1 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
6 2018-05-28 Mark Wielaard <mark@klomp.org>
8 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
9 if it is an expression containing a minus sign.
11 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
13 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
15 2018-05-27 Paul Koning <ni1d@arrl.net>
17 * config/pdp11/pdp11.md (truncsihi2): Remove.
19 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
20 Chung-Ju Wu <jasonwucj@gmail.com>
22 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
24 (unaligned_store_dw): Ditto.
25 * config/nds32/nds32-memory-manipulation.c
26 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
27 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
28 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
29 (emit_setmem_word_loop): Rename to ...
30 (emit_setmem_doubleword_loop): ... this.
31 (nds32_gen_dup_4_byte_to_word_value): New function.
32 (nds32_gen_dup_8_byte_to_double_word_value): New function.
33 (nds32_expand_setmem_loop): Refine implementation.
34 (nds32_expand_setmem_loop_v3m): Ditto.
35 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
38 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
40 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
42 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
44 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
45 (nds32_init_machine_status): Initialize machine->attr_naked_p and
46 machine->attr_no_prologue_p.
47 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
48 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
49 (nds32_expand_epilogue): Consider attr_naked_p.
50 (nds32_expand_epilogue_v3pop): Likewise.
51 (nds32_can_use_return_insn): Likewise.
52 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
53 attr_no_prologue_p fields.
54 * config/nds32/nds32.opt (mret-in-naked-func): New option.
56 2018-05-27 Jakub Jelinek <jakub@redhat.com>
59 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
62 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
64 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
66 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
68 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
70 (*<floatsuffix>floatv2div2sf2): Rename to ...
71 (*float<floatunssuffix>v2div2sf2): ... this.
72 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
73 (float<floatunssuffix>v2div2sf2_mask): ... this.
74 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
75 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
76 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
78 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
80 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
82 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
84 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
86 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
88 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
89 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
90 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
91 gen_ufix_truncv8dfv8si2.
92 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
93 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
94 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
95 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
96 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
97 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
98 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
99 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
101 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
105 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
107 2018-05-25 Jim Wilson <jimw@sifive.com>
109 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
110 * config/riscv/riscv.c (struct machine_function): Add
111 interrupt_handler_p and attribute_checked_p fields.
112 (riscv_attribute_table): Add interrupt.
113 (riscv_interrupt_type_p): New.
114 (riscv_save_reg_p): Save extra regs for interrupt handler.
115 (riscv_use_save_libcall): Return false for interrupt handler.
116 (riscv_first_stack_step): Add forward declaration.
117 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
118 for interrupt handler with large frame. Use it for saved reg list.
119 (riscv_expand_prologue): Move flag_stack_usage_info support to
120 eliminate duplication.
121 (riscv_expand_epilogue): Generate mret for interrupt handler.
122 (riscv_epilogue_uses): New.
123 (riscv_can_use_return_insn): Return false for interrupt handler.
124 (riscv_function_ok_for_sibcall): Likewise.
125 (riscv_set_current_function): Add interrupt handler support.
126 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
127 * config/riscv/riscv.md (UNSPECV_MRET): New.
129 (riscv_frflags, riscv_fsflags): Use tab after opcode.
131 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
133 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
135 PR tree-optimization/85712
136 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
137 this candidate has already been replaced in-situ by a copy.
139 2018-05-25 Jason Merrill <jason@redhat.com>
141 PR c++/80485 - inline function non-zero address.
142 * symtab.c (nonzero_address): Check DECL_COMDAT.
144 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
147 * config/alpha/alpha.md (ashlsi3): New insn pattern.
148 (*ashlsi_se): Rename from *ashldi_se. Define as sign
149 extension of SImode operation. Use const123_operand predicate.
151 (*saddl_se_1): Ditto.
153 (*ssubl_se_1): Ditto.
154 * config/alpha/predicates.md (const123_operand): New predicate.
155 * config/alpha/constraints.md (P): Use IN_RANGE.
157 2018-05-25 Richard Biener <rguenther@suse.de>
159 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
161 (ref_maybe_used_by_stmt_p): Likewise.
162 (stmt_may_clobber_ref_p): Likewise.
163 (stmt_may_clobber_ref_p_1): Likewise.
164 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
166 (ref_maybe_used_by_stmt_p): Likewise.
167 (stmt_may_clobber_ref_p): Likewise.
168 (stmt_may_clobber_ref_p_1): Likewise.
169 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
170 the alias oracle to disambiguate DRs with stmts DR analysis
172 (vect_analyze_data_refs): Do not give up on not analyzable
173 DRs for BB vectorization. Remove code truncating the dataref
176 2018-05-25 Jakub Jelinek <jakub@redhat.com>
179 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
180 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
181 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
183 2018-05-25 Richard Biener <rguenther@suse.de>
185 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
186 function, combining stmt data ref gathering and fatal analysis
188 (vect_analyze_data_refs): Remove now redudnant code and simplify.
189 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
190 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
191 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
192 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
194 2018-05-25 Bin Cheng <bin.cheng@arm.com>
196 PR tree-optimization/85720
197 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
198 SCC if all partitions are builtins.
199 (version_loop_by_alias_check): New parameter. Generate cancelable
200 runtime alias check if all partitions are builtins.
201 (distribute_loop): Update call to above function.
203 2018-05-25 Bin Cheng <bin.cheng@arm.com>
205 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
206 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
207 (parm_default_def_partition_arg): Ditto.
208 (set_parm_default_def_partition): Ditto.
209 (get_parm_default_def_partitions): Ditto and make it static.
210 (get_undefined_value_partitions): Ditto and make it static.
211 (remove_ssa_form): Refactor call to init_var_map here.
212 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
213 computation for loop region.
214 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
215 (register_default_def): Delete.
216 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
217 (parm_default_def_partition_arg): Ditto.
218 (set_parm_default_def_partition): Ditto.
219 (get_parm_default_def_partitions): Ditto and make it static.
220 (get_undefined_value_partitions): Ditto and make it static.
221 (coalesce_with_default, coalesce_with_default): Update comment.
222 (create_coalesce_list_for_region): New func factored out from
223 create_outofssa_var_map.
224 (populate_coalesce_list_for_outofssa): New func factored out from
225 create_outofssa_var_map and coalesce_ssa_name.
226 (create_outofssa_var_map): Delete.
227 (coalesce_ssa_name): Refactor to support live range computation.
228 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
229 (get_parm_default_def_partitions): Delete.
230 (get_undefined_value_partitions): Ditto.
231 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
232 computation for loop region.
233 (new_tree_live_info, loe_visit_block): Ditto.
234 (live_worklist, set_var_live_on_entry): Ditto.
235 (calculate_live_on_exit, verify_live_on_entry): Ditto.
236 * tree-ssa-live.h (struct _var_map): New fields.
237 (init_var_map): Change decl.
238 (region_contains_p): New.
240 2018-05-25 Bin Cheng <bin.cheng@arm.com>
242 * tree-ssa-live.h (live_merge_and_clear): Delete.
244 2018-05-25 Richard Biener <rguenther@suse.de>
247 * tree-dump.c (dequeue_and_dump): Remove access to removed
248 operand 2 of a SWITCH_EXPR.
250 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
252 * doc/sourcebuild.texi (vect_double_cond_arith): Include
253 multiplication and division.
254 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
255 (cond_udiv@var{m}, cond_umod@var{m}): Document.
256 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
257 (cond_udiv_optab, cond_umod_optab): New optabs.
258 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
259 (IFN_COND_RDIV): New internal functions.
260 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
261 TRUNC_MOD_EXPR and RDIV_EXPR.
262 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
263 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
265 (SVE_INT_BINARY): Include mult.
266 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
267 (optab, sve_int_op): Handle mult.
268 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
270 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
271 for SVE_INT_BINARY_SD.
273 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
275 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
276 (optab, sve_int_op): Handle div and udiv.
277 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
278 for SVE_INT_BINARY_SD.
279 (*<optab><mode>3): New insn for the same.
281 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
283 * tree-vect-patterns.c: Include predict.h.
284 (vect_recog_divmod_pattern): Restrict check for division support
285 to when optimizing for size.
287 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
289 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
290 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
291 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
292 (gimple_match_op::set_op): Likewise.
293 (gimple_resimplify4): Declare.
294 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
295 (expr::gen_transform): Likewise.
296 (decision_tree::gen): Generate a simplification routine for 4 operands.
297 * gimple-match-head.c (gimple_simplify): Add an overload for
298 4 operands. In the top-level function, handle up to 4 call
299 arguments and call gimple_resimplify4.
300 (gimple_resimplify4): New function.
301 (build_call_internal): Pass a fourth operand.
302 (maybe_push_to_seq): Likewise.
303 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
304 Fold VEC_COND_EXPRs of an operation and a default value into
305 an IFN_COND_* function if possible.
306 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
308 (SVE_COND_FP_BINARY): Include them.
309 (optab, sve_fp_op): Handle them.
310 (SVE_INT_BINARY_REV): New code iterator.
311 (SVE_COND_FP_BINARY_REV): New int iterator.
312 (commutative): New int attribute.
313 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
315 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
317 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
318 (*cond_<optab><mode>): New patterns for reversed operands.
320 2018-05-25 Richard Biener <rguenther@suse.de>
322 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
323 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
324 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
325 (STMT_VINFO_GROUPED_ACCESS): Adjust.
326 * tree-vect-data-refs.c (everywhere): Adjust users.
327 * tree-vect-loop.c (everywhere): Likewise.
328 * tree-vect-slp.c (everywhere): Likewise.
329 * tree-vect-stmts.c (everywhere): Likewise.
330 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
332 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
334 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
336 (gcc_cv_as_section_exclude): ... this.
337 Try Solaris as #exclude syntax.
338 * configure: Regenerate.
339 * config.in: Regenerate.
340 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
342 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
343 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
345 * varasm.c (default_elf_asm_named_section): Don't check if
346 HAVE_GAS_SECTION_EXCLUDE is defined.
348 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
350 * doc/md.texi: Update the documentation of the cond_* optabs
351 to mention the new final operand. Fix GET_MODE_NUNITS call.
352 Describe the scalar case too.
353 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
354 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
356 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
357 (get_conditional_internal_fn): Update comment.
358 * tree-vect-loop.c (vectorizable_reduction): Pass the original
359 accumulator value as a final argument to conditional functions.
360 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
361 a define_expand and add an "else" operand. Assert for now that
362 the else operand is equal to operand 2. Use SVE_INT_BINARY and
363 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
364 (*cond_<optab><mode>): New patterns.
365 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
366 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
367 (UNSPEC_COND_EOR): Delete.
368 (optab): Remove associated mappings.
369 (SVE_INT_BINARY): New code iterator.
370 (sve_int_op): Remove int attribute and add "minus" to the code
372 (SVE_COND_INT_OP): Delete.
373 (SVE_COND_FP_OP): Rename to...
374 (SVE_COND_FP_BINARY): ...this.
376 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
378 * optabs.c (can_reuse_operands_p): New function.
379 (maybe_legitimize_operands): Try to reuse the results for
382 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
384 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
385 Add {q} suffix to insn mnemonic.
387 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
389 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
390 (msp430_warn_func_return): New.
392 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
394 * fold-const.c (tree_nonzero_bits): New function.
395 * fold-const.h (tree_nonzero_bits): Likewise.
396 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
397 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
399 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
403 * varasm.c (assemble_alias): Check ifunc_resolver only on
406 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
409 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
410 when memory input operand is handled.
412 2018-05-24 Luis Machado <luis.machado@linaro.org>
414 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
416 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
418 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
420 * match.pd: Delay FMA folds until after vectorization.
422 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
425 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
426 address check not strict.
428 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
430 * gimple-match.h (gimple_match_op): New class.
431 (mprts_hook): Replace parameters with a gimple_match_op *.
432 (maybe_build_generic_op): Likewise.
433 (gimple_simplified_result_is_gimple_val): Replace parameters with
434 a const gimple_match_op *.
435 (gimple_simplify): Replace code_helper * and tree * parameters with
436 a gimple_match_op * parameter.
437 (gimple_resimplify1): Replace code_helper *, tree and tree *
438 parameters with a gimple_match_op * parameter.
439 (gimple_resimplify2): Likewise.
440 (gimple_resimplify3): Likewise.
441 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
442 parameters with a gimple_match_op * parameter.
443 * gimple-match-head.c (gimple_simplify): Change prototypes of
444 auto-generated functions to take a gimple_match_op * instead of
445 separate code_helper * and tree * parameters. Make the same
446 change in the top-level overload and update calls to the
447 gimple_resimplify routines. Update calls to the auto-generated
448 functions and to maybe_push_res_to_seq in the publicly-facing
449 operation-specific gimple_simplify overloads.
450 (gimple_match_op::MAX_NUM_OPS): Define.
451 (gimple_resimplify1): Replace rcode and ops with a single res_op
452 parameter. Update call to gimple_simplify.
453 (gimple_resimplify2): Likewise.
454 (gimple_resimplify3): Likewise.
455 (mprts_hook): Replace parameters with a gimple_match_op *.
456 (maybe_build_generic_op): Likewise.
457 (build_call_internal): Replace type, nargs and ops with
459 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
460 with a single gimple_match_op *. Update calls to mprts_hook,
461 build_call_internal and gimple_simplified_result_is_gimple_val.
462 Factor out code that is common to the tree_code and combined_fn cases.
463 * genmatch.c (expr::gen_transform): Replace tem_code and
464 tem_ops with a gimple_match_op called tem_op. Update calls
465 to the gimple_resimplify functions and maybe_push_res_to_seq.
466 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
467 res_ops. Update call to the gimple_resimplify functions.
468 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
469 (decision_tree::gen): Make the functions take a gimple_match_op *
470 called res_op instead of separate res_code and res_ops parameters.
471 Update call accordingly.
472 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
473 and ops with a single res_op parameter. Update calls to
474 maybe_build_generic_op and maybe_push_res_to_seq.
475 (fold_stmt_1): Update calls to gimple_simplify and
476 replace_stmt_with_simplification.
477 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
478 and gimple_simplified_result_is_gimple_val.
479 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
481 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
482 with a gimple_match_op *.
483 (vn_nary_build_or_lookup): Likewise. Update call to
484 vn_nary_build_or_lookup_1.
485 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
486 gimple_match_op *. Update calls to the gimple_resimplify routines
487 and to gimple_simplified_result_is_gimple_val.
488 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
489 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
490 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
491 (visit_nary_op): Likewise.
492 (visit_reference_op_load): Likewise.
494 2018-05-23 Luis Machado <luis.machado@linaro.org>
496 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
497 modifier for printing the step amount.
499 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
502 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
505 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
507 * doc/sourcebuild.texi (Endianness): New subsubsection.
509 2018-05-23 Luis Machado <luis.machado@linaro.org>
511 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
512 <prefetch_dynamic_strides>: New const bool field.
513 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
514 prefetch_dynamic_strides.
515 (exynosm1_prefetch_tune): Likewise.
516 (thunderxt88_prefetch_tune): Likewise.
517 (thunderx_prefetch_tune): Likewise.
518 (thunderx2t99_prefetch_tune): Likewise.
519 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
521 (aarch64_override_options_internal): Update to set
522 PARAM_PREFETCH_DYNAMIC_STRIDES.
523 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
524 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
525 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
526 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
527 prefetch-dynamic-strides setting.
529 2018-05-23 Luis Machado <luis.machado@linaro.org>
531 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
532 <minimum_stride>: New const int field.
533 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
534 minimum_stride field defaulting to -1.
535 (exynosm1_prefetch_tune): Likewise.
536 (thunderxt88_prefetch_tune): Likewise.
537 (thunderx_prefetch_tune): Likewise.
538 (thunderx2t99_prefetch_tune): Likewise.
539 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
540 <default_opt_level>: Set to 3.
541 (aarch64_override_options_internal): Update to set
542 PARAM_PREFETCH_MINIMUM_STRIDE.
543 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
544 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
545 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
546 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
547 stride is constant and is below the minimum stride threshold.
549 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
551 * config/arm/arm-cpus.in (mode26): Delete.
552 (armv4): Delete mode26 reference.
553 * config/arm/arm.c (arm_configure_build_target): Delete use of
556 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
558 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
560 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
561 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
563 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
564 (floatunsdidf2): Ditto.
566 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
568 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
569 (fixuns_trunc<mode>si2_avx512f): Ditto.
570 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
571 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
572 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
574 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
576 PR rtl-optimization/79985
577 * df-scan.c (df_insn_refs_collect): Remove special case for
578 global registers and asm statements.
580 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
582 * extend.texi (Global Register Variables): Rewrite the bullet list.
583 Note that the register is available for allocation. Note that access
584 via inline asm must use constraints. Add note about async-signal
585 handlers. Remove paragraph about automagic register selection.
587 2018-05-23 Richard Biener <rguenther@suse.de>
589 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
590 of fixed offset from memset VN.
592 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
594 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
596 (alloc_cand_and_find_basis): Initialize first_interp field.
597 (slsr_process_mul): Modify first_interp field.
598 (slsr_process_add): Likewise.
599 (slsr_process_cast): Modify first_interp field for each new
601 (slsr_process_copy): Likewise.
602 (dump_candidate): Dump first_interp field.
603 (replace_mult_candidate): Process all interpretations, not just
605 (replace_rhs_if_not_dup): Likewise.
606 (replace_one_candidate): Likewise.
608 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
610 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
612 (aarch64_needs_frame_chain): New function.
613 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
615 2018-05-23 Sudakshina Das <sudi.das@arm.com>
618 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
619 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
620 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
621 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
622 as true for strict-align.
623 (aarch64_can_inline_p): Perform checks even when callee has no
624 attributes to check for strict alignment.
625 * doc/extend.texi (AArch64 Function Attributes): Document
627 * doc/invoke.texi: (AArch64 Options): Likewise.
629 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
631 PR tree-optimization/85853
632 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
633 the handling of the root of the node to...
634 (vect_slp_analyze_node_operations_1): ...this new function,
635 and run the whole thing with the child nodes' def types
636 set according to their SLP node's def type.
638 2018-05-23 Richard Biener <rguenther@suse.de>
641 * tree-data-ref.c (create_runtime_alias_checks): Defer
642 and ignore overflow warnings.
644 2018-05-23 Yury Gribov <tetra2005@gmail.com>
646 PR tree-optimization/85822
647 * tree-vrp.c (is_masked_range_test): Fix handling of negative
650 2018-05-23 Richard Biener <rguenther@suse.de>
652 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
653 memset constants via native_interpret_expr.
655 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
658 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
660 (cgraph_node::create_alias): Likewise.
661 (cgraph_node::get_availability): Check ifunc_resolver instead
662 of looking up ifunc attribute.
663 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
664 * varasm.c (do_assemble_alias): Likewise.
665 (assemble_alias): Likewise.
666 (default_binds_local_p_3): Likewise.
667 * cgraph.h (cgraph_node): Add ifunc_resolver.
668 (cgraph_node::only_called_directly_or_aliased_p): Return false
670 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
672 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
673 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
674 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
675 instead of looking up ifunc attribute.
677 2018-05-22 Luis Machado <luis.machado@linaro.org>
679 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
681 2018-05-22 Martin Sebor <msebor@redhat.com>
684 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
685 only when expasion succeeds.
686 (expand_builtin_strcmp): Same.
687 (expand_builtin_strncmp): Same.
689 2018-05-22 Martin Sebor <msebor@redhat.com>
691 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
693 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
694 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
696 * config/aarch64/aarch64-ldpstp.md: Replace uses of
697 aarch64_mem_pair_operand with memory_operand and delete operand swapping
699 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
700 Add check for legitimate_address.
701 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
702 (aarch64_swap_ldrstr_operands): New.
703 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
706 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
707 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
709 * config/aarch64/aarch64.md: New patterns to generate stp
711 (store_pair_sw, store_pair_dw): New patterns to generate stp for
712 single words and double words.
713 (load_pair_sw, load_pair_dw): Likewise.
714 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
716 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
718 * config/aarch64/aarch64-ldpstp.md: Modify peephole
719 for different mode ldpstp and add peephole for merged zero stores.
721 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
723 (aarch64_gen_store_pair): Rename calls to match new patterns.
724 (aarch64_gen_load_pair): Rename calls to match new patterns.
725 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
726 (load_pair<DREG:mode><DREG2:mode>): ... This.
727 (store_pair<mode>): Rename to...
728 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
729 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
731 (V_INT_EQUIV): Handle SImode.
