14117 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14118 directly on the vec_perm_indices encoding.
14120 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14122 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14123 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14124 * rtx-vector-builder.h: New file.
14125 * rtx-vector-builder.c: Likewise.
14126 * rtl.h (rtx_def::u2): Add a const_vector field.
14127 (CONST_VECTOR_NPATTERNS): New macro.
14128 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14129 (CONST_VECTOR_DUPLICATE_P): Likewise.
14130 (CONST_VECTOR_STEPPED_P): Likewise.
14131 (CONST_VECTOR_ENCODED_ELT): Likewise.
14132 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14133 (unwrap_const_vec_duplicate): Likewise.
14134 (const_vec_series_p): Check for a non-duplicated vector encoding.
14135 Say that the function only returns true for integer vectors.
14136 * emit-rtl.c: Include rtx-vector-builder.h.
14137 (gen_const_vec_duplicate_1): Delete.
14138 (gen_const_vector): Call gen_const_vec_duplicate instead of
14139 gen_const_vec_duplicate_1.
14140 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14141 (gen_const_vec_duplicate): Use rtx_vector_builder.
14142 (gen_const_vec_series): Likewise.
14143 (gen_rtx_CONST_VECTOR): Likewise.
14144 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14145 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14146 Build a new vector rather than modifying a CONST_VECTOR in-place.
14147 (handle_special_swappables): Update call accordingly.
14148 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14149 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14150 Build a new vector rather than modifying a CONST_VECTOR in-place.
14151 (handle_special_swappables): Update call accordingly.
14153 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14155 * simplify-rtx.c (simplify_const_binary_operation): Use
14156 CONST_VECTOR_ELT instead of XVECEXP.
14158 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14160 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14161 the selector elements to be different from the data elements
14162 if the selector is a VECTOR_CST.
14163 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14164 ssizetype for the selector.
14166 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14168 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14169 before testing each element individually.
14170 * tree-vect-generic.c (lower_vec_perm): Likewise.
14172 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14174 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14175 * selftest-run-tests.c (selftest::run_tests): Call it.
14176 * vector-builder.h (vector_builder::operator ==): New function.
14177 (vector_builder::operator !=): Likewise.
14178 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14179 (vec_perm_indices::all_from_input_p): New function.
14180 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14181 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14182 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14183 instead of reading the VECTOR_CST directly. Detect whether both
14184 vector inputs are the same before constructing the vec_perm_indices,
14185 and update the number of inputs argument accordingly. Use the
14186 utility functions added above. Only construct sel2 if we need to.
14188 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14190 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14191 the broadcast of the low byte.
14192 (expand_mult_highpart): Use an explicit encoding for the permutes.
14193 * optabs-query.c (can_mult_highpart_p): Likewise.
14194 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14195 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14196 (vectorizable_bswap): Likewise.
14197 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14198 explicit encoding for the power-of-2 permutes.
14199 (vect_permute_store_chain): Likewise.
14200 (vect_grouped_load_supported): Likewise.
14201 (vect_permute_load_chain): Likewise.
14203 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14205 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14206 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14207 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14208 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14209 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14210 (vect_gen_perm_mask_any): Likewise.
14212 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14214 * int-vector-builder.h: New file.
14215 * vec-perm-indices.h: Include int-vector-builder.h.
14216 (vec_perm_indices): Redefine as an int_vector_builder.
14217 (auto_vec_perm_indices): Delete.
14218 (vec_perm_builder): Redefine as a stand-alone class.
14219 (vec_perm_indices::vec_perm_indices): New function.
14220 (vec_perm_indices::clamp): Likewise.
14221 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14222 (vec_perm_indices::new_vector): New function.
14223 (vec_perm_indices::new_expanded_vector): Update for new
14224 vec_perm_indices class.
14225 (vec_perm_indices::rotate_inputs): New function.
14226 (vec_perm_indices::all_in_range_p): Operate directly on the
14227 encoded form, without computing elided elements.
14228 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14229 encoding. Update for new vec_perm_indices class.
14230 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14231 the given vec_perm_builder.
14232 (expand_vec_perm_var): Update vec_perm_builder constructor.
14233 (expand_mult_highpart): Use vec_perm_builder instead of
14234 auto_vec_perm_indices.
14235 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14236 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14237 or double series encoding as appropriate.
14238 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14239 vec_perm_indices instead of auto_vec_perm_indices.
14240 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14241 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14242 (vect_permute_store_chain): Likewise.
14243 (vect_grouped_load_supported): Likewise.
14244 (vect_permute_load_chain): Likewise.
14245 (vect_shift_permute_load_chain): Likewise.
14246 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14247 (vect_transform_slp_perm_load): Likewise.
14248 (vect_schedule_slp_instance): Likewise.
14249 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14250 (vectorizable_mask_load_store): Likewise.
14251 (vectorizable_bswap): Likewise.
14252 (vectorizable_store): Likewise.
14253 (vectorizable_load): Likewise.
14254 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14255 vec_perm_indices instead of auto_vec_perm_indices. Use
14256 tree_to_vec_perm_builder to read the vector from a tree.
14257 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14258 vec_perm_builder instead of a vec_perm_indices.
14259 (have_whole_vector_shift): Use vec_perm_builder and
14260 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14261 truncation to calc_vec_perm_mask_for_shift.
14262 (vect_create_epilog_for_reduction): Likewise.
14263 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14264 from auto_vec_perm_indices to vec_perm_indices.
14265 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14266 instead of changing individual elements.
14267 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14268 the vector in d.perm.
14269 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14270 from auto_vec_perm_indices to vec_perm_indices.
14271 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14272 instead of changing individual elements.
14273 (arm_vectorize_vec_perm_const): Use new_vector to install
14274 the vector in d.perm.
14275 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14276 Update vec_perm_builder constructor.
14277 (rs6000_expand_interleave): Likewise.
14278 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14279 (rs6000_expand_interleave): Likewise.
14281 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14283 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14284 to qimode could truncate the indices.
14285 * optabs.c (expand_vec_perm_var): Likewise.
14287 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14289 * Makefile.in (OBJS): Add vec-perm-indices.o.
14290 * vec-perm-indices.h: New file.
14291 * vec-perm-indices.c: Likewise.
14292 * target.h (vec_perm_indices): Replace with a forward class
14294 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14295 * optabs.h: Include vec-perm-indices.h.
14296 (expand_vec_perm): Delete.
14297 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14298 (expand_vec_perm_const): Declare.
14299 * target.def (vec_perm_const_ok): Replace with...
14300 (vec_perm_const): ...this new hook.
14301 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14302 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14303 * doc/tm.texi: Regenerate.
14304 * optabs.def (vec_perm_const): Delete.
14305 * doc/md.texi (vec_perm_const): Likewise.
14306 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14307 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14308 expand_vec_perm for constant permutation vectors. Assert that
14309 the mode of variable permutation vectors is the integer equivalent
14310 of the mode that is being permuted.
14311 * optabs-query.h (selector_fits_mode_p): Declare.
14312 * optabs-query.c: Include vec-perm-indices.h.
14313 (selector_fits_mode_p): New function.
14314 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14315 is defined, instead of checking whether the vec_perm_const_optab
14316 exists. Use targetm.vectorize.vec_perm_const instead of
14317 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14318 fit in the vector mode before using a variable permute.
14319 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14320 vec_perm_indices instead of an rtx.
14321 (expand_vec_perm): Replace with...
14322 (expand_vec_perm_const): ...this new function. Take the selector
14323 as a vec_perm_indices rather than an rtx. Also take the mode of
14324 the selector. Update call to shift_amt_for_vec_perm_mask.
14325 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14326 Use vec_perm_indices::new_expanded_vector to expand the original
14327 selector into bytes. Check whether the indices fit in the vector
14328 mode before using a variable permute.
14329 (expand_vec_perm_var): Make global.
14330 (expand_mult_highpart): Use expand_vec_perm_const.
14331 * fold-const.c: Includes vec-perm-indices.h.
14332 * tree-ssa-forwprop.c: Likewise.
14333 * tree-vect-data-refs.c: Likewise.
14334 * tree-vect-generic.c: Likewise.
14335 * tree-vect-loop.c: Likewise.
14336 * tree-vect-slp.c: Likewise.
14337 * tree-vect-stmts.c: Likewise.
14338 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14340 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14341 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14342 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14343 (aarch64_vectorize_vec_perm_const): ...this new function.
14344 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14345 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14346 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14347 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14348 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14349 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14350 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14352 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14353 check for NEON modes.
14354 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14355 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14356 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14357 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14359 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14360 the old VEC_PERM_CONST conditions.
14361 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14362 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14363 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14364 (ia64_vectorize_vec_perm_const_ok): Merge into...