732 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
735 2018-05-22 Martin Sebor <msebor@redhat.com>
738 * calls.c (maybe_warn_nonstring_arg): Use string length to set
739 or ajust the presumed bound on an operation to avoid unnecessary
742 2018-05-22 Martin Sebor <msebor@redhat.com>
744 PR tree-optimization/85826
745 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
746 assuming that a DECL necesarily has a constant size.
748 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
751 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
753 2018-05-22 Richard Biener <rguenther@suse.de>
755 PR tree-optimization/85834
756 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
757 non-constant and non-zero memset arguments.
759 2018-05-22 Martin Liska <mliska@suse.cz>
762 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
764 2018-05-22 Richard Biener <rguenther@suse.de>
766 PR tree-optimization/85863
767 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
768 comparisons when vectype is specified.
769 (vectorizable_condition): Do not specify vectype for
770 vect_is_simple_cond when SLP vectorizing.
772 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
775 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
776 define __ibm128 as long double.
777 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
778 as a distinct type when IEEE 128-bit support is enabled.
779 (init_float128_ieee): Fix up conversions between IFmode and IEEE
780 128-bit types to use the correct functions.
781 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
782 convert between 128-bit floating point types that have different
783 modes but the same representation, instead of using gen_lowpart to
785 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
787 (IFKF_reg): New attributes to give the register constraints for
789 (extend<mode>tf2_internal): New insns to mark an explicit
790 conversion between 128-bit floating point types that have a
791 different mode but share the same representation.
793 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
795 PR tree-optimization/85814
796 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
797 a null return from get_strinfo when unsharing the next
798 strinfo in the chain.
800 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
803 * varasm.c (weak_finish): Clean up weak_decls.
805 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
807 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
808 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
809 UNSPEC_UADALP values.
810 * config/aarch64/iterators.md (ABAL): New int iterator.
813 (sur): Add mappings for the above.
814 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
816 (aarch64_<sur>abal<mode>_4): Likewise.
817 (aarch64_<sur>adalp<mode>_3): Likewise.
818 (<sur>sadv16qi): New define_expand.
820 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
822 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
823 (*movdf_internal): Ditto.
824 (*rcpsf2_sse): Ditto.
825 (*rsqrtsf2_sse): Ditto.
826 (*sqrt<mode>2_sse): Ditto.
828 2018-05-21 Tamar Christina <tamar.christina@arm.com>
830 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
832 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
833 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
834 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
835 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
837 * config/aarch64/arm_neon.h: Likewise.
838 * config/aarch64/iterators.md (VQ_I): New.
840 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
842 * config.gcc: Add arc/t-multilib-linux to tmake_file for
844 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
847 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
849 * config/nds32/constraints.md (S): New constraint.
850 * config/nds32/nds32.md (call_internal): Use constraint S.
851 (call_value_internal): Likewise.
852 (sibcall_internal): Likewise.
853 (sibcall_value_internal): Likewise.
855 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
856 Chung-Ju Wu <jasonwucj@gmail.com>
858 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
861 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
862 Chung-Ju Wu <jasonwucj@gmail.com>
864 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
865 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
866 (nds32_rtx_costs_impl): Simplify.
867 (nds32_address_cost_impl): Simplify.
868 (nds32_init_rtx_costs): New function.
869 (nds32_rtx_costs_speed_prefer): Likewise.
870 (nds32_rtx_costs_size_prefer): Likewise.
871 (nds32_address_cost_speed_prefer): Likewise.
872 (nds32_address_cost_speed_fwprop): Likewise.
873 (nds32_address_cost_size_prefer): Likewise.
874 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
875 * config/nds32/nds32.c (nds32_option_override): Use
876 nds32_init_rtx_costs function.
878 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
880 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
881 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
882 (TARGET_PIPELINE_N8): Likewise.
883 (TARGET_PIPELINE_N10): Likewise.
884 (TARGET_PIPELINE_N13): Likewise.
885 (TARGET_PIPELINE_GRAYWOLF): Likewise.
887 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
889 * config/nds32/nds32-fpu.md: Update copyright year.
891 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
893 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
895 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
897 * config/nds32/nds32.c
898 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
899 * config/nds32/nds32.opt (minline-asm-r15): New option.
901 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
903 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
905 * config/nds32/nds32.md (abssi2): New pattern.
907 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
909 * config/i386/i386.md (rex64namesuffix): New mode attribute.
910 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
911 Merge insn pattern from sse_cvtsi2ss<round_name> and
912 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
913 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
914 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
915 using SWI48 mode iterator.
916 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
917 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
918 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
919 pattern from sse_cvttss2si<round_saeonly_name>
920 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
921 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
922 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
923 using SWI48 mode iterator.
924 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
925 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
926 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
927 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
928 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
929 using SWI48 mode iterator.
930 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
931 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
932 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
933 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
934 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
936 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
937 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
938 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
939 pattern from sse_cvttsd2si<round_saeonly_name>
940 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
942 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
944 * config/nds32/nds32-md-auxiliary.c
945 (nds32_valid_smw_lwm_base_p): Refine.
946 (nds32_output_smw_single_word): Refine.
947 (nds32_output_smw_double_word): New.
948 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
950 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
952 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
953 (nds32_output_stack_pop): Refine.
954 (nds32_expand_unaligned_load): Refine.
955 (nds32_expand_unaligned_store): Refine.
957 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
958 Chung-Ju Wu <jasonwucj@gmail.com>
960 * config/nds32/constants.md: Add TP_REGNUM constant.
961 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
962 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
964 * config/nds32/nds32-doubleword.md: Consider flag_pic.
965 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
966 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
967 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
968 and PIC code generation.
969 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
971 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
973 * config/nds32/nds32.md: Support TLS and PIC.
974 * config/nds32/nds32.c: Support TLS and PIC.
975 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
976 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
979 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
981 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
984 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
985 Chung-Ju Wu <jasonwucj@gmail.com>
987 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
988 * config/nds32/nds32-md-auxiliary.c
989 (symbolic_reference_mentioned_p): New.
990 (nds32_legitimize_ict_address): New.
991 (nds32_expand_ict_move): New.
992 (nds32_indirect_call_referenced_p): New.
993 (nds32_symbol_binds_local_p): Delete.
994 (nds32_long_call_p): Modify.
995 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
996 * config/nds32/nds32-protos.h
997 (symbolic_reference_mentioned_p): Declare.
998 (nds32_legitimize_ict_address): Declare.
999 (nds32_expand_ict_move): Declare.
1000 (nds32_indirect_call_referenced_p): Declare.
1001 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1002 (nds32_relax_group): Use nds32_ict_const_p as condition.
1003 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1004 (nds32_asm_file_start): Output ict_model directive in asm code.
1005 (nds32_legitimate_address_p): Consider indirect call.
1006 (nds32_print_operand): Consider indirect call.
1007 (nds32_print_operand_address): Consider indirect call.
1008 (nds32_insert_attributes): Handle "indirect_call" attribute.
1009 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1010 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1011 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1012 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1013 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1014 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1015 (TARGET_ICT_MODEL_SMALL): Define.
1016 (TARGET_ICT_MODEL_LARGE): Define.
1017 * config/nds32/nds32.md (movsi): Consider ict model.
1018 (call, call_value): Consider ict model.
1019 (sibcall, sibcall_value): Consider ict model.
1020 * config/nds32/nds32.opt (mict-model): New option.
1021 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1024 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1025 Monk Chiang <sh.chiang04@gmail.com>
1026 Jim Wilson <jimw@sifive.com>
1028 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1029 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1030 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1031 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1032 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1033 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1034 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1035 compute save_libcall_adjustment properly.
1036 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1037 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1038 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1039 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1040 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1041 (ABI_SPEC): Handle mabi=ilp32e.
1042 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1043 (RVE): Add RVE mask.
1044 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1045 <-march>: Add rv32e as an example.
1047 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1050 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1051 (intra_create_variable_infos): Handle C++ constructors.
1053 2018-05-18 Martin Liska <mliska@suse.cz>
1055 * passes.def: Remove a redundant pass.
1057 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1060 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1062 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1064 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1066 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1068 (armv2, armv2a, armv3, armv3m): Delete architectures.
1069 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1070 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1071 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1073 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1074 (*mulsidi3adddi): Likewise.
1075 (mulsidi3): Likewise.
1076 (*mulsidi3_nov6): Likewise.
1077 (umulsidi3): Likewise.
1078 (umulsidi3_nov6): Likewise.
1079 (umaddsidi4): Likewise.
1080 (*umulsidi3adddi): Likewise.
1081 (smulsi3_highpart): Likewise.
1082 (*smulsi3_highpart_nov6): Likewise.
1083 (umulsi3_highpart): Likewise.
1084 (*umulsi3_highpart_nov6): Likewise.
1085 * config/arm/arm.h (arm_arch3m): Delete.
1086 * config/arm/arm.c (arm_arch3m): Delete.
1087 (arm_option_override_internal): Update armv3-related comment.
1088 (arm_configure_build_target): Delete use of isa_bit_mode32.
1089 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1090 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1091 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1093 (mulusa3): Likewise.
1094 * config/arm/arm-protos.h (arm_arch3m): Delete.
1095 * config/arm/arm-tables.opt: Regenerate.
1096 * config/arm/arm-tune.md: Likewise.
1097 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1098 deleted architectures.
1100 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1102 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1103 (armv5t, armv5te): New features.
1104 (ARMv5, ARMv5e): Delete fgroups.
1105 (ARMv5t, ARMv5te): Adjust for above changes.
1107 (armv5, armv5e): Delete arches.
1108 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1110 (*call_reg_arm): Likewise.
1111 (*call_value_reg_armv5): Likewise.
1112 (*call_value_reg_arm): Likewise.
1113 (*call_symbol): Likewise.
1114 (*call_value_symbol): Likewise.
1115 (*sibcall_insn): Likewise.
1116 (*sibcall_value_insn): Likewise.
1118 (prefetch): Likewise.
1119 (define_split and define_peephole2 dependent on arm_arch5):
1121 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1123 (TARGET_ARM_QBIT): Likewise.
1124 (TARGET_DSP_MULTIPLY): Likewise.
1125 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1126 (arm_arch5, arm_arch5e): Delete.
1127 (arm_arch5t, arm_arch5te): Declare.
1128 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1129 (arm_arch5t): Declare.
1130 (arm_option_reconfigure_globals): Update for the above.
1131 (arm_options_perform_arch_sanity_checks): Update comment, replace
1132 use of arm_arch5 with arm_arch5t.
1133 (use_return_insn): Likewise.
1134 (arm_emit_call_insn): Likewise.
1135 (output_return_instruction): Likewise.
1136 (arm_final_prescan_insn): Likewise.
1137 (arm_coproc_builtin_available): Likewise.
1138 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1139 arm_arch5e with arm_arch5t and arm_arch5te.
1140 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1141 (arm_arch5t, arm_arch5te): Declare.
1142 * config/arm/arm-tables.opt: Regenerate.
1143 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1144 * config/arm/t-multilib: Likewise.
1145 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1146 instead of arm_arch5.
1147 (*call_reg_thumb1): Likewise.
1148 (*call_value_reg_thumb1_v5): Likewise.
1149 (*call_value_reg_thumb1): Likewise.
1150 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1152 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1154 2018-05-18 Martin Liska <mliska@suse.cz>
1156 PR gcov-profile/84846
1157 * doc/gcov.texi: Document -t option of gcov tool.
1159 2018-05-18 Martin Liska <mliska@suse.cz>
1161 PR gcov-profile/84846
1162 * gcov.c (print_usage): Add new -t option.
1163 (process_args): Handle the option.
1164 (generate_results): Use stdout as output when requested by
1167 2018-05-18 Martin Liska <mliska@suse.cz>
1169 PR gcov-profile/84846
1170 * coverage.c (coverage_init): Write PWD to .gcno file.
1171 * doc/gcov.texi: Document how working directory is printed.
1172 * gcov-dump.c (dump_gcov_file): Print PWD.
1173 * gcov.c (output_intermediate_file): Likewise.
1174 (read_graph_file): Read PWD string.
1175 (output_lines): Print PWD.
1177 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1180 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1181 for retval and return false if all args to phi are zero.
1183 2018-05-18 Richard Biener <rguenther@suse.de>
1185 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1187 (evrp_dom_walker::before_dom_children): Call it.
1189 2018-05-18 Richard Biener <rguenther@suse.de>
1191 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1192 results when processing array refs with variable index.
1194 2018-05-18 Toon Moene <toon@moene.org>
1196 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1197 directly after that of -floop-interchange. Indicate that both
1198 options are enabled by default when specifying -O3.
1200 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1202 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1203 iterator. Delete separate integer-mode vec_set<mode> expander.
1204 (aarch64_simd_vec_setv2di): Delete.
1205 (vec_setv2di): Delete.
1206 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1207 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1208 the "w, r" alternative.
1210 2018-05-18 Martin Liska <mliska@suse.cz>
1212 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1213 * tree-pass.h (make_pass_lower_switch_O0): New function.
1214 * tree-switch-conversion.c (node_has_low_bound): Remove.
1215 (node_has_high_bound): Likewise.
1216 (node_is_bounded): Likewise.
1217 (class pass_lower_switch): Make it a template type and create
1219 (pass_lower_switch::execute): Add template argument.
1220 (make_pass_lower_switch): New function.
1221 (make_pass_lower_switch_O0): New function.
1222 (do_jump_if_equal): Remove.
1223 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1224 all the hard work to tree optimization passes.
1226 2018-05-18 Martin Liska <mliska@suse.cz>
1228 * dbgcnt.c (limit_low): Renamed from limit.
1229 (limit_high): New variable.
1230 (dbg_cnt_is_enabled): Check for upper limit.
1231 (dbg_cnt): Adjust dumping.
1232 (dbg_cnt_set_limit_by_index): Add new argument for high
1234 (dbg_cnt_set_limit_by_name): Likewise.
1235 (dbg_cnt_process_single_pair): Parse new format.
1236 (dbg_cnt_process_opt): Use strtok.
1237 (dbg_cnt_list_all_counters): Remove 'value' and add
1239 * doc/invoke.texi: Document changes.
1241 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1243 * doc/sourcebuild.texi (scalar_all_fma): Document.
1244 * tree.def (FMA_EXPR): Delete.
1245 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1246 * internal-fn.c (ternary_direct): New macro.
1247 (expand_ternary_optab_fn): Likewise.
1248 (direct_ternary_optab_supported_p): Likewise.
1249 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1250 * builtins.c (fold_builtin_fma): Delete.
1251 (fold_builtin_3): Don't call it.
1252 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1253 * expr.c (expand_expr_real_2): Likewise.
1254 * fold-const.c (operand_equal_p): Likewise.
1255 (fold_ternary_loc): Likewise.
1256 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1257 * gimple.c (DEFTREECODE): Likewise.
1258 * gimplify.c (gimplify_expr): Likewise.
1259 * optabs-tree.c (optab_for_tree_code): Likewise.
1260 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1261 * tree-eh.c (operation_could_trap_p): Likewise.
1262 (stmt_could_throw_1_p): Likewise.
1263 * tree-inline.c (estimate_operator_cost): Likewise.
1264 * tree-pretty-print.c (dump_generic_node): Likewise.
1265 (op_code_prio): Likewise.
1266 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1267 * tree-ssa-operands.c (get_expr_operands): Likewise.
1268 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1269 * fold-const-call.h (fold_fma): Delete.
1270 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1271 CFN_FNMA and CFN_FNMS.
1273 * genmatch.c (combined_fn): New enum.
1274 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1275 (commutative_op): New function.
1276 (commutate): Use it. Handle more than 2 operands.
1277 (dt_operand::gen_gimple_expr): Use commutative_op.
1278 (parser::parse_expr): Allow :c to be used with non-binary
1279 operators if the commutative operand is known.
1280 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1281 CFN_FMS, CFN_FNMA and CFN_FNMS.
1282 (backprop::process_assign_use): Remove FMA_EXPR handling.
1283 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1284 (gen_hsa_fma): New function.
1285 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1286 IFN_FNMA and IFN_FNMS.
1287 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1288 * gimple-fold.h (follow_all_ssa_edges): Declare.
1289 * gimple-fold.c (follow_all_ssa_edges): New function.
1290 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1291 gimple_build interface and use follow_all_ssa_edges to fold the result.
1292 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1293 instead of checking for optabs directly.
1294 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1295 rather than FMA_EXPRs.
1296 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1297 call to IFN_FMA instead of an FMA_EXPR.
1299 2018-05-17 Jim Wilson <jimw@sifive.com>
1301 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1302 sign extended promoted subreg, and the range does not have the sign bit
1303 set, then do a sign extend.
1305 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1306 test, check for sign extended subreg and/or constant operands, and
1307 do a sign extend in that case.
1309 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1311 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1312 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1314 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1315 Change logics_shift_reg to logics_shift_imm.
1316 (thunderx2t99_fp_loadpair_basic): Delete.
1317 (thunderx2t99_fp_storepair_basic): Delete.
1318 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1319 (thunderx2t99_asimd_polynomial): Delete.
1320 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1321 and neon_fp_mul_d_scalar_q.
1322 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1323 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1324 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1325 (thunderx2t99_asimd_lut): Add missing tbl types.
1326 (thunderx2t99_asimd_ext): Delete.
1327 (thunderx2t99_asimd_load1_1_mult): Delete.
1328 (thunderx2t99_asimd_load1_2_mult): Delete.
1329 (thunderx2t99_asimd_load1_ldp): New.
1330 (thunderx2t99_asimd_load1): New.
1331 (thunderx2t99_asimd_load2): Add missing *load2* types.
1332 (thunderx2t99_asimd_load3): New.
1333 (thunderx2t99_asimd_load4): New.
1334 (thunderx2t99_asimd_store1_1_mult): Delete.
1335 (thunderx2t99_asimd_store1_2_mult): Delete.
1336 (thunderx2t99_asimd_store2_mult): Delete.
1337 (thunderx2t99_asimd_store2_onelane): Delete.
1338 (thunderx2t99_asimd_store_stp): New.
1339 (thunderx2t99_asimd_store1): New.
1340 (thunderx2t99_asimd_store2): New.
1341 (thunderx2t99_asimd_store3): New.
1342 (thunderx2t99_asimd_store4): New.
1344 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1346 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1347 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1349 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1350 Segher Boessenkool <segher@kernel.crashing.org>
1353 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1356 2018-05-17 Richard Biener <rguenther@suse.de>
1358 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1359 for pruning loop and prune defs feeding only already visited PHIs.
1361 2018-05-17 Richard Biener <rguenther@suse.de>
1363 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1365 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1366 Richard Biener <rguenther@suse.de>
1368 PR tree-optimization/85793
1369 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1370 for VMAT_ELEMENTWISE.
1372 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1374 * internal-fn.h (lookup_internal_fn): Declare
1375 * internal-fn.c (lookup_internal_fn): New function.
1376 * gimple.c (gimple_build_call_from_tree): Handle calls to
1378 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1379 internal function names.
1380 * tree-pretty-print.c (dump_generic_node): Likewise.
1381 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1383 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1385 * gimple-fold.h (gimple_build): Make the function forms take
1386 combined_fn rather than built_in_function.
1387 (gimple_simplify): Likewise.
1388 * gimple-match-head.c (gimple_simplify): Likewise.
1389 * gimple-fold.c (gimple_build): Likewise.
1390 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1391 rather than gimple_build_call_internal.
1392 (get_initial_defs_for_reduction): Likewise.
1393 (vect_create_epilog_for_reduction): Likewise.
1394 (vectorizable_live_operation): Likewise.
1396 2018-05-17 Martin Liska <mliska@suse.cz>
1398 * gimple-ssa-sprintf.c (format_directive): Do not use
1399 space in between 'G_' and '('.
1401 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1404 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1405 even if the mask is not all ones.
1408 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1410 (ix86_gimple_fold_builtin): Likewise.
1413 * config/i386/i386.c: Include tree-vector-builder.h.
1414 (ix86_vector_shift_count): New function.
1415 (ix86_fold_builtin): Fold shift builtins by scalar count.
1416 (ix86_gimple_fold_builtin): Likewise.
1418 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1419 _mm512_setzero): New intrinsics.
1421 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1422 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1424 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1425 code generation for cases where splatting a value is not useful.
1426 * simplify-rtx.c (simplify_ternary_operation): Simplify
1427 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1428 mode to a vec_concat.
1430 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1432 * config.gcc: Support "goldmont-plus".
1433 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1435 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1436 PROCESSOR_GOLDMONT_PLUS.
1437 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1438 (processor_target_table): Add "goldmont-plus".
1439 (PTA_GOLDMONT_PLUS): Define.
1440 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1441 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1442 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1443 (fold_builtin_cpu): Add "goldmont-plus".
1444 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1445 (ix86_option_override_internal): Add "goldmont-plus".
1446 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1447 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1448 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1449 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1451 2018-05-17 Richard Biener <rguenther@suse.de>
1453 PR tree-optimization/85757
1454 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1455 remove defs that only feed that PHI from further processing.
1457 2018-05-16 Jim Wilson <jimw@sifive.com>
1459 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1461 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1463 2018-05-16 Mark Wielaard <mark@klomp.org>
1465 * dwarf2out.c (count_index_strings): New function.
1466 (output_indirect_strings): Call count_index_strings and generate
1467 header for dwarf_version >= 5.
1469 2018-05-16 Mark Wielaard <mark@klomp.org>
1471 * dwarf2out.c (dwarf_FORM): New function.
1472 (set_indirect_string): Use dwarf_FORM.
1473 (reset_indirect_string): Likewise.
1474 (size_of_die): Likewise.
1475 (value_format): Likewise.
1476 (output_die): Likewise.
1477 (add_skeleton_AT_string): Likewise.
1478 (output_macinfo_op): Likewise.
1479 (index_string): Likewise.
1480 (output_index_string_offset): Likewise.
1481 (output_index_string): Likewise.
1482 (count_index_strings): Likewise.
1484 2018-05-16 Carl Love <cel@us.ibm.com>
1486 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1487 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1489 2018-05-16 Martin Jambor <mjambor@suse.cz>
1491 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1492 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1494 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1496 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1497 (fnma<mode>4): Likewise.
1498 (fms<mode>4): Likewise.
1499 (fnms<mode>4): Likewise.
1500 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1501 (aarch64_fnma<mode>4): Likewise.
1502 (aarch64_fms<mode>4): Likewise.
1503 (aarch64_fnms<mode>4): Likewise.
1504 (aarch64_fnmadd<mode>4): Likewise.
1506 2018-05-16 Jason Merrill <jason@redhat.com>
1508 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1510 2018-05-16 Richard Biener <rguenther@suse.de>
1512 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1513 (dump_stmt_cost): Declare.
1514 (add_stmt_cost): Dump cost we add.
1515 (add_stmt_costs): New function.
1516 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1518 (vect_analyze_stmt): Adjust prototype.
1519 (vectorizable_condition): Likewise.
1520 (vectorizable_live_operation): Likewise.
1521 (vectorizable_reduction): Likewise.
1522 (vectorizable_induction): Likewise.
1523 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1524 cost vector to pass to vectorizable_ and record afterwards.
1525 (vect_model_reduction_cost): Take cost vector argument and adjust.
1526 (vect_model_induction_cost): Likewise.
1527 (vectorizable_reduction): Likewise.
1528 (vectorizable_induction): Likewise.
1529 (vectorizable_live_operation): Likewise.
1530 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1531 SLP_TREE_NUMBER_OF_VEC_STMTS.
1532 (vect_analyze_slp_cost_1): Remove.
1533 (vect_analyze_slp_cost): Likewise.
1534 (vect_slp_analyze_node_operations): Take visited args and
1535 a target cost vector. Avoid processing already visited stmt sets.
1536 (vect_slp_analyze_operations): Use a local cost vector to gather
1537 costs and register those of non-discarded instances.
1538 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1539 (vect_schedule_slp_instance): Remove copying of
1540 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1542 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1543 adding cost. Record cost entry location.
1544 (vect_prologue_cost_for_slp_op): Function to compute cost of
1545 a constant or invariant generated for SLP vect in the prologue,
1546 split out from vect_analyze_slp_cost_1.
1547 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1548 (vect_model_promotion_demotion_cost): Likewise.
1549 (vect_model_store_cost): Likewise, make static.
1550 (vect_model_load_cost): Likewise.
1551 (vectorizable_bswap): Add cost vector arg and adjust.
1552 (vectorizable_call): Likewise.
1553 (vectorizable_simd_clone_call): Likewise.
1554 (vectorizable_conversion): Likewise.
1555 (vectorizable_assignment): Likewise.
1556 (vectorizable_shift): Likewise.
1557 (vectorizable_operation): Likewise.
1558 (vectorizable_store): Likewise.
1559 (vectorizable_load): Likewise.
1560 (vectorizable_condition): Likewise.
1561 (vectorizable_comparison): Likewise.
1562 (can_vectorize_live_stmts): Likewise.
1563 (vect_analyze_stmt): Likewise.
1564 (vect_transform_stmt): Adjust calls to vectorizable_*.
1565 * tree-vectorizer.c: Include gimple-pretty-print.h.
1566 (dump_stmt_cost): New function.
1568 2018-05-16 Richard Biener <rguenther@suse.de>
1570 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1571 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1572 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1573 (check_name): New callback.
1574 (dse_classify_store): Track cycles via a visited bitmap of PHI
1575 defs and simplify handling of in-loop and across loop dead stores
1576 and properly fail for loop-variant refs. Handle byte-tracking with
1577 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1580 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1582 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1583 (vect_get_mask_type_for_stmt): Likewise.
1584 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1586 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1587 to determine the statement's vector type and the vector type that
1588 should be used for calculating nunits. Deal with cases in which
1589 the type has to be deferred.
1590 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1591 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1592 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1593 (vect_determine_vf_for_stmt): New functions, split out from...
1594 (vect_determine_vectorization_factor): ...here.
1595 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1596 (vect_get_mask_type_for_stmt): New functions, split out from
1597 vect_determine_vectorization_factor.
1599 2018-05-16 Richard Biener <rguenther@suse.de>
1601 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1602 verify the [VEC_]COND_EXPR embedded comparison.
1604 2018-05-15 Martin Sebor <msebor@redhat.com>
1606 PR tree-optimization/85753
1607 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1608 RECORD_TYPE in addition to ARRAY_TYPE.
1610 2018-05-15 Martin Sebor <msebor@redhat.com>
1613 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1615 2018-05-15 Richard Biener <rguenther@suse.de>
1617 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1618 add by_clobber_p one. Change algorithm to collect all defs
1619 representing uses we need to walk and try reducing them to
1620 a single one before failing.
1621 (dse_dom_walker::dse_optimize_stmt): Adjust.
1623 2018-05-13 Mark Wielaard <mark@klomp.org>
1625 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1626 (size_of_loc_descr): Likewise.
1627 (output_loc_operands): Likewise.
1628 (output_loc_operands_raw): Likewise.
1629 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1630 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1631 (hash_loc_operands): Likewise.
1632 (compare_loc_operands): Likewise.
1634 2018-05-14 Mark Wielaard <mark@klomp.org>
1636 * dwarf2out.c (count_index_addrs): New function.
1637 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1639 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1641 PR tree-optimization/83648
1642 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1643 return value as malloc candidate.
1645 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1648 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1649 as true in call to suggest_attribute.
1651 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1653 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1654 -mreadonly-in-sdata.
1656 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1658 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1660 (aarch64_crypto_aesd_fused): Likewise.
1662 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1664 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1665 (movsi_aarch64): Likewise.
1666 (load_pairsi): Likewise.
1667 (load_pairdi): Likewise.
1668 (store_pairsi): Likewise.
1669 (store_pairdi): Likewise.
1670 (load_pairsf): Likewise.
1671 (load_pairdf): Likewise.
1672 (store_pairsf): Likewise.
1673 (store_pairdf): Likewise.
1674 (zero_extend): Likewise.
1675 (trunc): Swap alternatives.
1676 (fcvt_target): Add '?' to prefer w over r.
1678 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1681 * config/i386/i386.md: Disallow non-commutative arithmetics in
1682 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1683 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1684 in the peephole2 before it.
1686 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1688 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1689 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1690 (ix86_handle_option): Handle -mcldemote.
1691 * config.gcc: New header.
1692 * config/i386/cldemoteintrin.h: New file.
1693 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1694 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1696 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1697 OPTION_MASK_ISA_CLDEMOTE.
1698 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1699 (ix86_valid_target_attribute_inner_p): Ditto.
1700 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1701 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1702 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1703 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1704 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1706 * config/i386/i386.opt: Add -mcldemote.
1707 * config/i386/x86intrin.h: New header.
1708 * doc/invoke.texi: Add -mcldemote.
1710 2018-05-14 Richard Biener <rguenther@suse.de>
1712 * doc/match-and-simplify.texi: Adjust :s documentation.
1714 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1716 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1717 intended memcpy size.
1718 (REORDER_45): Likewise.
1720 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1722 * sort.cc: New file.
1723 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1724 * vec.c (qsort_chk): Use gcc_qsort.
1725 * Makefile.in (OBJS-libcommon): Add sort.o.
1726 (build/sort.o): New target. Use it...
1727 (BUILD_RTL): ... here, and...
1728 (build/gencfn-macros): ... here, and...
1729 (build/genmatch): ... here.
1731 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1732 Chung-Ju Wu <jasonwucj@gmail.com>
1734 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1735 * config/nds32/nds32-graywolf.md: New file.
1736 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1737 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1739 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1740 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1741 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1742 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1743 * config/nds32/pipelines.md: Include n15 settings.
1745 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1746 Chung-Ju Wu <jasonwucj@gmail.com>
1748 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1749 * config/nds32/nds32-n13.md: New file.
1750 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1751 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1753 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1754 * config/nds32/nds32.md (pipeline_model): Add n13.
1755 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1756 * config/nds32/pipelines.md: Include n13 settings.
1758 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1759 Chung-Ju Wu <jasonwucj@gmail.com>
1761 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1762 * config/nds32/nds32-n10.md: New file.
1763 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1764 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1766 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1767 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1768 * config/nds32/nds32.md (pipeline_model): Add n10.
1769 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1770 * config/nds32/pipelines.md: Include n10 settings.
1772 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1773 Kito Cheng <kito.cheng@gmail.com>
1774 Chung-Ju Wu <jasonwucj@gmail.com>
1776 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1777 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1778 Add enum values for DSP extension instructions.
1779 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1781 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1782 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1784 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1785 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1786 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1787 * config/nds32/nds32-intrinsic.md: Likewise.
1788 * config/nds32/nds32_intrinsic.h: Likewise.
1789 * config/nds32/nds32-md-auxiliary.c: Likewise.
1790 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1791 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1792 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1793 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1794 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1795 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1796 TYPE_DMAC in switch statement.
1797 * config/nds32/nds32.c: New checking and implementation for DSP
1798 extension instructions.
1799 * config/nds32/nds32.h: Likewise.
1800 * config/nds32/nds32.md: Likewise.
1801 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1802 * config/nds32/predicates.md: Implement new predicates for DSP
1805 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1807 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1808 Reformat alternatives and attributes so it is easier to identify
1809 which constraints/attributes go with which instruction.
1810 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1811 (mov<mode>_softfloat32, FMOVE64): Likewise.
1812 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1813 (mov<mode>_softfloat64, FMOVE64): Likewise.
1815 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1817 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1819 (Basic PowerPC Built-in Functions): The new name of the
1820 subsection previously known as "PowerPC Built-in Functions".
1821 (Basic PowerPC Built-in Functions Available on all Configurations):
1823 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1824 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1825 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1826 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1828 2018-05-11 Martin Jambor <mjambor@suse.cz>
1831 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1834 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1837 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1839 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1841 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1842 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1843 (ix86_handle_option): Handle -mwaitpkg.
1844 * config.gcc: New header.
1845 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1846 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1847 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1849 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1850 OPTION_MASK_ISA_WAITPKG.
1851 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1852 (ix86_option_override_internal): Add PTA_WAITPKG.
1853 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1854 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1855 IX86_BUILTIN_TPAUSE.
1856 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1857 __builtin_ia32_umwait and __builtin_ia32_tpause.
1858 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1859 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1860 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1861 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1862 UNSPECV_TPAUSE): New.
1863 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1864 * config/i386/i386.opt: Add -mwaitpkg.
1865 * config/i386/waitpkgintrin.h: New file.
1866 * config/i386/x86intrin.h: New header.
1867 * doc/invoke.texi: Add -mwaitpkg.
1869 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1872 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1874 (cortex-m0): Use armv6s-m isa.
1875 (cortex-m0plus): Likewise.
1876 (cortex-m1): Likewise.
1877 (cortex-m0.small-multiply): Likewise.
1878 (cortex-m0plus.small-multiply): Likewise.
1879 (cortex-m1.small-multiply): Likewise.
1881 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1882 Jakub Jelinek <jakub@redhat.com>
1884 PR tree-optimization/85692
1885 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1886 source permute as well.
1888 2018-05-11 Martin Liska <mliska@suse.cz>
1891 * doc/extend.texi: Document LLVM style format for no_sanitize
1894 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1896 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1897 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1898 (mode_supports_vsx_dform_quad): Likewise.
1899 (mode_supports_vmx_dform): Move these functions to be next to the
1900 other mode_supports functions.
1901 (mode_supports_dq_form): Likewise.
1902 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1903 mode_supports_dq_form.
1904 (reg_offset_addressing_ok_p): Likewise.
1905 (offsettable_ok_by_alignment): Likewise.
1906 (rs6000_legitimate_offset_address_p): Likewise.
1907 (legitimate_lo_sum_address_p): Likewise.
1908 (rs6000_legitimize_address): Likewise.
1909 (rs6000_legitimize_reload_address): Likewise.
1910 (rs6000_secondary_reload_inner): Likewise.
1911 (rs6000_preferred_reload_class): Likewise.
1912 (rs6000_output_move_128bit): Likewise.
1914 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1916 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1917 Generate SImode target register for null target.
1918 <case IX86_BUILTIN_XGETBV>: Ditto.
1919 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1920 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1922 2018-05-10 Carl Love <cel@us.ibm.com>
1924 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1925 dcbtt and dcbtstt if operands[2] is 0.
1927 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1930 * config/i386/sse.md (usadv64qi): New expander.
1932 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1934 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1935 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1936 -maltivec=be support.
1937 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1938 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1939 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1940 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1941 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1942 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1943 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1944 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1945 altivec_vsumsws): Adjust.
1946 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1947 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1948 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1950 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1951 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1952 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1953 (altivec_lve<VI_char>x): Delete expand.
1954 (*altivec_lve<VI_char>x_internal): Rename to...
1955 (altivec_lve<VI_char>x): ... this.
1956 (altivec_lvxl_<mode>): Delete expand.
1957 (*altivec_lvxl_<mode>_internal): Rename to ...
1958 (altivec_lvxl_<mode>): ... this.
1959 (altivec_stvxl_<mode>): Delete expand.
1960 (*altivec_stvxl_<mode>_internal): Rename to ...
1961 (altivec_stvxl_<mode>): ... this.
1962 (altivec_stve<VI_char>x): Delete expand.
1963 (*altivec_stve<VI_char>x_internal): Rename to ...
1964 (altivec_stve<VI_char>x): ... this.
1965 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1966 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1967 reduc_plus_scal_<mode>): Adjust.
1968 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1970 (rs6000_cpu_cpp_builtins): Adjust.
1971 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1972 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1973 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1974 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1975 -maltivec=be support.
1976 (rs6000_split_vec_extract_var): Adjust.
1977 (rs6000_split_v4si_init): Adjust.
1978 (swap_selector_for_mode): Delete.
1979 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1980 altivec_expand_stvex_be): Delete.
1981 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1982 -maltivec=be support.
1983 (rs6000_gimple_fold_builtin): Ditto.
1984 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1986 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1987 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1988 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1989 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1990 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1991 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1992 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1993 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1994 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1995 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1996 anonymous split): Adjust.
1997 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1998 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2000 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2002 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2003 when --with-gxx-include-dir is also specified.
2004 * configure: Regenerate.
2006 2018-05-09 Jim Wilson <jimw@sifive.com>
2009 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2010 * config/riscv/t-withmultilib: New.
2011 * config/riscv/withmultilib.h: New.
2012 * doc/install.texi: Document RISC-V --with-multilib-list support.
2014 2018-05-09 Richard Biener <rguenther@suse.de>
2016 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2018 (vect_bb_vectorization_profitable_p): Adjust. Compute
2019 actual scalar cost using the cost vector and the add_stmt_cost
2022 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2024 PR rtl-optimization/85645
2025 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2026 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2028 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2030 PR rtl-optimization/85645
2031 * shrink-wrap.c (spread_components): Return a boolean saying if
2032 anything was changed.
2033 (try_shrink_wrapping_separate): Iterate spread_components until
2034 nothing changes anymore.
2036 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2038 PR rtl-optimization/85645
2039 * regrename.c (build_def_use): Also kill the chains that include the
2040 destination of a REG_CFA_REGISTER note.
2042 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2044 PR rtl-optimization/85645
2045 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2046 insn that has a REG_CFA_REGISTER note.
2048 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2050 * cfgexpand.c (expand_clobber): New function.
2051 (expand_gimple_stmt_1): Use it.
2052 * tree-vect-stmts.c (vect_clobber_variable): New function,
2054 (vectorizable_simd_clone_call): ...here.
2055 (vectorizable_store): Emit a clobber either side of an
2056 IFN_STORE_LANES sequence.
2057 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2059 2018-05-09 Tom de Vries <tom@codesourcery.com>
2062 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2063 (define_insn "trap_if_false"): Add exit after trap.
2065 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2067 PR rtl-optimization/85638
2068 * bb-reorder.c: Include common/common-target.h.
2069 (create_forwarder_block): New function extracted from...
2070 (fix_up_crossing_landing_pad): ...here. Rename into...
2071 (dw2_fix_up_crossing_landing_pad): ...this.
2072 (sjlj_fix_up_crossing_landing_pad): New function.
2073 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2074 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2075 from both partitions and exit the loop after one iteration.
2077 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2080 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2082 (Basic PowerPC Built-in Functions): The new name of the
2083 subsection previously known as "PowerPC Built-in Functions".
2084 (Basic PowerPC Built-in Functions Available on all Configurations):
2086 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2088 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2090 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2092 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2095 2018-05-08 Jim Wilson <jimw@sifive.com>
2097 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2098 (LD_EMUL_SUFFIX): New.
2099 (LINK_SPEC): Use it.
2101 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2103 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2105 (Basic PowerPC Built-in Functions): The new name of the
2106 subsection previously known as "PowerPC Built-in Functions".
2107 (Basic PowerPC Built-in Functions Available on all Configurations):
2109 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2111 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2113 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2115 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2118 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2121 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2122 after cmpelim optimization.
2124 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2126 * config.gcc: Support "goldmont".
2127 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2128 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2130 * config/i386/i386.c (m_GOLDMONT): Define.
2131 (processor_target_table): Add "goldmont".
2132 (PTA_GOLDMONT): Define.
2133 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2134 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2135 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2136 (fold_builtin_cpu): Add "goldmont".
2137 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2138 (ix86_option_override_internal): Add "goldmont".
2139 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2140 (processor_type): Add PROCESSOR_GOLDMONT.
2141 * config/i386/i386.md: Add CPU "glm".
2142 * config/i386/glm.md: New file.
2143 * config/i386/x86-tune.def: Add m_GOLDMONT.
2144 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2146 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2149 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2151 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2152 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2153 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2156 * config/i386/i386.c (ix86_fold_builtin): Handle
2157 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2160 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2161 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2163 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2166 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2167 (check_arch): Likewise.
2168 (check_fpu): Return the result rather than printing it.
2169 (end arch): Fix operator precedence.
2170 (end cpu): Likewise.
2171 (END): Print the result from check_fpu.
2173 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2174 Alan Hayward <alan.hayward@arm.com>
2175 David Sherwood <david.sherwood@arm.com>
2177 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2178 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2179 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2180 (*fcmuo<mode>_and): New patterns.
2182 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2184 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2185 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2186 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2187 (cmp_op, sve_imm_con): New code attributes.
2188 (SVE_COND_INT_CMP, imm_con): Delete.
2189 (cmp_op): Remove above unspecs from int attribute.
2190 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2192 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2193 comparison-specific unspecs.
2194 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2195 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2196 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2197 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2198 (*vec_fcm<cmp_op><mode>): Rename to...
2199 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2200 (*vec_fcmuo<mode>): Rename to...
2201 (*fcmuo<mode>): ...this and adjust likewise.
2202 (*pred_fcm<cmp_op><mode>): New pattern.
2203 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2204 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2206 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2208 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2209 (aarch64_emit_sve_predicated_cond): New function.
2210 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2211 (aarch64_emit_unspec_cond_or): Replace with...
2212 (aarch64_emit_sve_or_conds): ...this new function. Use
2213 aarch64_emit_sve_ptrue_op for the individual comparisons and
2214 aarch64_emit_binop to OR them together.