14365 (ia64_vectorize_vec_perm_const): ...this new function.
14366 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14367 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14368 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14369 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14370 * config/mips/mips.c (mips_expand_vec_perm_const)
14371 (mips_vectorize_vec_perm_const_ok): Merge into...
14372 (mips_vectorize_vec_perm_const): ...this new function.
14373 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14374 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14375 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14376 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14377 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14378 (rs6000_expand_vec_perm_const): Delete.
14379 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14381 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14382 (altivec_expand_vec_perm_const_le): Take each operand individually.
14383 Operate on constant selectors rather than rtxes.
14384 (altivec_expand_vec_perm_const): Likewise. Update call to
14385 altivec_expand_vec_perm_const_le.
14386 (rs6000_expand_vec_perm_const): Delete.
14387 (rs6000_vectorize_vec_perm_const_ok): Delete.
14388 (rs6000_vectorize_vec_perm_const): New function.
14389 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14390 an element count and rtx array.
14391 (rs6000_expand_extract_even): Update call accordingly.
14392 (rs6000_expand_interleave): Likewise.
14393 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14394 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14395 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14396 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14397 (rs6000_expand_vec_perm_const): Delete.
14398 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14399 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14400 (altivec_expand_vec_perm_const_le): Take each operand individually.
14401 Operate on constant selectors rather than rtxes.
14402 (altivec_expand_vec_perm_const): Likewise. Update call to
14403 altivec_expand_vec_perm_const_le.
14404 (rs6000_expand_vec_perm_const): Delete.
14405 (rs6000_vectorize_vec_perm_const_ok): Delete.
14406 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14407 reference to the SPE evmerge intructions.
14408 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14409 an element count and rtx array.
14410 (rs6000_expand_extract_even): Update call accordingly.
14411 (rs6000_expand_interleave): Likewise.
14412 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14413 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14415 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14417 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14419 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14420 vector mode and that that mode matches the mode of the data
14422 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14423 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14424 directly using expand_vec_perm_1 when forcing selectors into
14426 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14428 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14430 * optabs-query.h (can_vec_perm_p): Delete.
14431 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14432 * optabs-query.c (can_vec_perm_p): Split into...
14433 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14434 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14435 particular selector is valid.
14436 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14437 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14438 (vect_grouped_load_supported): Likewise.
14439 (vect_shift_permute_load_chain): Likewise.
14440 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14441 (vect_transform_slp_perm_load): Likewise.
14442 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14443 (vectorizable_bswap): Likewise.
14444 (vect_gen_perm_mask_checked): Likewise.
14445 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14446 implementations of variable permutation vectors into account
14447 when deciding which selector to use.
14448 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14449 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14450 with a false third argument.
14451 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14452 to test whether the constant selector is valid and can_vec_perm_var_p
14453 to test whether a variable selector is valid.
14455 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14457 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14458 * optabs-query.c (can_vec_perm_p): Likewise.
14459 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14460 instead of vec_perm_indices.
14461 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14462 (vect_gen_perm_mask_checked): Likewise,
14463 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14464 (vect_gen_perm_mask_checked): Likewise,
14466 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14468 * optabs-query.h (qimode_for_vec_perm): Declare.
14469 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14470 (qimode_for_vec_perm): ...this new function.
14471 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14473 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14475 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14476 does not have a conditional at the top.
14478 2018-01-02 Richard Biener <rguenther@suse.de>
14480 * ipa-inline.c (big_speedup_p): Fix expression.
14482 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14485 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14488 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14492 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14493 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14494 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14495 cond_taken_branch_cost 3->4.
14497 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14499 PR tree-optimization/83581
14500 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14501 TODO_cleanup_cfg if any changes have been made.
14503 PR middle-end/83608
14504 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14505 convert_modes if target mode has the right side, but different mode
14508 PR middle-end/83609
14509 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14510 last argument when extracting from CONCAT. If either from_real or
14511 from_imag is NULL, use expansion through memory. If result is not
14512 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14513 the parts directly to inner mode, if even that fails, use expansion
14516 PR middle-end/83623
14517 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14518 check for bswap in mode rather than HImode and use that in expand_unop
14521 Copyright (C) 2018 Free Software Foundation, Inc.
14523 Copying and distribution of this file, with or without modification,
14524 are permitted in any medium without royalty provided the copyright
14525 notice and this notice are preserved.