2215 (aarch64_emit_inverted_unspec_cond): Replace with...
2216 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2217 aarch64_emit_sve_ptrue_op for the comparison and
2218 aarch64_emit_unop to invert the result.
2219 (aarch64_expand_sve_vec_cmp_float): Update after the above
2220 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2222 2018-05-07 Nathan Sidwell <nathan@acm.org>
2224 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2225 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2226 (Backwards Compatibility): Likewise.
2228 2018-05-07 Luis Machado <luis.machado@linaro.org>
2232 2018-05-07 Luis Machado <luis.machado@linaro.org>
2234 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2235 <prefetch_dynamic_strides>: New const bool field.
2236 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2237 prefetch_dynamic_strides.
2238 (exynosm1_prefetch_tune): Likewise.
2239 (thunderxt88_prefetch_tune): Likewise.
2240 (thunderx_prefetch_tune): Likewise.
2241 (thunderx2t99_prefetch_tune): Likewise.
2242 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2243 (aarch64_override_options_internal): Update to set
2244 PARAM_PREFETCH_DYNAMIC_STRIDES.
2245 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2246 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2247 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2248 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2249 prefetch-dynamic-strides setting.
2251 2018-05-07 Luis Machado <luis.machado@linaro.org>
2253 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2254 <minimum_stride>: New const int field.
2255 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2256 minimum_stride field.
2257 (exynosm1_prefetch_tune): Likewise.
2258 (thunderxt88_prefetch_tune): Likewise.
2259 (thunderx_prefetch_tune): Likewise.
2260 (thunderx2t99_prefetch_tune): Likewise.
2261 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2262 (aarch64_override_options_internal): Update to set
2263 PARAM_PREFETCH_MINIMUM_STRIDE.
2264 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2265 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2266 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2267 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2268 stride is constant and is below the minimum stride threshold.
2270 2018-05-07 Luis Machado <luis.machado@linaro.org>
2272 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2275 2018-05-07 Luis Machado <luis.machado@linaro.org>
2277 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2278 <prefetch_dynamic_strides>: New const bool field.
2279 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2280 prefetch_dynamic_strides.
2281 (exynosm1_prefetch_tune): Likewise.
2282 (thunderxt88_prefetch_tune): Likewise.
2283 (thunderx_prefetch_tune): Likewise.
2284 (thunderx2t99_prefetch_tune): Likewise.
2285 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2286 (aarch64_override_options_internal): Update to set
2287 PARAM_PREFETCH_DYNAMIC_STRIDES.
2288 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2289 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2290 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2291 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2292 prefetch-dynamic-strides setting.
2294 2018-05-07 Luis Machado <luis.machado@linaro.org>
2296 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2297 <minimum_stride>: New const int field.
2298 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2299 minimum_stride field.
2300 (exynosm1_prefetch_tune): Likewise.
2301 (thunderxt88_prefetch_tune): Likewise.
2302 (thunderx_prefetch_tune): Likewise.
2303 (thunderx2t99_prefetch_tune): Likewise.
2304 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2305 (aarch64_override_options_internal): Update to set
2306 PARAM_PREFETCH_MINIMUM_STRIDE.
2307 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2308 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2309 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2310 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2311 stride is constant and is below the minimum stride threshold.
2313 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2316 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2317 the type is addressable. Don't force op into register if it has
2320 2018-05-05 Roland McGrath <mcgrathr@google.com>
2323 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2324 any section for which we don't know a specific type it should have,
2325 regardless of name. Previously this was done only for the exact
2326 names ".init_array", ".fini_array", and ".preinit_array".
2327 (default_elf_asm_named_section): Add comment about
2328 relationship with default_section_type_flags and SECTION_NOTYPE.
2329 (get_section): Don't consider it a type conflict if one side has
2330 SECTION_NOTYPE and the other doesn't, as long as neither has the
2331 SECTION_BSS et al used in the default_section_type_flags logic.
2333 2018-05-05 Tom de Vries <tom@codesourcery.com>
2336 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2337 (workaround_barsyncs): New function.
2338 (nvptx_reorg): Use workaround_barsyncs.
2339 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2340 (define_expand "nvptx_membar_cta"): New define_expand.
2341 (define_insn "*nvptx_membar_cta"): New insn.
2343 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2345 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2346 To improve optimization opportunities.
2347 * builtin-types.def: The new needed builtin types for the above.
2349 2018-05-04 Richard Biener <rguenther@suse.de>
2351 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2352 * gimple-ssa-store-merging.c
2353 (imm_store_chain_info::output_merged_store): Remove redundant create,
2354 release split_store vector contents on failure.
2355 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2356 scalar stmt vector on cache hit.
2358 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2360 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2362 * config.gcc (powerpc-xilinx-eabi*): Remove.
2363 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2365 (fusion_addis_mem_combo_load): Ditto.
2366 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2368 (rs6000_cpu_cpp_builtins): Ditto.
2369 * config/rs6000/rs6000-linux.c
2370 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2371 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2372 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2374 (rs6000_setup_reg_addr_masks): Ditto.
2375 (rs6000_init_hard_regno_mode_ok): Ditto.
2376 (rs6000_option_override_internal): Ditto.
2377 (legitimate_lo_sum_address_p): Ditto.
2378 (rs6000_legitimize_address): Ditto.
2379 (rs6000_legitimize_reload_address): Ditto.
2380 (rs6000_legitimate_address_p): Ditto.
2381 (abi_v4_pass_in_fpr): Ditto.
2382 (setup_incoming_varargs): Ditto.
2383 (rs6000_gimplify_va_arg): Ditto.
2384 (rs6000_split_multireg_move): Ditto.
2385 (rs6000_savres_strategy): Ditto.
2386 (rs6000_emit_prologue_components): Ditto.
2387 (rs6000_emit_epilogue_components): Ditto.
2388 (rs6000_emit_prologue): Ditto.
2389 (rs6000_emit_epilogue): Ditto.
2390 (rs6000_elf_file_end): Ditto.
2391 (rs6000_function_value): Ditto.
2392 (rs6000_libcall_value): Ditto.
2393 * config/rs6000/rs6000.h: Ditto.
2394 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2395 (TARGET_MINMAX): ... this. New.
2396 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2397 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2398 (*movsi_internal1_single): Delete.
2399 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2400 mfpu=, mxilinx-fpu): Delete.
2401 * config/rs6000/singlefp.h: Delete.
2402 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2403 * config/rs6000/t-rs6000: Ditto.
2404 * config/rs6000/t-xilinx: Delete.
2405 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2406 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2407 (VStype_simple): Delete.
2408 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2409 * config/rs6000/xfpu.h: Delete.
2410 * config/rs6000/xfpu.md: Delete.
2411 * config/rs6000/xilinx.h: Delete.
2412 * config/rs6000/xilinx.opt: Delete.
2413 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2414 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2416 2018-05-04 Tom de Vries <tom@codesourcery.com>
2419 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2422 2018-05-04 Richard Biener <rguenther@suse.de>
2425 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2426 (expand_complex_div_wide): Likewise.
2427 (expand_complex_operations_1): Likewise.
2428 (expand_complex_libcall): Preserve EH info of the original stmt.
2429 (tree_lower_complex): Handle removed blocks.
2430 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2431 on complex multiplication and division libcall builtins.
2433 2018-05-04 Richard Biener <rguenther@suse.de>
2436 * fold-const.c (negate_expr_p): Restrict negation of operand
2437 zero of a division to when we know that can happen without
2439 (fold_negate_expr_1): Likewise.
2441 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2444 * real.h (real_nextafter): Declare.
2445 * real.c (real_nextafter): New function.
2446 * fold-const-call.c (fold_const_nextafter): New function.
2447 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2448 CASE_CFN_NEXTTOWARD.
2449 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2450 even when arg1_mode is different from arg0_mode.
2452 2018-05-03 Nathan Sidwell <nathan@acm.org>
2454 * doc/extend.texi (Deprecated Features): Remove
2456 (Backwards Compatibility): Likewise.
2457 * doc/invoke.texi (C++ Language Options): Likewise.
2458 (C++ Dialect Options): Likewise.
2460 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2463 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2464 _mm512_mask_mullox_epi64): New intrinsics.
2466 2018-05-03 Tom de Vries <tom@codesourcery.com>
2469 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2470 dump files): Add offload-tree.
2472 2018-05-03 Richard Biener <rguenther@suse.de>
2474 PR tree-optimization/85615
2475 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2476 to loops not nested in BBs loop father to avoid creating multi-entry
2479 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2481 PR tree-optimization/70291
2482 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2483 arguments. Change return type to tree. Emit libcall as a new
2484 statement rather than replacing existing one when inplace_p is true.
2485 (expand_complex_multiplication_components): New function.
2486 (expand_complex_multiplication): Expand floating-point complex
2487 multiplication using the above.
2488 (expand_complex_division): Rename inner_type parameter to type.
2489 Update expand_complex_libcall call-site.
2490 (expand_complex_operations_1): Update expand_complex_multiplication
2491 and expand_complex_division call-sites.
2493 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2496 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2497 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2498 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2499 the highest significant bit of the shift count mask is clear. In
2500 check whether and[sq]i3 is needed verify that all significant bits
2501 of the shift count other than the highest are set.
2503 2018-05-02 Tom de Vries <tom@codesourcery.com>
2506 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2507 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2508 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2509 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2510 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2511 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2512 __builtin_goacc_parlevel_size.
2514 2018-05-02 Richard Biener <rguenther@suse.de>
2516 PR tree-optimization/85597
2517 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2518 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2521 2018-05-02 Tom de Vries <tom@codesourcery.com>
2524 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2525 dump files): Add ltrans-tree.
2527 2018-05-02 Tom de Vries <tom@codesourcery.com>
2530 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2531 dump files): Add wpa-ipa.
2533 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2535 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2536 powerpc*-*-linux*paired* target.
2537 * config/rs6000/750cl.h: Delete.
2538 * config/rs6000/paired.h: Delete.
2539 * config/rs6000/paired.md: Delete.
2540 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2542 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2543 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2544 comment. Remove paired float support.
2545 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2546 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2548 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2549 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2551 * config/rs6000/rs6000.c: Remove paired float support.
2552 (paired_expand_vector_init, paired_expand_vector_move,
2553 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2554 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2555 paired_expand_builtin, paired_expand_predicate_builtin,
2556 paired_init_builtins): Delete.
2557 * config/rs6000/rs6000.h: Remove paired float support.
2558 * config/rs6000/rs6000.md: Remove paired float support.
2559 (move_from_CR_ov_bit): Delete.
2560 * config/rs6000/rs6000.opt (mpaired): Delete.
2561 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2562 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2564 2018-05-02 Richard Biener <rguenther@suse.de>
2567 * gimplify.c (gimplify_save_expr): When in SSA form allow
2568 SAVE_EXPRs to compute to SSA vars.
2570 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2573 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2574 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2575 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2576 clobber operands[2], instead use a new pseudo. Formatting fixes.
2578 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2580 PR tree-optimization/85586
2581 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2582 exit early for statements in the same group if the accesses are
2585 2018-05-02 Tom de Vries <tom@codesourcery.com>
2588 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2591 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2593 PR tree-optimization/85143
2594 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2596 2018-05-01 Tom de Vries <tom@codesourcery.com>
2599 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2600 not found" error message.
2602 2018-05-01 Tom de Vries <tom@codesourcery.com>
2605 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2606 * vec.c (test_ordered_remove_if): New function.
2607 (vec_c_tests): Call test_ordered_remove_if.
2608 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2609 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2610 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2611 VEC_ORDERED_REMOVE_IF.
2613 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2615 PR tree-optimization/82665
2616 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2617 pointer subtraction where arguments come from a memchr call.
2619 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2621 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2622 --push-state --as-needed and --pop-state instead of --as-needed and
2623 --no-as-needed if ld supports it.
2624 * configure: Regenerated.
2627 * doc/install.texi2html: Replace _002d with - and _002a with * in
2628 generated html files using sed.
2630 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2633 * gcc-rich-location.c (blank_line_before_p): New function.
2634 (use_new_line): New function.
2635 (gcc_rich_location::add_fixit_insert_formatted): New function.
2636 * gcc-rich-location.h
2637 (gcc_rich_location::add_fixit_insert_formatted): New function.
2639 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2641 * selftest.c (assert_streq): Rename "expected" and "actual" to
2642 "val1" and "val2". Extend NULL-handling to cover both inputs
2643 symmetrically, while still requiring both to be non-NULL for a pass.
2644 * selftest.h (assert_streq): Rename "expected" and "actual" to
2646 (ASSERT_EQ): Likewise.
2647 (ASSERT_EQ_AT): Likewise.
2648 (ASSERT_KNOWN_EQ): Likewise.
2649 (ASSERT_KNOWN_EQ_AT): Likewise.
2650 (ASSERT_NE): Likewise.
2651 (ASSERT_MAYBE_NE): Likewise.
2652 (ASSERT_MAYBE_NE_AT): Likewise.
2653 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2654 the assertion to pass.
2655 (ASSERT_STREQ_AT): Likewise.
2657 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2659 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2660 interaction with -pie.
2662 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2664 * selftest.h: Fix alphabetization of per-source-file selftest
2667 2018-04-30 Jason Merrill <jason@redhat.com>
2669 PR c++/61982 - dead stores to destroyed objects.
2670 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2673 2018-04-30 Jason Merrill <jason@redhat.com>
2675 * tree.c (build_clobber): New.
2676 * tree.h: Declare it.
2677 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2679 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2681 * diagnostic-show-locus.c (layout::layout): Update for
2682 location_get_source_line returning a char_span.
2683 (struct char_span): Move to input.h.
2684 (struct correction): Update for fields in char_span becoming
2686 (struct source_line): Update for location_get_source_line
2687 returning a char_span.
2688 (layout::print_line): Likewise.
2689 * edit-context.c (edited_file::print_content): Likewise.
2690 (edited_file::print_diff_hunk): Likewise.
2691 (edited_file::print_run_of_changed_lines): Likewise.
2692 (edited_file::get_num_lines): Likewise.
2693 (edited_line::edited_line): Likewise.
2694 * final.c (asm_show_source): Likewise.
2695 * input.c (location_get_source_line): Convert return type
2696 from const char * to char_span, losing the final "line_len"
2698 (dump_location_info): Update for the above.
2699 (get_substring_ranges_for_loc): Likewise. Use a char_span
2700 when handling the literal within the line.
2701 (test_reading_source_line): Update for location_get_source_line
2702 returning a char_span.
2703 * input.h (class char_span): Move here from
2704 diagnostic-show-locus.c, converting from a struct to a class.
2705 Make data members private.
2706 (char_span::operator bool): New.
2707 (char_span::length): New.
2708 (char_span::get_buffer): New.
2709 (char_span::operator[]): New.
2710 (char_span::subspan): Make const.
2711 (char_span::xstrdup): New.
2712 (location_get_source_line): Convert return type from const char *
2713 to char_span, losing the final "line_size" param.
2715 2018-04-30 Jan Hubicka <jh@suse.cz>
2717 * lto-wrapper.c (ltrans_priorities): New static var.
2718 (cmp_priority): New.
2719 (run_gcc): Read priorities and if doing parallel build order
2720 the Makefile by them.
2722 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2724 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2726 2018-04-30 Richard Biener <rguenther@suse.de>
2728 * tree-cfg.c (verify_address): Remove base argument, add
2729 flag whether to check TREE_ADDRESSABLE and do that.
2730 (verify_expr): Remove.
2731 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2732 (verify_gimple_assign_single): Likewise.
2733 (verify_gimple_switch): Likewise.
2734 (verify_expr_location_1): Dereference tp once. Add (disabled)
2735 piece from verify_expr.
2736 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2738 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2740 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2742 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2744 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2745 (small_data_pattern): Likewise.
2746 (arc_rewrite_small_data): Likewise.
2747 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2748 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2749 (get_symbol_alignment): New function.
2750 (legitimate_small_data_address_p): Likewise.
2751 (legitimate_scaled_address): Update, call
2752 legitimate_small_data_address_p.
2753 (output_sdata): New static variable.
2754 (arc_print_operand): Update how we handle small data operands.
2755 (arc_print_operand_address): Likewise.
2756 (arc_legitimate_address_p): Update, use
2757 legitimate_small_data_address_p.
2758 (arc_rewrite_small_data_p): Remove.
2759 (arc_rewrite_small_data_1): Likewise.
2760 (arc_rewrite_small_data): Likewise.
2761 (small_data_pattern): Likewise.
2762 (compact_sda_memory_operand): Update to use
2763 legitimate_small_data_address_p and get_symbol_alignment.
2764 (prepare_move_operands): Don't rewite sdata pattern.
2765 (prepare_extend_operands): Remove.
2766 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2768 (zero_extendqisi2): Likewise.
2769 (zero_extendhisi2): Likewise.
2770 (extendqihi2): Likewise.
2771 (extendqisi2): Likewise.
2772 (extendhisi2): Likewise.
2776 * config/arc/constraints.md (Usd): Change it to memory constraint.
2778 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2780 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2781 as source of std instructions.
2782 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2783 allow 6-bit constants as source for store instructions.
2784 (movdi_insn): Update instruction pattern to allow 6-bit constants
2785 as source for store instructions.
2787 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2789 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2791 2018-04-30 Nathan Sidwell <nathan@acm.org>
2792 Sandra Loosemore <sandra@codesourcery.com>
2794 * dumpfile.c (dump_open): Allow '-' for stdout.
2795 * doc/invoke.texi (Developer Options): Document dump filename
2796 determination early. Document stdin/stdout selection.
2798 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2800 Microblaze Target: PIC data text relative
2802 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2803 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2805 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2806 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2807 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2808 New addressing mode for data-text relative position indepenedent code.
2809 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2810 'ADDRESS_SYMBOLIC_TXT_REL'.
2811 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2812 (microblaze_legitimate_pic_operand): Exclude function calls from
2813 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2814 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2816 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2817 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2818 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2819 for 'address + offset'.
2820 (microblaze_expand_prologue): Add new function prologue call for
2822 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2823 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2824 table in case of TARGET_PIC_DATA_TEXT_REL.
2825 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2826 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2827 Add new macros 'UNSPEC_TEXT',
2828 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2829 + exclude function calls from 'UNSPEC_PLT' in case of data text
2831 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2832 new target hook for generating address diff vector tables in case of
2834 * doc/tm.texi : Regenerate.
2835 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2836 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2837 of addr diff vector generation.
2838 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2839 target hook definition.
2840 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2841 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2842 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2843 option for data text relative.
2845 2018-04-30 Richard Biener <rguenther@suse.de>
2847 * tree-chrec.h (evolution_function_is_constant_p): Remove
2849 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2851 2018-04-30 Richard Biener <rguenther@suse.de>
2854 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2856 2018-04-30 Richard Biener <rguenther@suse.de>
2858 PR tree-optimization/28364
2859 PR tree-optimization/85275
2860 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2861 copying first exit test.
2863 2018-04-28 Mark Wielaard <mark@klomp.org>
2865 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2867 (dwarf_AT): Handle DW_AT_addr_base.
2868 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2870 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2873 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2874 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2875 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2876 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2878 2018-04-28 Richard Biener <rguenther@suse.de>
2880 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2881 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2882 to reflect use. Only add interesting stmts.
2884 2018-04-27 Martin Jambor <mjambor@suse.cz>
2887 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2888 the jump function allows for passing through aggregate values.
2890 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2892 * input.h (in_system_header_at): Convert from macro to inline
2894 (from_macro_expansion_at): Likewise.
2895 (from_macro_definition_at): Likewise.
2897 2018-04-27 Jeff Law <law@redhat.com>
2899 * config.gcc: Mark tile* targets as deprecated/obsolete.
2901 2018-04-27 Richard Biener <rguenther@suse.de>
2903 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2906 2018-04-27 Richard Biener <rguenther@suse.de>
2908 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2910 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2912 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2913 with Yd constraint. Set "preferred_for_speed" attribute from
2914 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2916 (*movdi_internal): Ditto.
2917 (movti_interunit splitters): Remove
2918 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2919 (movdi_interunit splitters): Ditto.
2920 * config/i386/constraints.md (Ye): Remove.
2921 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2923 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2926 * config/aarch64/constraints.md (Usg): Limit to 31.
2929 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2931 PR tree-optimization/85529
2932 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2933 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2934 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2935 zero extension or masking of the MSB bit.
2936 (optimize_range_tests): Add FIRST_BB argument, pass it through
2937 to optimize_range_tests_var_bound.
2938 (maybe_optimize_range_tests, reassociate_bb): Adjust
2939 optimize_range_tests callers.
2941 2018-04-26 Richard Biener <rguenther@suse.de>
2942 Jakub Jelinek <jakub@redhat.com>
2944 * cgraph.h (symbol_table): Just declare debug method here.
2945 * symtab.c (symbol_table::debug): Define.
2947 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2949 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2951 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2953 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2954 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2955 (*movdi_internal): Substitute Yi and Yj constraint with x
2956 and Ym and Yn constraint with y constraint. Update "isa"
2957 attribute and set "preferred_for_speed" attribute from
2958 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2959 (*movsi_internal): Ditto.
2960 (*movdf_internal): Ditto.
2961 (*movsf_internal): Ditto.
2962 (*zero_extendsidi2): Ditto.
2963 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2964 (sse2_loadld): Ditto.
2965 (*vec_extract<ssevecmodelower>_0): Ditto.
2966 (*vec_extractv4si_0_zext_sse4): Ditto.
2967 (vec_concatv2di): Ditto.
2968 (*vec_dup<mode>): Ditto.
2969 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2970 * config/i386/constraints.md (Yi): Remove.
2975 2018-04-26 Nathan Sidwell <nathan@acm.org>
2977 * dumpfile.c (dump_open): New.
2978 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2979 (dump_finish): Detect stdio/stderr by value not name.
2981 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2983 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2985 2018-04-26 Tom de Vries <tom@codesourcery.com>
2988 * config/nvptx/nvptx.c (verify_neutering_jumps)
2989 (verify_neutering_labels): New function
2990 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2992 2018-04-26 Tom de Vries <tom@codesourcery.com>
2995 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2996 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2999 2018-04-26 Richard Biener <rguenther@suse.de>
3000 Tom de Vries <tom@codesourcery.com>
3003 * lto-streamer-out.c (output_function): Fixup loops if required to match
3004 discovery done in the reader.
3006 2018-04-26 Richard Biener <rguenther@suse.de>
3008 PR tree-optimization/85116
3009 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3010 have a loop exit from the single latch predecessor. Remove
3011 case of header with just condition.
3012 (ch_base::copy_headers): Exclude infinite loops from any
3014 (pass_ch::execute): Record exits.
3016 2018-04-26 Richard Biener <rguenther@suse.de>
3018 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3019 prologue cost vector and pass it to vect_get_load_cost.
3020 (vect_get_peeling_costs_all_drs): Likewise.
3021 (vect_peeling_hash_get_lowest_cost): Likewise.
3022 (vect_enhance_data_refs_alignment): Likewise.
3024 2018-04-26 Richard Biener <rguenther@suse.de>
3027 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3028 checking of integer<->pointer conversions.
3029 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3030 sign-/zero-extending pointer types.
3031 (expand_omp_for_static_chunk): Likewise.
3033 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3034 Jean Lee <xiaoyur347@gmail.com>
3036 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3037 (TARGET_ASAN_SHADOW_OFFSET): Define.
3038 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3039 true for -fsanitize=address.
3041 2018-04-25 Mark Wielaard <mark@klomp.org>
3043 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3046 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3048 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3049 than "alu", remove explicit "memory" and "imm_disp" attributes.
3050 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3053 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3054 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3055 gen_lowpart_no_emit.
3057 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3060 * config/i386/i386.c (ix86_expand_builtin): Change memory
3061 operand to XI, extend p0 to Pmode.
3062 * config/i386/i386.md: Change unspec volatile and operand
3063 1 mode to XI, change operand 0 mode to P.
3065 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3067 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3068 GET_MODE_MASK before any checking.
3069 (nds32_can_use_bset_p): Likewise.
3070 (nds32_can_use_btgl_p): Likewise.
3072 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3074 * config/nds32/nds32-doubleword.md: New define_split pattern for
3075 illegal register number.
3077 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3079 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3081 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3083 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3085 2018-04-25 Richard Biener <rguenther@suse.de>
3087 * lto-streamer.h (LTO_major_version): Bump to 8.
3089 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3091 * BASE-VER: Set to 9.0.0.
3093 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3095 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3096 in __abskf2 and __powikf2.
3098 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3101 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3102 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3103 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3104 Use the above on operand 2. Reindent.
3105 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3107 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3110 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3112 * config/i386/i386.opt (mcet): Removed.
3113 * doc/install.texi: Remove -mcet documentation.
3114 * doc/invoke.texi: Likewise.
3116 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3119 * doc/install.texi: Remove -mcet from bootstrap-cet.
3121 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3124 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3125 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3129 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3130 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3131 containing a CONST_VECTOR.
3133 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3135 * doc/install.texi: Update newlib dependency for nvptx.
3137 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3140 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3141 instead of INTVAL when shifting x left.
3143 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3145 PR tree-optimization/85478
3146 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3147 vect_grouped_store_supported for single element vectors.
3149 2018-04-24 Richard Biener <rguenther@suse.de>
3152 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3153 load cost increase to the case of non-constant step.
3155 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3158 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3159 destination if any_malformed_asm.
3161 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3164 * expr.c (store_field): In the bitfield case, if the value comes from
3165 a function call and is returned in registers by means of a PARALLEL,
3166 do not change the mode of the temporary unless BLKmode and VOIDmode.
3168 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3170 PR rtl-optimization/85423
3171 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3172 dependencies to debug insns when the previous insn is non-debug.
3174 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3176 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3177 enums into a single definition.
3178 (fls): Fix predicates and printing.
3181 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3183 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3184 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3185 and short u6 immediate.
3186 (check_if_valid_sleep_operand): Remove.
3187 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3189 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3191 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3192 flag_always_save_lp condition.
3193 * config/nds32/nds32.opt (malways-save-lp): New option.
3195 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3197 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3198 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3199 * config/nds32/nds32.h
3200 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3201 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3203 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3205 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3206 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3208 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3209 Chung-Ju Wu <jasonwucj@gmail.com>
3211 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3213 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3214 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3216 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3218 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3220 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3222 * config/nds32/nds32-protos.h (nds32_data_alignment,
3223 nds32_local_alignment): Declare.
3224 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3225 nds32_local_alignment): New functions.
3226 (TARGET_CONSTANT_ALIGNMENT): Define.
3227 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3229 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3231 * config/nds32/nds32.c
3232 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3233 (TARGET_MODES_TIEABLE_P): Likewise.
3235 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3237 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3240 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3241 Chung-Ju Wu <jasonwucj@gmail.com>
3243 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3244 for unaligned access.
3245 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3247 * config/nds32/nds32-intrinsic.md: Likewise.
3248 * config/nds32/nds32_intrinsic.h: Likewise.
3249 * config/nds32/nds32.h (nds32_builtins): Likewise.
3250 * config/nds32/nds32.opt (munaligned-access): New option.
3251 * config/nds32/nds32.c (nds32_asm_file_start): Display
3252 flag_unaligned_access status.
3254 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3256 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3257 -mno-relax is present.
3258 * config/riscv/linux.h (LINK_SPEC): Ditto.
3260 2018-04-20 Martin Sebor <msebor@redhat.com>
3263 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3265 (gimple_fold_builtin_stxcpy_chk): Same.
3266 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3268 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3271 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3272 __powikf2 when long double is IEEE 128-bit.
3274 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3276 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3277 step to make sure stack always aligned.
3279 2018-04-20 Carl Love <cel@us.ibm.com>
3282 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3283 size check for arg0.
3285 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3286 Tom de Vries <tom@codesourcery.com>
3289 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3290 Emit insns for calls too.
3291 (nvptx_find_par): Always look for worker-level predecessor insn.
3292 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3294 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3295 (nvptx_process_pars): Propagate frames for calls.
3297 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3300 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3302 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3303 (ix86_handle_option): Don't handle OPT_mibt.
3304 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3306 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3308 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3309 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3310 (ix86_target_macros): Define __CET__ with flag_cf_protection
3311 for -fcf-protection.
3312 * config/i386/i386.c (isa2_opts): Remove -mibt.
3313 * config/i386/i386.h (TARGET_IBT): Removed.
3314 (TARGET_IBT_P): Likewise.
3315 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3316 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3317 * config/i386/i386.opt (mcet): Update help message.
3320 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3321 -mcet as an alias for -mshstk.
3323 2018-04-20 Richard Biener <rguenther@suse.de>
3326 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3327 complexity by forcing a single use of the multiply operand.
3329 2018-04-20 Martin Jambor <mjambor@suse.cz>
3332 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3333 recursion dependency to only apply to non-clones.
3335 2018-04-20 Martin Jambor <mjambor@suse.cz>
3338 * ipa-cp.c (create_specialized_node): Check that clones of
3339 self-recursive edges exist during IPA-CP.
3341 2018-04-19 Toon Moene <toon@moene.org>
3343 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3346 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3348 PR tree-optimization/85467
3349 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3350 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3351 VECTOR_CST element to type.
3353 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3356 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3357 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3358 (builtin_longjmp): Likewise.
3359 (save_stack_nonlocal): New pattern.
3360 (restore_stack_nonlocal): Likewise.
3362 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3365 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3366 Replace ASM_OUTPUT_LABEL with fprintf.
3368 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3371 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3372 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3373 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3374 define __IBT__ and __SHSTK__ for -fcf-protection.
3375 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3377 (ix86_trampoline_init): Likewise.
3378 (x86_output_mi_thunk): Likewise.
3379 (ix86_notrack_prefixed_insn_p): Likewise.
3380 (ix86_option_override_internal): Don't disallow -fcf-protection.
3381 * config/i386/i386.md (rdssp<mode>): Also enable for
3383 (incssp<mode>): Likewise.
3384 (nop_endbr): Likewise.
3385 * config/i386/i386.opt (mcet): Change help message to built-in
3389 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3390 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3391 enable CET built-in functions.
3393 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3395 * common/config/i386/i386-common.c
3396 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3397 OPTION_MASK_ISA_MOVDIRI_UNSET,
3398 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3399 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3400 * config.gcc (movdirintrin.h): New header.
3401 * config/i386/cpuid.h (bit_MOVDIRI,
3402 bit_MOVDIR64B): New bits.
3403 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3405 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3406 (VOID, PVOID, PCVOID)): New function types.
3407 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3408 __builtin_ia32_directstoreu_u64,
3409 __builtin_ia32_movdir64b): New builtins.
3410 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3411 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3413 (ix86_valid_target_attribute_inner_p): Ditto.
3414 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3415 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3416 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3417 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3418 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3419 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3420 (movdiri<mode>, movdir64b_<mode>): New.
3421 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3422 * config/i386/immintrin.h: Include movdirintrin.h.
3423 * config/i386/movdirintrin.h: New file.
3424 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3426 2018-04-19 Richard Biener <rguenther@suse.de>
3429 * cfg.c (clear_bb_flags): When loop state says we have
3430 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3432 2018-04-19 Richard Biener <rguenther@suse.de>
3434 PR tree-optimization/84737
3435 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3436 copying restrict info.
3437 (vect_setup_realignment): Use it.
3438 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3439 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3440 the first DR to all generated stores.
3441 (vectorizable_load): Likewise for loads.
3443 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3445 PR tree-optimization/85446
3446 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3447 the integral and pointer types to have the same precision.
3449 * doc/install.texi: Document --disable-cet being the default and
3452 2018-04-18 Martin Liska <mliska@suse.cz>
3454 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3457 2018-04-18 Martin Liska <mliska@suse.cz>
3460 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3463 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3464 arguments if they are comparable.
3466 2018-04-18 Martin Liska <mliska@suse.cz>
3469 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3472 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3475 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3478 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3479 ENDBR after calling __morestack.
3481 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3484 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3485 by using gcc_base_ver to generate a gcc_driver_version, and use
3486 it when generating GCC_DRIVER_NAME.
3487 * configure: Regenerate.
3489 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3492 * config.gcc: Obsolete powerpc*-*-*spe*.
3494 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3497 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3498 (stabstr_D): Change type of unum from unsigned int to
3499 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3502 2018-04-17 Jim Wilson <jimw@sifive.com>
3505 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3506 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3507 Set arg_pointer_offset after using pretend_args_size.
3509 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3511 PR rtl-optimization/85431
3512 * dse.c (record_store): Ignore zero width stores.
3515 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3516 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3517 __builtin_stack_restore rather than after it.
3518 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3519 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3520 argument instead of virtual_dynamic_stack_rtx.
3522 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3524 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3526 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3527 Add note to error message to explain internal mapping of overloaded
3528 built-in function name to non-overloaded built-in function name.
3529 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3532 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3535 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3536 where the inputs overlap with the output.
3538 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3540 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3541 (=v, v) alternative and explicit "memory" attribute.
3542 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3543 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3545 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3546 "sselog1" type instead of "sselog".
3547 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3548 "sselog". Remove explicit "memory" attribute.
3549 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3550 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3552 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3553 "sselog1" type instead of "sselog". Remove explicit "memory"
3555 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3556 use "sselog1" type instead of "sselog". Remove explicit "memory"
3558 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3559 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3561 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3562 "sselog1" type instead of "sselog". Remove explicit "memory"
3564 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3565 use "sselog1" type instead of "sselog". Remove explicit "memory"
3569 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3572 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3575 2018-04-17 Martin Jambor <mjambor@suse.cz>
3578 * ipa-cp.c (create_specialized_node): Call
3579 expand_all_artificial_thunks if necessary.
3581 2018-04-17 Martin Liska <mliska@suse.cz>
3584 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3585 in message, remote space in between '_G' and '('.
3587 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3590 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3591 avx512f_vmcmp<mode>3<round_saeonly_name>,
3592 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3593 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3594 avx512f_rndscale<mode><round_saeonly_name>,
3595 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3596 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3597 Use %<iptr>2 instead of %2 for -masm=intel.
3598 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3599 avx512f_vcvttss2usi<round_saeonly_name>,
3600 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3602 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3603 avx512f_vcvttsd2usi<round_saeonly_name>,
3604 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3605 Use %q1 instead of %1 for -masm=intel.
3606 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3607 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3608 of %3 for -masm=intel.
3609 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3611 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3613 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3615 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3616 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3618 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3620 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3621 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3622 %g1 and one with %0 and %1.
3623 (avx512er_vmrcp28<mode><round_saeonly_name>,
3624 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3626 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3627 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3628 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3629 of %0 and %{%4%} for -masm=intel.
3630 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3631 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3632 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3633 order of %0 and %{%5%}%{z%} for -masm=intel.
3635 2018-04-17 Jan Hubicka <jh@suse.cz>
3638 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3640 2018-04-17 Martin Liska <mliska@suse.cz>
3643 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3644 for target_clone error message. Make default implementation
3645 clone to be a local declaration.
3646 (separate_attrs): Add new argument and check for an empty
3648 (expand_target_clones): Handle it.
3649 (ipa_target_clone): Make redirection just for target_clones
3652 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3653 Tom de Vries <tom@codesourcery.com>
3656 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3657 tiled basic blocks without omp continue statements.
3659 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3662 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3663 vec_extract expression as having side effects to make sure it gets
3666 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3669 * config/i386/i386.c (get_builtin_code_for_version): Check
3672 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3675 * gcc/config.gcc: Support "skylake".
3676 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3678 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3679 (processor_target_table): Add "skylake".
3680 (ix86_option_override_internal): Add "skylake".
3681 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3682 PROCESSOR_CANNONLAKE.
3683 (get_builtin_code_for_version): Fix priority for
3684 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3685 PROCESSOR_SKYLAKE-AVX512.
3686 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3687 (processor_type): Add PROCESSOR_SKYLAKE.
3689 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3690 Jason Merrill <jason@redhat.com>
3693 * convert.c (convert_to_integer_1): Use direct recursion for
3694 enumeral types and types with a precision less than the number
3695 of bits in their mode.
3697 2018-04-16 Julia Koval <julia.koval@intel.com>
3700 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3701 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3703 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3706 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3707 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3708 and -mno-direct-move.
3710 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3713 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3714 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3715 Ensure negative shifts result in {0}.
3717 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3719 PR rtl-optimization/79916
3720 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3721 regs (if any) to define how to gnerate SD moves when LRA is in
3724 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3726 PR rtl-optimization/85393
3727 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3728 * except.c (expand_dw2_landing_pad_for_region): Make static.
3729 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3730 a label and unconditional jump to old_bb, rather than
3731 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3734 PR rtl-optimization/85376
3735 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3736 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3737 instead of a specific value.
3739 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3740 Bin Cheng <bin.cheng@arm.com>
3742 PR tree-optimization/82965
3743 PR tree-optimization/83991
3744 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3745 by_profile_only parameter.
3746 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3747 information if the loop was predicted to iterate too many times.
3748 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3750 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3753 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3756 2018-04-13 Martin Liska <mliska@suse.cz>
3757 Jakub Jelinek <jakub@redhat.com>
3760 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3761 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3762 * builtins.c (expand_builtin_memory_copy_args): Use
3763 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3764 handle dest_addr == pc_rtx.
3766 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3769 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3770 asked to not generate direct moves.
3771 (fix_trunc<mode>si2_stfiwx): Similar.
3772 (fix_trunc<mode>si2_internal): Similar.
3774 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3777 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3778 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3779 lookup if dest in some wider mode is known to be const0_rtx and
3780 if so, record permanent equivalence for it to be ZERO_EXTEND of
3781 the narrower mode destination.
3783 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3785 * lto-streamer-out.c (output_function): Revert 259346.
3786 * omp-expand.c (expand_oacc_for): Likewise.
3788 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3790 PR rtl-optimization/85354
3791 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3792 * sel-sched.c (sel_global_init): ... here.
3794 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3797 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3798 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3799 mode for PE-COFF targets.
3800 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3801 (i386_pe_asm_lto_end): Likewise.
3802 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3803 (TARGET_ASM_LTO_END): Likewise.
3804 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3805 (i386_pe_asm_lto_start): New function.
3806 (i386_pe_asm_lto_end): Likewise.
3808 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3809 Richard Biener <rguenther@suse.de>
3812 * lto-streamer-out.c (output_function): Fix CFG loop state before
3814 * omp-expand.c (expand_oacc_for): Handle calls to internal
3815 functions like regular functions.
3817 2018-04-12 Richard Biener <rguenther@suse.de>
3820 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3821 for the early LTO debug to properly generate references to it
3822 during DIE emission. Do not re-use that for the skeleton for
3824 (dwarf2out_early_finish): Likewise.
3826 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3829 * config/i386/sse.md
3830 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3831 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3832 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3833 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3834 and output is a reg, avoid creating invalid lowpart subreg, but
3835 instead split into a 512-bit move. Don't split if not AVX512VL,
3836 input is xmm16+ reg and output is a mem.
3837 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3838 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3839 xmm16+ reg and output is a mem.
3841 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3843 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3844 also for flag_dwarf2_cfi_asm.
3846 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3848 PR rtl-optimization/85342
3849 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3850 a bool scalar var inside of the loop instead. Don't try to update
3851 recog_data.operand after failed apply_change_group.
3853 2018-04-12 Tom de Vries <tom@codesourcery.com>
3856 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3857 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3858 array with flexible array member as array without given dimension.
3859 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3860 argument for undefined param to true.
3862 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3865 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3866 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3867 from PowerPC section.
3868 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3869 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3870 help text that is too long.
3871 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3872 help text that is too long.
3873 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3874 help text that is too long.
3876 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3878 * config/alpha/alpha.md (stack_probe_internal): Rename
3879 from "probe_stack". Update all callers.
3881 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3883 PR rtl-optimization/84566
3884 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3885 sched_macro_fuse_insns.
3887 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3890 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3891 (compute_block_dependences): ... from here.
3893 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3895 PR tree-optimization/85331
3896 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3897 from int to HOST_WIDE_INT.
3899 2018-04-11 Martin Jambor <mjambor@suse.cz>
3902 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3903 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3904 not the same as the source val.
3905 (cgraph_edge_brings_value_p): New parameter.
3906 (gather_edges_for_value): Pass destination value to
3907 cgraph_edge_brings_value_p.
3908 (perhaps_add_new_callers): Likewise.
3909 (get_info_about_necessary_edges): Likewise and exclude values brought
3910 only by self-recursive edges.
3911 (create_specialized_node): Redirect only clones of self-calling edges.
3912 (+self_recursive_pass_through_p): New function.
3913 (find_more_scalar_values_for_callers_subset): Use it.
3914 (find_aggregate_values_for_callers_subset): Likewise.
3915 (known_aggs_to_agg_replacement_list): Removed.
3916 (decide_whether_version_node): Re-calculate known constants for all
3917 remaining context clones.
3919 2018-04-11 Richard Biener <rguenther@suse.de>
3922 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3923 from early DWARF output.
3924 (dwarf2out_early_finish): Output line info unconditionally into
3925 early DWARF and add reference to it.
3927 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3930 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3931 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3932 other than V2DFmode using iptr mode attribute.
3933 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3935 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3937 PR rtl-optimization/84659
3938 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3940 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3943 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3945 (output_loc_list): Pass address of a dummy size variable even in the
3946 locview handling loop.
3947 (index_location_lists): Add comment on why skip_loc_list_entry can't
3950 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3953 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3956 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3959 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3960 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3961 and -mstring-compare-inline-limit.
3963 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3966 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3967 for stack clash protection in a register whenever we need it to be in
3970 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3972 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3973 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3975 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3978 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3980 (mlong-double-): Ditto.
3981 * config/rs6000/sysv4.opt (msdata=): Ditto.
3982 (mtls-size=): Ditto.
3984 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3986 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3987 erroneous entries for
3988 "vector int vec_ldl (int, long int *)", and
3989 "vector unsigned int vec_ldl (int, unsigned long int *)".
3990 Add comments and entries for
3991 "vector bool char vec_ldl (int, bool char *)",
3992 "vector bool short vec_ldl (int, bool short *)",
3993 "vector bool int vec_ldl (int, bool int *)",
3994 "vector bool long long vec_ldl (int, bool long long *)",
3995 "vector pixel vec_ldl (int, pixel *)",
3996 "vector long long vec_ldl (int, long long *)",
3997 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3998 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3999 type tree bool_long_long_type_node and correct definition of
4000 bool_V2DI_type_node to make reference to this new type tree.
4001 (rs6000_mangle_type): Replace erroneous reference to
4002 bool_long_type_node with bool_long_long_type_node.
4003 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4004 comments to emphasize sign distinctions for char and int types and
4005 replace RS6000_BTI_bool_long constant with
4006 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4007 use of RS6000_BTI_pixel.
4008 (bool_long_type_node): Remove this macro definition.
4009 (bool_long_long_type_node): New macro definition
4011 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4013 PR rtl-optimization/85300
4014 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4015 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4016 simplify_unary_operation fails.
4018 2018-04-10 Martin Liska <mliska@suse.cz>
4020 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4021 cgraph_edge and ipa_ref.
4023 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4027 * config/i386/sse.md
4028 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4029 computation of the VEC_MERGE selector from mask.
4030 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4031 Fix decoding of the VEC_MERGE selector into mask.
4033 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4035 PR tree-optimization/85286
4036 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4038 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4040 * final.c (final_1): Set insn_last_address as well as
4041 insn_current_address.
4043 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4046 * explow.c (emit_stack_probe): Call validize_mem on memory location
4047 before passing it to gen_probe_stack. Create address operand and
4048 legitimize it for the probe_stack_address case.
4050 2018-04-09 Jan Hubicka <jh@suse.cz>
4053 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4054 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4055 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4057 (free_lang_data): Rebuild type inheritance graph.
4059 2018-04-09 Martin Sebor <msebor@redhat.com>
4061 * invoke.texi (-finline-small-functions): Mention other optimization
4063 (-findirect-inlining, -fpartial-inlining): Same.
4064 (-finline-functions-called-once): Same.
4065 (-freorder-blocks-and-partition): Same.
4067 2018-04-09 Jan Hubicka <jh@suse.cz>
4070 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4071 jumps; choose last target that matches the criteria (i.e.
4072 no partition changes for non-crossing jumps).
4073 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4074 support for redirecting crossing jumps to non-crossing.
4076 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4078 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4079 also for naked functions.
4081 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4083 * config/arc/arc.md (add_shift): New pattern.
4084 (add_shift2): Likewise.
4085 (sub_shift): Likewise.
4086 (sub_shift_cmp0_noout): Likewise.
4087 (compare_si_ashiftsi): Likewise.
4088 (xbfu_cmp0_noout): New combine pattern.
4089 (xbfu_cmp0"): Likewise.
4090 (movsi_set_cc_insn): Place the predicable variant first.
4091 (commutative_binary_cmp0_noout): Remove clobber.
4092 (commutative_binary_cmp0): New pattern.
4093 (noncommutative_binary_cmp0): Likewise.
4094 (noncommutative_binary_cmp0_noout): Likewise.
4095 (noncommutative_binary_comparison_result_used): Removed.
4096 (rsub_cmp0): New pattern.
4097 (rsub_cmp0_noout): Likewise.
4098 (extzvsi): Changed, keep only meaningful variants.
4099 (SQH, SEZ): New iterators.
4100 (SQH_postfix): New mode attribute.
4101 (SEZ_prefix): New code attribute.
4102 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4103 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4104 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4106 (noncommutative_operator): Check the availability of barrel
4109 2018-04-09 Richard Biener <rguenther@suse.de>
4111 PR tree-optimization/85284
4112 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4113 Only use the niter constraining form of simple_iv when the exit
4116 2018-04-09 Tom de Vries <tom@codesourcery.com>
4119 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4120 (define_expand "*memory_barrier"): New define_expand.
4121 (define_insn "memory_barrier"): New insn.
4123 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4125 PR rtl-optimization/80463
4126 PR rtl-optimization/83972
4127 PR rtl-optimization/83480
4129 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4130 correct producer for the insn.
4131 (tidy_control_flow): Fixup seqnos in case of debug insns.
4133 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4135 PR rtl-optimization/83913
4137 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4138 different sched-times when merging exprs.
4140 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4142 PR rtl-optimization/83962
4144 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4145 tidy_fallthru_edge and tidy_control_flow.
4147 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4149 PR rtl-optimization/83530
4151 * sel-sched.c (force_next_insn): New global variable.
4152 (remove_insn_for_debug): When force_next_insn is true, also leave only
4153 next insn in the ready list.
4154 (sel_sched_region): When the region wasn't scheduled, make another pass
4155 over it with force_next_insn set to 1.
4157 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4159 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4161 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4162 for interrupt control.
4163 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4164 functions for interrupt control.
4165 * config/nds32/nds32-intrinsic.md: Likewise.
4166 * config/nds32/nds32_intrinsic.h: Likewise.
4167 * config/nds32/nds32.h (nds32_builtins): Likewise.
4169 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4171 * config/nds32/nds32.c (nds32_init_machine_status,
4172 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4173 strict_aligned_p field.
4174 (nds32_expand_to_rtl_hook): New function.
4175 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4176 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4178 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4179 Chung-Ju Wu <jasonwucj@gmail.com>
4181 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4182 * config/nds32/nds32-n7.md: New file.
4183 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4184 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4186 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4187 * config/nds32/nds32.md (pipeline_model): Add n7.
4188 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4189 * config/nds32/pipelines.md: Include n7 settings.
4191 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4192 Chung-Ju Wu <jasonwucj@gmail.com>
4194 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4195 * config/nds32/nds32-e8.md: New file.
4196 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4197 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4199 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4200 * config/nds32/nds32.md (pipeline_model): Add e8.
4201 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4202 * config/nds32/pipelines.md: Include e8 settings.
4204 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4205 Chung-Ju Wu <jasonwucj@gmail.com>
4207 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4208 * config/nds32/nds32-n8.md: New file.
4209 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4210 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4212 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4213 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4214 * config/nds32/nds32.md (pipeline_model): Add n8.
4215 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4216 * config/nds32/pipelines.md: Include n8 settings.
4218 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4219 Chung-Ju Wu <jasonwucj@gmail.com>
4221 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4222 * config/nds32/nds32-n9-2r1w.md: New file.
4223 * config/nds32/nds32-n9-3r2w.md: New file.
4224 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4225 nds32_register_ports): New or modify for cpu n9.
4226 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4228 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4229 * config/nds32/nds32-utils.c: New file.
4230 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4231 TARGET_MUL_SLOW): Define.
4232 * config/nds32/nds32.md (pipeline_model): New attribute.
4233 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4234 New options that support cpu n9.
4235 * config/nds32/pipelines.md: Include n9 settings.
4236 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4238 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4240 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4241 information if necessary.
4242 (output_cond_branch_compare_zero): Likewise.
4243 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4244 (nds32_target_alignment): Refine for alignment.
4245 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4246 (FUNCTION_BOUNDARY): Modify.
4247 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4249 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4251 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4253 * config/nds32/constants.md (unspec_volatile_element): Add values for
4254 TLB operation and data prefetch.
4255 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4256 functions for TLB operation and data prefetch.
4257 * config/nds32/nds32-intrinsic.md: Likewise.
4258 * config/nds32/nds32_intrinsic.h: Likewise.
4259 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4260 (nds32_print_operand): Likewise.
4261 * config/nds32/nds32.h (nds32_builtins): Likewise.
4263 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4264 Andrew Pinski <pinsika@gcc.gnu.org>
4267 * match.pd: Use constant_boolean_node of correct type instead of
4268 boolean_true_node or boolean_false_node for simplifying
4269 pointer comparisons to zero.
4271 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4273 PR tree-optimization/80021
4274 * tree.c (verify_type_variant): Make error call in verify_variant_match
4275 translatable and remove final full stop.
4277 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4279 * config/nds32/constants.md (unspec_volatile_element): Add
4280 UNSPEC_VOLATILE_EH_RETURN.
4281 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4282 nds32_output_stack_pop): Support dwarf exception handling process.
4283 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4284 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4285 exception handling process.
4286 (nds32_compute_stack_frame): Likewise.
4287 (nds32_return_addr_rtx): Likewise.
4288 (nds32_initial_elimination_offset): Likewise.
4289 (nds32_expand_prologue): Likewise.
4290 (nds32_expand_epilogue): Likewise.
4291 (nds32_dynamic_chain_address): New function.
4292 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4294 (DYNAMIC_CHAIN_ADDRESS): Define.
4295 (EH_RETURN_DATA_REGNO): Define.
4296 (EH_RETURN_STACKADJ_RTX): Define.
4297 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4298 patterns for dwarf exception handling.
4300 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4302 * config/nds32/nds32.h: Clean up obsolete macros.
4304 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4306 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4307 Add enum values for particular instructions.
4308 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4309 particular intrinsic functions.
4310 * config/nds32/nds32-intrinsic.md: Likewise.
4311 * config/nds32/nds32_intrinsic.h: Likewise.
4312 * config/nds32/nds32.h (nds32_builtins): Likewise.
4313 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4314 (btst, ave): New patterns for particular instructions.
4316 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4318 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4319 Add enum values for atomic load/store and memory sync.
4320 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4322 * config/nds32/nds32-intrinsic.md: Likewise.
4323 * config/nds32/nds32_intrinsic.h: Likewise.
4324 * config/nds32/nds32.h (nds32_builtins): Likewise.
4326 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4328 PR tree-optimization/85257
4329 * fold-const.c (native_encode_vector): If not all elts could fit
4330 and off is -1, return 0 rather than offset.
4331 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4332 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4333 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4334 adjust buffer in native_interpret_expr call.
4336 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4338 * config/nds32/constants.md (unspec_volatile_element): Add cache
4339 control enum values.
4340 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4341 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4342 * config/nds32/nds32.c (nds32_cctl_names): New.
4343 (nds32_print_operand): Handle cache control register names.
4344 * config/nds32/nds32.h (nds32_builtins): New enum values.
4345 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4347 * config/nds32/nds32.md (type): Add mmu.
4348 * config/nds32/pipelines.md (simple_insn): Add mmu.
4350 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4352 * config/nds32/nds32.md (type): Remove call.
4353 * config/nds32/pipelines.md (simple_insn): Likewise.
4355 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4357 * config/nds32/constants.md (unspec_volatile_element): Add
4358 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4359 UNSPEC_VOLATILE_FMFCFG.
4360 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4361 description for fmfcfg and fmfcsr.
4362 (bdesc_1arg): Add fmtcsr.
4363 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4364 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4365 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4366 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4367 unspec_fmfcfg): New patterns.
4368 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4369 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4370 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4371 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4372 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4373 __nds32__fmfcfg): Define.
4375 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4377 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4378 intrinsic register names.
4379 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4380 intrinsic register enum values and macros.
4382 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4384 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4385 for load/store addressing form.
4386 (nds32_print_operand_address): Likewise.
4388 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4391 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4392 based on LABEL_REF. Remove useless assertion.
4393 (pic_address_needs_scratch): Fix formatting.
4394 (sparc_legitimize_pic_address): Minor tweaks.
4395 (sparc_delegitimize_address): Adjust assertion accordingly.
4396 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4397 into symbolic_operand.
4398 (movsi_high_pic_label_ref): Likewise.
4399 (movsi_lo_sum_pic_label_ref): Likewise.
4400 (movdi_pic_label_ref): Likewise.
4401 (movdi_high_pic_label_ref): Likewise.
4402 (movdi_lo_sum_pic_label_ref): Likewise.
4404 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4406 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4407 custom LIB_SPEC setup.
4409 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4410 Kito Cheng <kito.cheng@gmail.com>
4412 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4413 * config/riscv/freebsd.h: New.
4415 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4417 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4418 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4421 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4422 Kito Cheng <kito.cheng@gmail.com>
4424 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4425 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4426 * config/nds32/nds32-protos.h (nds32_output_call,
4427 nds32_output_return): Declare.
4428 * config/nds32/nds32.md: Refine all the call and return patterns.
4430 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4433 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4434 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4436 PR rtl-optimization/84872
4437 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4438 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4441 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4443 * expr.c (copy_blkmode_to_reg): Revert 254862.
4444 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4446 2018-04-06 Richard Biener <rguenther@suse.de>
4449 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4450 after seeing a component reference with an adjacent field. Treat
4451 refs to arrays at struct end of external decls similar to
4452 refs to unconstrained commons.
4454 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4457 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4458 look through SAVE_EXPRs with non-side-effects argument. Adjust
4460 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4463 2018-04-06 Richard Biener <rguenther@suse.de>
4466 * alias.c (find_base_term): New wrapper around find_base_term
4467 unwinding CSELIB_VAL_PTR changes.
4468 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4471 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4473 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4475 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4476 constant definitions.
4477 ("nop"): lr 0,0 -> nopr r0
4478 ("nop_lr0", "nop_lr1"): New insn definitions.
4480 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4482 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4483 NDS32_V3PUSH_AVAILABLE_P macro.
4485 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4486 Chung-Ju Wu <jasonwucj@gmail.com>
4488 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4489 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4490 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4491 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4492 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4493 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4494 * config/nds32/constraints.md: New constraints and checking for hard
4495 float configuration.
4496 * config/nds32/iterators.md: New mode iterator and attribute for hard
4497 float configuration.
4498 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4500 * config/nds32/nds32-fpu.md: New file.
4501 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4502 deal with hard float code generation.
4503 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4505 (abi_type, float_reg_number): New enum type.
4506 * config/nds32/nds32-predicates.c: New predicates for hard float.
4507 * config/nds32/nds32-protos.h: Declare functions for hard float.
4508 * config/nds32/nds32.c: Implementation for hard float configuration.
4509 * config/nds32/nds32.h: Definitions for hard float configuration.
4510 * config/nds32/nds32.md: Include hard float machine description and
4511 modify patterns for hard float configuration.
4512 * config/nds32/nds32.opt: New options for hard float configuration.
4513 * config/nds32/predicates.md: New predicates for hard float
4516 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4518 * common/config/nds32/nds32-common.c
4519 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4521 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4524 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4525 CONSTRUCTOR_ELT (ctor, ...)->value.
4527 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4530 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4532 2018-04-05 Tom de Vries <tom@codesourcery.com>
4535 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4538 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4539 Kito Cheng <kito.cheng@gmail.com>
4541 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4542 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4543 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4544 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4546 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4547 Kito Cheng <kito.cheng@gmail.com>
4549 * config/nds32/constraints.md (Ufe): New memory constraint.
4550 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4551 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4552 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4554 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4555 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4557 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4559 * config/nds32/nds32.md: Use optimize_size in the condition for
4560 alu-shift instructions.
4562 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4564 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4566 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4568 * config/nds32/nds32.md (negsi2): Refine pattern.
4570 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4571 Chung-Ju Wu <jasonwucj@gmail.com>
4573 * config/nds32/iterators.md (shift_rotate): New code iterator.
4574 (shift): New code attribute.
4575 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4576 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4577 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4578 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4579 bit-wise operations.
4580 (andsi3, *andsi3): Ditto.
4581 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4582 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4583 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4584 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4585 nds32_ior_operand, nds32_xor_operand): New predicates.
4587 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4589 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4590 (addsi3, subsi3): ... this.
4592 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4594 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4596 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4598 * config/nds32/nds32.md: Adjust indention.
4600 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4602 * config/nds32/nds32.md (feature): New attribute.
4604 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4606 * config/nds32/nds32.md (subtype): New attribute.
4608 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4611 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4612 expansion to perform a bitwise AND of the argument followed by a
4613 boolean negation of the result.
4615 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4617 PR rtl-optimization/84878
4618 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4619 the basic block. Assert the use reference is not artificial and that
4620 it has an associated insn.
4622 2018-04-04 Michael Matz <matz@suse.de>
4624 * builtins.c (compute_objsize): Pass correct operand
4625 to array_at_struct_end_p.
4627 2018-04-04 Richard Biener <rguenther@suse.de>
4630 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4631 from contexts for DINFO_LEVEL_TERSE and below.
4633 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4635 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4636 resiter_operand condition.
4637 * config/nds32/nds32.md (*move<mode>): Ditto.
4639 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4640 Monk Chiang <sh.chiang04@gmail.com>
4642 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4644 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4646 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4648 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4649 Kito Cheng <kito.cheng@gmail.com>
4651 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4652 nds32_cond_code_str, output_cond_branch,
4653 output_cond_branch_compare_zero, nds32_expand_cbranch,
4654 nds32_expand_cstore, nds32_expand_movcc,
4655 nds32_output_cbranchsi4_equality_zero,
4656 nds32_output_cbranchsi4_equality_reg,
4657 nds32_output_cbranchsi4_equality_reg_or_const_int,
4658 nds32_output_cbranchsi4_greater_less_zero: New functions.
4659 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4660 nds32_expand_cstore, nds32_expand_movcc,
4661 nds32_output_cbranchsi4_equality_zero,
4662 nds32_output_cbranchsi4_equality_reg,
4663 nds32_output_cbranchsi4_equality_reg_or_const_int,
4664 nds32_output_cbranchsi4_greater_less_zero): Declare.
4665 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4666 nds32_rimm11s_operand): New predicates.
4667 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4668 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4671 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4673 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4674 * config/nds32/nds32.md: Ditto.
4675 * config/nds32/pipelines.md: Ditto.
4677 2018-04-04 Richard Biener <rguenther@suse.de>
4679 PR tree-optimization/85168
4680 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4681 propagating abnormals.
4683 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4685 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4687 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4688 Kito Cheng <kito.cheng@gmail.com>
4690 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4691 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4692 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4693 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4694 * config/nds32/nds32.md (sibcall_internal): New.
4695 (sibcall_register): Remove.
4696 (sibcall_immediate): Remove.
4697 (sibcall_value_internal): New.
4698 (sibcall_value_register): Remove.
4699 (sibcall_value_immediate): Remove.
4700 * config/nds32/predicates.md (nds32_general_register_operand): New.
4701 (nds32_call_address_operand): New.
4703 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4705 PR rtl-optimization/85167
4706 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4707 bb_defs if *split_p, instead preinitialize it to NULL.
4709 PR tree-optimization/85156
4710 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4711 evaluating the argument multiple times.
4713 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4715 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4717 (_mm_cvtpd_ps): Likewise.
4718 (_mm_cvttpd_epi32): Likewise.
4719 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4720 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4721 vector, pixel, and bool following altivec.h include.
4723 2018-04-03 Martin Sebor <msebor@redhat.com>
4725 * doc/extend.texi (Common Function Attributes): Clarify.
4726 (const attribute): Likewise.
4727 (pure attribute): Likewise.
4729 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4732 * config/i386/i386.c (ix86_expand_vector_set): Use
4733 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4735 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4737 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4738 instructions when changing rounding bits to preserve precision bits
4739 in the x87 control word.
4741 2018-04-03 Martin Liska <mliska@suse.cz>
4743 PR tree-optimization/82491
4744 * rtl.h (strip_offset_and_add): Replace += suboffset with
4745 poly_uint64 () + suboffset.
4747 2018-03-29 Martin Liska <mliska@suse.cz>
4748 Martin Jambor <mjambor@suse.cz>
4751 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4752 param_type is not an integral or pointer type.
4754 2018-04-03 Richard Biener <rguenther@suse.de>
4756 * sese.h (recompute_all_dominators): Remove.
4758 2018-04-02 Martin Sebor <msebor@redhat.com>
4760 * doc/invoke.texi (-Wrestrict): Fix typos.
4762 2018-04-02 Jim Wilson <jimw@sifive.com>
4764 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4765 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4766 (<optab>di3, <optab>si3_extend): Likewise.
4767 (<optab>si3_mask, <optab>si3_mask_1): New.
4768 (<optab>di3_mask, <optab>di3_mask_1): New.
4769 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4770 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4771 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4773 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4775 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4778 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4780 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4781 (nds32_canonicalize_comparison): New function.
4783 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4784 Kito Cheng <kito.cheng@gmail.com>
4785 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4787 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4788 * config/nds32/constants.md (unspec_volatile_element): Add
4789 UNSPEC_VOLATILE_RELAX_GROUP.
4790 * config/nds32/nds32-relax-opt.c: New file.
4791 * config/nds32/nds32-predicates.c
4792 (nds32_symbol_load_store_p): New function.
4793 * config/nds32/nds32-protos.h
4794 (nds32_symbol_load_store_p): Declare function.
4795 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4796 * config/nds32/nds32.c
4797 (nds32_register_pass): New function to register pass.
4798 (nds32_register_passes): New function to register passes.
4799 * config/nds32/nds32.md (relax_group): New pattern.
4800 * config/nds32/nds32.opt (mrelax-hint): New option.
4801 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4803 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4805 * config/nds32/t-nds32: Modify files dependency.
4807 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4809 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4810 (PROFILE_HOOK): Define its implementation.
4812 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4814 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4815 type and 32-bit size.
4817 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4820 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4821 (V_128_256): New mode iterator.
4822 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4823 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4824 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4826 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4827 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4829 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4832 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4833 NaN inputs correctly.
4835 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4838 * config/rs6000/vsx.md (??r): New mode attribute.
4839 (*vsx_mov<mode>_64bit): Use it.
4840 (*vsx_mov<mode>_32bit): Likewise.
4842 2018-03-30 Martin Sebor <msebor@redhat.com>
4844 PR tree-optimization/84818
4845 * builtins.c (check_access): Use warning_n.
4847 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4850 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4852 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4855 2018-03-30 Julia Koval <julia.koval@intel.com>
4858 * x86-tune.def (movx, partial_reg_dependency): Enable for
4861 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4864 * lra-constraints.c (process_alt_operands): Move setting
4865 this_alternative_matches below.
4867 2018-03-29 Martin Liska <mliska@suse.cz>
4870 * doc/invoke.texi: Document how LTO works with debug info.
4871 Describe auto-load support of binutils. Mention 'x86-64'
4872 as valid option value of -march option.
4874 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4876 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4879 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4880 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4881 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4884 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4887 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4888 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4889 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4890 for RS6000_BTM_POWERPC64.
4891 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4892 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4893 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4898 2018-03-28 Carl Love <cel@us.ibm.com>
4901 2017-09-27 Carl Love <cel@us.ibm.com>
4903 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4904 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4905 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4908 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4910 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4911 instead of __vector bool.
4912 (_mm_max_pu8): Likewise.
4913 (_mm_min_pi16): Likewise.
4915 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4918 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4919 (DIVWEUO): Likewise.
4921 (DIVDEUO): Likewise.
4922 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4923 DIVWEUO and DIVDEUO.
4924 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4925 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4926 (div_extend): Likewise.
4927 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4929 (__builtin_divweuo): Likewise.
4930 (__builtin_divdeo): Likewise.
4931 (__builtin_divdeuo): Likewise.
4933 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4936 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4937 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4939 PR tree-optimization/82004
4940 * gimple-match-head.c (optimize_pow_to_exp): New function.
4941 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4942 Don't fold to exp if optimize_pow_to_exp is false.
4944 2018-03-28 Martin Liska <mliska@suse.cz>
4947 * calls.c (initialize_argument_information): Fix trailing space.
4948 * common.opt: Fix typo and provide better explanation for
4949 -fsanitize-coverage option.
4950 * config/i386/i386.opt: Fix typo.
4952 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4953 Martin Liska <mliska@suse.cz>
4956 * gimplify.c (asan_poison_variable): Don't do the check for
4957 gimplify_omp_ctxp here.
4958 (gimplify_decl_expr): Do it here.
4959 (gimplify_target_expr): Likewise.
4961 2018-03-28 Martin Liska <mliska@suse.cz>
4964 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4965 chkp_type_bounds_count if MPX is not enabled.
4967 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4969 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4971 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4974 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4975 function to create the function decl for complex long double
4976 multiply and divide for -mabi=ieeelongdouble.
4977 (init_float128_ieee): Call it.
4979 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4982 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4983 -fcf-protection=branch -mibt.
4984 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4986 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4989 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4991 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4994 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4995 extern array declarations.
4997 2018-03-27 Richard Biener <rguenther@suse.de>
5000 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5001 explicit single_use checks.
5003 2018-03-27 Richard Biener <rguenther@suse.de>
5005 PR tree-optimization/85082
5006 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5009 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5011 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5012 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5013 Turn on fasynchronous-unwind-tables and funwind-tables.
5015 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5018 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5019 (*bmi_blsr_<mode>_ccz): Ditto.
5021 2018-03-26 Tom de Vries <tom@codesourcery.com>
5023 PR tree-optimization/85063
5024 * omp-general.c (offloading_function_p): New function. Factor out
5026 * omp-offload.c (pass_omp_target_link::gate): ... here.
5027 * omp-general.h (offloading_function_p): Declare.
5028 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5029 with attribute omp declare target for offloading functions.
5031 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5033 PR tree-optimization/84005
5034 * tree-data-ref.h (get_base_for_alignment): Declare.
5035 * tree-data-ref.c (get_base_for_alignment_1): New function.
5036 (get_base_for_alignment): Likewise.
5037 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5038 get_base_for_alignment to find a suitable base object, instead
5039 of always using drb->base_address.
5041 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5044 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5045 known size by default.
5047 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5050 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5051 and non BLKmode operands.
5053 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5056 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5057 Clean up attributes.
5059 2018-03-23 Richard Biener <rguenther@suse.de>
5062 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5063 we are going to emit early debug for LTO.
5065 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5068 * function.c (match_asm_constraints_1): Don't optimize if input
5069 doesn't satisfy general_operand predicate for output's mode.
5072 * alias.c (write_dependence_p): Don't require for x_canonicalized
5073 non-VOIDmode if x has VOIDmode.
5076 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5077 just don't try to optimize it rather than assert it never happens.
5079 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5081 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5082 macro expansions for definition of ST_INTERNAL_<mode> and
5083 LD_INTERNAL_<mode> builtins.
5084 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5086 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5088 (altivec_expand_st_builtin): Likewise.
5089 (altivec_expand_builtin): Remove calls to deleted functions.
5090 (rs6000_address_for_altivec): Delete this function.
5091 * config/rs6000/vector.md: Remove expands for
5092 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5094 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5097 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5098 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5099 re-computing once computed.
5100 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5101 (arm_init_machine_status): Initialize
5102 machine->static_chain_stack_bytes.
5104 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5107 * doc/extend.texi: Add four new prototypes for vec_ld.
5108 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5109 definitions for more logical presentation.
5110 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5111 entries for V1TI variants of __builtin_altivec_ld builtin.
5112 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5113 handling of V1TI variant of LVX icode pattern.
5114 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5115 (rs6000_gimple_fold_builtin): Likewise.
5116 (altivec_init_builtins): Add code to define
5117 __builtin_altivec_lvx_v1ti function.
5119 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5122 * function.c (match_asm_constraints_1): Don't do the optimization
5123 if input isn't a REG, SUBREG, MEM or constant.
5125 2018-03-22 Tom de Vries <tom@codesourcery.com>
5127 PR tree-optimization/84956
5128 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5129 bb_has_abnormal_pred.
5131 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5134 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5135 DECL_INITIAL (decl) to decl at the end.
5136 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5139 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5141 * doc/extend.texi (__builtin_tgmath): Document when complex
5142 integer types are treated as _Complex _Float64.
5144 2018-03-21 Tom de Vries <tom@codesourcery.com>
5146 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5148 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5150 PR tree-optimization/84960
5151 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5152 if it is ENTRY block, move them into single succ of ENTRY in that case.
5154 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5156 PR tree-optimization/84811
5157 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5158 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5159 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5160 (known_subrange_p): Update accordingly. Cast each value involved
5161 in the size comparison, rather than casting the result of the
5164 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5166 PR tree-optimization/84982
5167 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5168 by flipping the least significant bit rather than all bits from
5169 bitpos to bitpos + bitsize - 1.
5171 2018-03-21 Nathan Sidwell <nathan@acm.org>
5173 * doc/extend.texi (Deprecated Features): Remove mention of
5174 long-deleted deprecations.
5176 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5179 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5180 * configure: Regenerate.
5182 2018-03-21 Tom de Vries <tom@codesourcery.com>
5184 PR tree-optimization/83126
5185 * tree-parloops.c (num_phis): New function.
5186 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5188 2018-03-21 Nathan Sidwell <nathan@acm.org>
5190 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5191 mention anon-struct/union members and trailing attributes.
5193 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5195 PR tree-optimization/84969
5196 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5197 builtin memset partitions if they set different rhs values.
5199 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5201 PR rtl-optimization/84989
5202 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5203 VEC_DUPLICATE with scalar result mode.
5205 2018-03-21 Martin Liska <mliska@suse.cz>
5208 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5209 not intended return statement.
5211 2018-03-21 Martin Liska <mliska@suse.cz>
5214 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5215 (chkp_find_bound_slots_1): Limit number of iterations.
5217 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5220 * Minor grammar fixes for x86 options.
5222 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5225 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5226 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5228 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5231 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5232 (altivec_lvx_<mode>_1op): Likewise.
5233 (altivec_stvx_<mode>_2op): Likewise.
5234 (altivec_stvx_<mode>_1op): Likewise.
5235 (altivec_lvx_<VM2:mode>): New define_expand.
5236 (altivec_stvx_<VM2:mode>): Likewise.
5237 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5238 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5239 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5240 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5241 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5242 (rs6000_gen_lvx): Likewise.
5243 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5244 (altivec_expand_stv_builtin): Likewise.
5245 (altivec_expand_builtin): Likewise.
5246 * config/rs6000/vector.md: Likewise.
5248 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5251 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5254 2018-03-20 Richard Biener <rguenther@suse.de>
5257 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5258 sign-conversions as zero, fall back to standard scalar_stmt
5261 2018-03-20 Martin Liska <mliska@suse.cz>
5264 * predict.c (rebuild_frequencies): Handle case when we have
5265 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5267 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5270 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5271 flag_section_anchors.
5272 * varasm.c (use_blocks_for_decl_p): Remove hack for
5273 dw2_force_const_mem.
5276 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5278 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5279 be created, use lowpart_subreg of operands[0] rather than operands[0]
5281 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5282 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5283 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5284 and n constraint instead of aarch64_shift_imm_di and Usd.
5285 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5286 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5288 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5291 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5292 to favor GPR over NEON registers.
5293 (<shift>di3_neon): Likewise.
5295 2018-03-20 Tom de Vries <tom@codesourcery.com>
5298 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5299 (nvptx_process_pars): Emit bar.sync asap and alap.
5301 2018-03-20 Tom de Vries <tom@codesourcery.com>
5304 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5305 seen_label if seen_label is already set.
5307 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5310 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5311 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5312 Use 1U instead of 1. Formatting fixes.
5315 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5316 instead of TREE_TYPE (s1) for the return value.
5318 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5320 PR tree-optimization/84946
5321 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5322 bitsize + bitsize in poly_uint64 rather than poly_int64.
5325 * dwarf2asm.c: Include fold-const.c.
5326 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5327 of decl rather than decl itself.
5329 PR rtl-optimization/84643
5330 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5332 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5335 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5336 calling assemble_variable.
5338 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5341 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5342 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5344 2018-03-19 Jim Wilson <jimw@sifive.com>
5347 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5348 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5349 (riscv_first_stack_step): Likewise.
5350 (riscv_option_override): Use STACK_BOUNDARY instead of
5352 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5354 (BIGGEST_ALIGNMENT): Set to 128.
5355 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5356 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5359 2018-03-19 Richard Biener <rguenther@suse.de>
5361 PR tree-optimization/84933
5362 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5363 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5365 2018-03-19 Richard Biener <rguenther@suse.de>
5367 PR tree-optimization/84859
5368 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5369 (cond_if_else_store_replacement): Perform sinking operation on
5370 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5371 Generalize what a BB with a single eligible store is.
5373 2018-03-19 Richard Biener <rguenther@suse.de>
5375 PR tree-optimization/84929
5376 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5377 chrec_is_positive against non-chrec arg.
5379 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5382 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5384 2018-03-18 Martin Liska <mliska@suse.cz>
5386 PR rtl-optimization/84635
5387 * regrename.c (build_def_use): Use matches_mode only when
5390 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5392 PR tree-optimization/84913
5393 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5394 vectorize chains of COND_EXPRs.
5396 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5398 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5400 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5402 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5404 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5406 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5408 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5409 Kito Cheng <kito.cheng@gmail.com>
5411 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5412 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5413 (nds32_adjust_reg_alloc_order): New function.
5414 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5416 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5418 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5419 nds32_print_operand, nds32_print_operand_address): Use
5420 HOST_WIDE_INT_PRINT_DEC instead.
5422 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5424 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5426 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5429 * config/i386/i386.c (initial_ix86_tune_features,
5430 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5432 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5433 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5434 rather than 1u << ix86_tune. Formatting fix.
5435 (ix86_option_override_internal): Change ix86_arch_mask from
5436 unsigned int to unsigned HOST_WIDE_INT, initialize to
5437 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5438 (ix86_function_specific_restore): Likewise.
5440 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5443 * postreload.c (reload_combine_recognize_pattern): Perform
5444 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5445 truncate_int_for_mode the result for the destination's mode.
5448 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5449 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5451 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5454 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5457 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5458 regno_allocno_class_array and sorted_pseudos.
5459 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5460 insns where regno is used.
5462 2018-03-16 Martin Liska <mliska@suse.cz>
5465 * multiple_target.c (create_dispatcher_calls): Redirect
5466 reference in the symbol table.
5468 2018-03-16 Martin Liska <mliska@suse.cz>
5471 * multiple_target.c (create_dispatcher_calls): Redirect also
5474 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5478 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5479 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5480 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5482 2018-03-16 Julia Koval <julia.koval@intel.com>
5484 * doc/invoke.texi (Skylake Server): Add CLWB.
5485 Cannonlake): Remove CLWB.
5487 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5489 PR tree-optimization/84841
5490 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5492 (FLOAT_ONE_CONST_TYPE): Define.
5493 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5494 (sort_by_operand_rank): Put entries with higher constant_type last
5495 rather than first to match comments.
5497 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5499 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5502 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5505 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5506 check and use address of target always.
5508 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5511 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5512 (indirect_thunk_bnd_needed): Likewise.
5513 (indirect_thunks_used): Likewise.
5514 (indirect_thunks_bnd_used): Likewise.
5515 (indirect_return_needed): New.
5516 (indirect_return_bnd_needed): Likewise.
5517 (output_indirect_thunk_function): Add a bool argument for
5519 (output_indirect_thunk_function): Don't generate alias for
5520 function return thunk.
5521 (ix86_code_end): Call output_indirect_thunk_function to generate
5522 function return thunks.
5523 (ix86_output_function_return): Set indirect_return_bnd_needed
5524 and indirect_return_needed instead of indirect_thunk_bnd_needed
5525 and indirect_thunk_needed.
5527 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5529 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5530 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5531 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5533 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5534 Paul Hua <paul.hua.gm@gmail.com>
5537 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5539 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5541 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5542 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5545 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5548 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5549 instead of GET_MODE_SIZE when comparing Units.
5551 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5554 * varasm.c (hash_section): Return an unchangeble hash value
5555 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5556 Return !aarch64_can_use_per_function_literal_pools_p ().
5558 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5561 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5562 rather than address of cmode as last argument to prepare_cmp_insn.
5564 2018-03-15 Julia Koval <julia.koval@intel.com>
5566 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5567 F_AVX512VNNI, F_AVX512BITALG): New.
5569 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5572 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5573 insn for floating-point loads and stores.
5575 2018-03-14 Carl Love <cel@us.ibm.com>
5577 * config/rs6000/rs6000-c.c: Add macro definitions for
5578 ALTIVEC_BUILTIN_VEC_PERMXOR.
5579 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5580 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5581 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5582 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5584 * config/doc/extend.texi: Add prototypes for vec_permxor.
5586 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5589 * diagnostic-show-locus.c (class layout_point): Convert m_line
5590 from int to linenum_type.
5591 (line_span::comparator): Use linenum "compare" function when
5592 comparing line numbers.
5593 (test_line_span): New function.
5594 (layout_range::contains_point): Convert param "row" from int to
5596 (layout_range::intersects_line_p): Likewise.
5597 (layout::will_show_line_p): Likewise.
5598 (layout::print_source_line): Likewise.
5599 (layout::should_print_annotation_line_p): Likewise.
5600 (layout::print_annotation_line): Likewise.
5601 (layout::print_leading_fixits): Likewise.
5602 (layout::annotation_line_showed_range_p): Likewise.
5603 (struct line_corrections): Likewise for field m_row.
5604 (line_corrections::line_corrections): Likewise for param "row".
5605 (layout::print_trailing_fixits): Likewise.
5606 (layout::get_state_at_point): Likewise.
5607 (layout::get_x_bound_for_row): Likewise.
5608 (layout::print_line): Likewise.
5609 (diagnostic_show_locus): Likewise for locals "last_line" and
5611 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5612 * input.c (selftest::test_linenum_comparisons): New function.
5613 (selftest::input_c_tests): Call it.
5614 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5615 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5616 * selftest.h (ASSERT_GT): New macro.
5617 (ASSERT_GT_AT): New macro.
5618 (ASSERT_LT): New macro.
5619 (ASSERT_LT_AT): New macro.
5621 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5623 PR rtl-optimization/84780
5624 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5626 2018-03-14 Martin Liska <mliska@suse.cz>
5628 * tree.c (record_node_allocation_statistics): Use
5629 get_stats_node_kind.
5630 (get_stats_node_kind): New function extracted from
5631 record_node_allocation_statistics.
5632 (free_node): Use get_stats_node_kind.
5634 2018-03-14 Richard Biener <rguenther@suse.de>
5636 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5637 that the value-set of ANTIC_IN doesn't grow.
5640 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5642 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5643 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5645 2018-03-14 Julia Koval <julia.koval@intel.com>
5647 * config.gcc (icelake-client, icelake-server): New.
5649 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5650 (initial_ix86_arch_features): Ditto.
5651 (PTA_SKYLAKE): Add SGX.
5652 (PTA_ICELAKE): Remove.
5653 (PTA_ICELAKE_CLIENT): New.
5654 (PTA_ICELAKE_SERVER): New.
5655 (ix86_option_override_internal): Split up icelake on icelake client and
5657 (get_builtin_code_for_version): Ditto.
5658 (fold_builtin_cpu): Ditto.
5659 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5660 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5661 * config/i386/i386.h (processor_type): Ditto.
5662 * doc/invoke.texi: Ditto.
5664 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5667 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5668 INTEGER_CST offset, add it together with bitpos / 8 and
5669 sign extend based on POINTER_SIZE.
5673 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5676 * config/i386/constraints.md (Yc): New register constraint.
5677 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5678 Use Yc constraint for alternative 2 of operand 0. Remove
5679 preferred_for_speed attribute.
5681 2018-03-14 Richard Biener <rguenther@suse.de>
5683 PR tree-optimization/84830
5684 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5685 with the old one to avoid oscillations.
5687 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5690 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5692 (assign_by_spills): Return a flag of reload assignment failure.
5693 Do not process the reload assignment failures. Do not spill other
5694 reload pseudos if they has the same reg class. Update n if
5696 (lra_assign): Add a return arg. Set up from the result of
5697 assign_by_spills call.
5698 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5699 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5700 usage_insns if it is not NULL.
5701 (spill_hard_reg_in_range): New function.
5702 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5703 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5704 function prototypes.
5705 (lra_assign): Change prototype.
5706 * lra.c (lra): Add code to deal with fails by splitting hard reg
5709 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5711 * config/riscv/riscv.opt (mrelax): New option.
5712 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5713 "norelax" when riscv_mrelax is disabled.
5714 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5716 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5719 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5720 reassociation for int modes.
5722 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5724 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5725 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5727 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5728 * config/aarch64/aarch64-sve.md
5729 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5730 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5731 (*extend<mode><Vwide>2): Rename to...
5732 (aarch64_sve_extend<mode><Vwide>2): ...this.
5733 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5734 renaming the old pattern to...
5735 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5737 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5738 define_expand, renaming the old pattern to...
5739 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5740 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5741 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5742 account when deciding which SVE instruction the optab should use.
5743 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5745 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5747 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5748 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5749 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5750 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5751 (tlsdesc_small_<mode>): Turn a define_expand and use
5752 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5753 (tlsdesc_small_advsimd_<mode>): ...this.
5754 (tlsdesc_small_sve_<mode>): New pattern.
5756 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5758 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5759 (UNSPEC_UMUL_HIGHPART): New constants.
5760 (MUL_HIGHPART): New int iteraor.
5761 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5762 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5764 (*<su>mul<mode>3_highpart): New define_insn.
5766 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5769 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5772 2018-03-13 Martin Liska <mliska@suse.cz>
5775 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5777 (sem_item_optimizer::~sem_item_optimizer): Release it.
5778 (sem_item_optimizer::merge_classes): Register variable aliases.
5779 (sem_item_optimizer::fixup_pt_set): New function.
5780 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5781 * ipa-icf.h: Declare new variables and functions.
5783 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5786 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5787 integer_pow2p@2 and test integer_pow2p in condition.
5788 (A < 0 ? C : 0): Similarly for @1.
5791 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5792 characters starting at p contain '\0' character, don't look beyond
5796 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5797 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5800 * reg-stack.c (change_stack): Change update_end var from int to
5801 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5802 also call set_block_for_insn on the newly added insns and rescan.
5805 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5806 on the last operand.
5809 * tree.c (stabilize_reference_1): Return save_expr (e) for
5810 STATEMENT_LIST even if it doesn't have side-effects.
5812 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5814 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5816 2018-03-12 Renlin Li <renlin.li@arm.com>
5818 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5819 aarch64_output_scalar_simd_mov_immediate.
5821 2018-03-12 Martin Sebor <msebor@redhat.com>
5823 PR tree-optimization/83456
5824 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5825 for perfectly overlapping calls to memcpy.
5826 (gimple_fold_builtin_memory_chk): Same.
5827 (gimple_fold_builtin_strcpy): Handle no-warning.
5828 (gimple_fold_builtin_stxcpy_chk): Same.
5829 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5831 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5833 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5834 parameter. Use it for SFmode.
5835 (rs6000_function_arg_advance_1): Adjust.
5836 (rs6000_function_arg): Adjust.
5837 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5839 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5841 PR rtl-optimization/84169
5842 PR rtl-optimization/84780
5843 * combine.c (can_combine_p): Check for a 2-insn combination whether
5844 the destination register is used between the two insns, too.
5846 2018-03-12 Richard Biener <rguenther@suse.de>
5848 PR tree-optimization/84803
5849 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5850 for refs DR analysis didn't process.
5852 2018-03-12 Richard Biener <rguenther@suse.de>
5854 PR tree-optimization/84777
5855 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5856 force-vectorize loops ignore whether we are optimizing for size.
5858 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5860 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5861 (TARGET_MD_ASM_ADJUST): Define.
5863 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5864 Kito Cheng <kito.cheng@gmail.com>
5865 Chung-Ju Wu <jasonwucj@gmail.com>
5867 * config/nds32/nds32.c (nds32_compute_stack_frame,
5868 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5869 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5870 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5871 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5872 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5873 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5874 * config/nds32/nds32.md (prologue, epilogue): Use macro
5875 NDS32_V3PUSH_AVAILABLE_P to do checking.
5877 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5880 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5881 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5882 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5883 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5884 addition of most attributes on !orig_type_die or the attribute not
5885 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5887 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5888 Chung-Ju Wu <jasonwucj@gmail.com>
5890 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5892 * config/nds32/nds32.opt (mvh): New option.
5894 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5895 Chung-Ju Wu <jasonwucj@gmail.com>
5897 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5899 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5900 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5903 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5904 Chung-Ju Wu <jasonwucj@gmail.com>
5906 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5908 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5909 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5911 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5912 Kito Cheng <kito.cheng@gmail.com>
5913 Chung-Ju Wu <jasonwucj@gmail.com>
5915 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5916 UNSPEC_FFMISM and UNSPEC_FLMISM.
5917 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5918 for ffb, ffmism and flmism.
5919 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5920 (unspec_ffmism): Ditto.
5921 (unspec_flmism): Ditto.
5922 (nds32_expand_builtin_impl): Check if string extension is available.
5923 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5924 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5926 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5929 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5932 * lra-assigns.c (assign_by_spills): Return a flag of reload
5933 assignment failure. Do not process the reload assignment
5934 failures. Do not spill other reload pseudos if they has the same
5936 (lra_assign): Add a return arg. Set up from the result of
5937 assign_by_spills call.
5938 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5939 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5940 usage_insns if it is not NULL.
5941 (spill_hard_reg_in_range): New function.
5942 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5943 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5944 function prototypes.
5945 (lra_assign): Change prototype.
5946 * lra.c (lra): Add code to deal with fails by splitting hard reg
5949 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5952 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5954 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5957 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5958 (dw_val_node): Add val_symbolic_view.
5959 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5960 (symview_upper_bound): New.
5961 (new_line_info_table): Initialize symviews_since_reset.
5962 (dwarf2out_source_line): Count symviews_since_reset and set
5963 symview_upper_bound.
5964 (dw_val_equal_p): Handle symview.
5965 (add_AT_symview): New.
5966 (print_dw_val): Handle symview.
5967 (attr_checksum, attr_checksum_ordered): Likewise.
5968 (same_dw_val_p, size_of_die): Likewise.
5969 (value_format, output_die): Likewise.
5970 (add_high_low_attributes): Use add_AT_symview for entry_view.
5971 (dwarf2out_finish): Reset symview_upper_bound, clear
5974 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5977 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5978 Add strict argument and use it.
5979 (rs6000_split_multireg_move): Update for new strict argument.
5980 (mem_operand_gpr): Disallow all non-offsettable addresses.
5981 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5983 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5986 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5987 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5988 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5991 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5992 decl, use remap_type if we want to use the type.
5994 2018-03-09 Martin Sebor <msebor@redhat.com>
5996 PR tree-optimization/84526
5997 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5999 (builtin_access::generic_overlap): Be prepared to handle non-array
6002 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6004 PR rtl-optimization/84682
6005 * lra-constraints.c (process_address_1): Check is_address flag
6006 for address constraints.
6007 (process_alt_operands): Likewise.
6008 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6009 preprocess_constraints.
6010 * recog.h (preprocess_constraints): Add oploc parameter.
6012 * recog.c (preprocess_constraints): Test address_operand for
6013 CT_ADDRESS constraints.
6015 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6018 * lra-assigns.c (assign_by_spills): Return a flag of reload
6019 assignment failure. Do not process the reload assignment
6020 failures. Do not spill other reload pseudos if they has the same
6022 (lra_assign): Add a return arg. Set up from the result of
6023 assign_by_spills call.
6024 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6025 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6026 usage_insns if it is not NULL.
6027 (spill_hard_reg_in_range): New function.
6028 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6029 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6030 function prototypes.
6031 (lra_assign): Change prototype.
6032 * lra.c (lra): Add code to deal with fails by splitting hard reg
6035 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6038 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6039 Accept complain bool parameter. Only emit errors if it is true.
6040 (arm_parse_cpu_option_name): Likewise.
6041 (arm_target_thumb_only): Adjust callers of the above.
6042 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6043 prototype to take a default true bool parameter.
6044 (arm_parse_arch_option_name): Likewise.
6046 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6047 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6051 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6052 * configure: Regenerate.
6053 * configure.ac ("linker --version-script option"): New.
6054 ("linker soname option"): New.
6056 2018-03-09 Richard Biener <rguenther@suse.de>
6058 PR tree-optimization/84775
6059 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6060 immediate uses of predicate stmts and mark them modified.
6063 PR tree-optimization/84178
6064 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6066 (version_loop_for_if_conversion): Delay update_ssa call.
6067 (tree_if_conversion): Delay update_ssa until after predicate
6070 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6073 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6074 when the function accesses prior frames.
6076 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6079 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6080 gen_llsym, otherwise call maybe_gen_llsym.
6083 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6084 has ',' character inside of it.
6086 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6089 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6090 as clobbering CC_REGNUM.
6092 2018-03-08 Richard Biener <rguenther@suse.de>
6095 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6096 (follow_copies_to_constant): Do not follow SSA names registered
6099 2018-03-08 Richard Biener <rguenther@suse.de>
6101 PR tree-optimization/84178
6102 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6104 (version_loop_for_if_conversion): Delay update_ssa call.
6105 (tree_if_conversion): Delay update_ssa until after predicate
6108 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6110 PR tree-optimization/84178
6111 * tree-if-conv.c (release_bb_predicate): Remove the
6112 the assertion that the stmts have NULL use_ops.
6113 Discard the statements, asserting that they haven't
6114 yet been added to a BB.
6116 2018-03-08 Richard Biener <rguenther@suse.de>
6118 PR tree-optimization/84746
6119 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6120 (phi_translate): Pass in destination ANTIC_OUT set.
6121 (phi_translate_1): Likewise. For a simplified result lookup
6122 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6123 (phi_translate_set): Adjust.
6124 (do_pre_regular_insertion): Likewise.
6125 (do_pre_partial_partial_insertion): Likewise.
6127 2018-03-08 Martin Liska <mliska@suse.cz>
6129 PR gcov-profile/84735
6130 * doc/gcov.texi: Document usage of profile files.
6131 * gcov-io.h: Document changes in the format.
6133 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6137 * dwarf2out.c (struct dw_line_info_table): Update comments for
6139 (FORCE_RESET_NEXT_VIEW): New.
6140 (FORCE_RESETTING_VIEW_P): New.
6141 (RESETTING_VIEW_P): Check for -1 too.
6142 (ZERO_VIEW_P): Likewise.
6143 (new_line_info_table): Force-reset next view.
6144 (dwarf2out_begin_function): Likewise.
6145 (dwarf2out_source_line): Simplify zero_view_p initialization.
6146 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6147 view directly. Omit view when omitting .loc at line 0.
6149 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6151 PR tree-optimization/84740
6152 * tree-switch-conversion.c (process_switch): Call build_constructors
6153 only if info.phi_count is non-zero.
6155 PR tree-optimization/84739
6156 * tree-tailcall.c (find_tail_calls): Check call arguments against
6157 DECL_ARGUMENTS (current_function_decl) rather than
6158 DECL_ARGUMENTS (func) when checking for tail recursion.
6160 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6162 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6163 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6164 Volker Reichelt's entry and add entries for people that perform
6165 GCC fuzzy testing and report numerous bugs.
6167 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6170 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6171 readonly data in sdata, if that is disabled.
6172 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6173 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6174 -mreadonly-in-sdata option.
6176 2018-03-07 Martin Sebor <msebor@redhat.com>
6178 PR tree-optimization/84468
6179 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6180 basic block when looking for nul assignment.
6182 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6185 * except.h (output_function_exception_table): Adjust prototype.
6186 * except.c (output_function_exception_table): Remove FNNAME parameter
6187 and add SECTION parameter. Ouput one part of the table at a time.
6188 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6189 the first part of the exception table and emit unwind directives.
6190 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6191 (i386_pe_seh_cold_init): Likewise.
6192 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6193 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6194 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6195 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6196 * config/i386/winnt.c: Include except.h.
6197 (struct seh_frame_state): Add reg_offset, after_prologue and
6198 in_cold_section fields.
6199 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6200 (i386_pe_seh_cold_init): New function.
6201 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6202 to seh->in_cold_section.
6203 (seh_emit_push): Record the offset of the push.
6204 (seh_emit_save): Record the offet of the save.
6205 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6206 Test seh->after_prologue to disregard the epilogue.
6207 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6208 (i386_pe_end_cold_function): New function.
6210 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6213 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6214 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6217 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6219 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6220 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6223 * multiple_target.c: Include tree-inline.h and intl.h.
6224 (expand_target_clones): Diagnose and fail if node->definition and
6225 !tree_versionable_function_p (node->decl).
6227 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6229 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6231 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6232 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6233 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6235 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6238 * combine.c (try_combine): Use reg_or_subregno instead of handling
6239 just paradoxical SUBREGs and REGs.
6241 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6243 * config/arc/arc.c (arc_finalize_pic): Remove function.
6244 (arc_must_save_register): We use single base PIC register, remove
6245 checks to save/restore the PIC register.
6246 (arc_expand_prologue): Likewise.
6247 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6249 (arc_verify_short): Likewise.
6250 (arc_attr_type): Likewise.
6251 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6252 (walk_stores): Likewise.
6253 (arc_address_cost): Make it static.
6254 (arc_verify_short): Likewise.
6255 (branch_dest): Likewise.
6256 (arc_attr_type): Likewise.
6257 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6258 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6259 (arc_final_prescan_insn): Remove inserting the nops due to
6260 hardware hazards. It is done in reorg step.
6261 (insn_length_variant_t): Remove.
6262 (insn_length_parameters_t): Likewise.
6263 (arc_insn_length_parameters): Likewise.
6264 (arc_get_insn_variants): Likewise.
6265 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6267 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6270 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6273 PR tree-optimization/84687
6274 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6276 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6278 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6280 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6281 Rename to ppc_speculation_barrier.
6282 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6283 __builtin_ppc_speculation_barrier.
6285 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6288 * combine.c (combine_simplify_rtx): Don't try to simplify if
6289 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6292 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6294 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6295 to 32 bytes when compiling for POWER9.
6297 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6300 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6301 regparm >= 3 with no arg reg available also for calls with
6302 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6305 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6307 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6309 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6312 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6314 2018-03-05 Richard Biener <rguenther@suse.de>
6316 PR tree-optimization/84486
6317 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6318 When inserting a __builtin_assume_aligned call set the LHS
6319 SSA name alignment info accordingly.
6321 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6323 PR tree-optimization/84114
6324 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6325 Avoid reassociation of FLOAT_MODE addition.
6327 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6329 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6330 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6331 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6332 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6333 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6334 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6335 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6337 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6338 __builtin_ia32_wbinvd): New builtins.
6339 (SPECIAL_ARGS2): New.
6340 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6341 (SPECIAL_ARGS2): New.
6342 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6343 (ix86_valid_target_attribute_inner_p): Ditto.
6344 (ix86_init_mmx_sse_builtins): Add special_args2.
6345 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6346 TARGET_WBNOINVD_P): New.
6347 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6348 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6349 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6350 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6351 * config/i386/pconfigintrin.h: New file.
6352 * config/i386/wbnoinvdintrin.h: Ditto.
6353 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6354 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6356 2018-03-05 Richard Biener <rguenther@suse.de>
6358 PR tree-optimization/84670
6359 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6361 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6362 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6363 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6364 doesn't grow if all successors (recursively) were visited at least
6367 2018-03-05 Richard Biener <rguenther@suse.de>
6369 PR tree-optimization/84650
6370 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6371 if executed in the loop pipeline.
6373 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6375 * doc/configfiles.texi (Configuration Files): Move info about
6376 conditionalizing $target-protos.h to...
6377 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6378 differs from $target-protos.h.
6380 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6381 Chung-Ju Wu <jasonwucj@gmail.com>
6383 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6384 * config/nds32/nds32-multiple.md (setmemsi): Define.
6385 * config/nds32/nds32-memory-manipulation.c
6386 (nds32_gen_dup_4_byte_to_word_value): New.
6387 (emit_setmem_word_loop): New.
6388 (emit_setmem_byte_loop): New.
6389 (nds32_expand_setmem_loop): New.
6390 (nds32_expand_setmem_loop_v3m): New.
6391 (nds32_expand_setmem_unroll): New.
6392 (nds32_expand_setmem): New.
6394 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6395 Chung-Ju Wu <jasonwucj@gmail.com>
6397 * config/nds32/nds32-memory-manipulation.c
6398 (nds32_emit_load_store): New.
6399 (nds32_emit_post_inc_load_store): New.
6400 (nds32_emit_mem_move): New.
6401 (nds32_emit_mem_move_block): New.
6402 (nds32_expand_movmemsi_loop_unknown_size): New.
6403 (nds32_expand_movmemsi_loop_known_size): New.
6404 (nds32_expand_movmemsi_loop): New.
6405 (nds32_expand_movmemsi_unroll): New.
6406 (nds32_expand_movmemqi): Rename ...
6407 (nds32_expand_movmemsi): ... to this.
6408 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6409 (movmemsi): ... to this.
6410 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6411 (nds32_expand_movmemsi): ... to this.
6413 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6414 Monk Chiang <sh.chiang04@gmail.com>
6415 Chung-Ju Wu <jasonwucj@gmail.com>
6417 * config/nds32/nds32-protos.h
6418 (nds32_expand_load_multiple): New arguments.
6419 (nds32_expand_store_multiple): Ditto.
6420 (nds32_valid_multiple_load_store): Rename ...
6421 (nds32_valid_multiple_load_store_p): ... to this.
6422 * config/nds32/nds32-memory-manipulation.c
6423 (nds32_expand_load_multiple): Refine implementation.
6424 (nds32_expand_store_multiple): Ditto.
6425 * config/nds32/nds32-multiple.md
6426 (load_multiple): Update nds32_expand_load_multiple interface.
6427 (store_multiple): Update nds32_expand_store_multiple interface.
6428 * config/nds32/nds32-predicates.c
6429 (nds32_valid_multiple_load_store): Rename ...
6430 (nds32_valid_multiple_load_store_p): ... to this and refine
6432 * config/nds32/predicates.md
6433 (nds32_load_multiple_and_update_address_operation): New predicate.
6434 (nds32_store_multiple_and_update_address_operation): New predicate.
6436 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6437 Chung-Ju Wu <jasonwucj@gmail.com>
6439 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6440 (combo): New attribute.
6441 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6443 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6445 * config/nds32/nds32.opt: Change -mcmodel= default value.
6447 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6448 Monk Chiang <sh.chiang04@gmail.com>
6449 Chung-Ju Wu <jasonwucj@gmail.com>
6451 * config/nds32/constants.md (unspec_element): New enum.
6452 * config/nds32/constraints.md (Umw): New constraint.
6453 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6454 * config/nds32/nds32-intrinsic.md: Likewise.
6455 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6456 (nds32_valid_smw_lwm_base_p): New.
6457 (nds32_output_smw_single_word): New.
6458 (nds32_output_lmw_single_word): New.
6459 (nds32_expand_unaligned_load): New.
6460 (nds32_expand_unaligned_store): New.
6461 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6462 (nds32_output_smw_single_word): Declare.
6463 (nds32_output_lmw_single_word): Declare.
6464 (nds32_expand_unaligned_load): Declare.
6465 (nds32_expand_unaligned_store): Declare.
6466 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6467 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6468 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6469 NDS32_BUILTIN_UASTORE_DW.
6470 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6473 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6474 Kito Cheng <kito.cheng@gmail.com>
6475 Chung-Ju Wu <jasonwucj@gmail.com>
6477 * config/nds32/nds32-intrinsic.c
6478 (nds32_expand_builtin_null_ftype_reg): Delete.
6479 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6480 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6481 (nds32_read_argument): New.
6482 (nds32_legitimize_target): Ditto.
6483 (nds32_legitimize_argument): Ditto.
6484 (nds32_check_constant_argument): Ditto.
6485 (nds32_expand_unop_builtin): Ditto.
6486 (nds32_expand_unopimm_builtin): Ditto.
6487 (nds32_expand_binop_builtin): Ditto.
6488 (nds32_builtin_decl_impl): Ditto.
6489 (builtin_description): Ditto.
6490 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6491 (nds32_init_builtins_impl): Ditto.
6492 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6493 (nds32_builtin_decl): New.
6494 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6495 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6497 2018-03-02 Jeff Law <law@redhat.com>
6499 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6500 (redundant_insn, fill_simple_delay_slots): Likewise.
6501 (fill_slots_from_thread): Likewise.
6502 * resource.c (mark_referenced_resources): Likewise.
6503 (mark_set_resources, find_dead_or_set_registers): Likewise.
6505 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6507 * substring-locations.h (format_warning_va): Formatting fix for
6509 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6511 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6512 * substring-locations.c: Include intl.h.
6513 (format_warning_va): Turned into small wrapper around
6514 format_warning_n_va, renamed to ...
6515 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6516 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6518 (format_warning_at_substring_n): New function.
6519 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6520 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6521 format_warning_at_substring with just a shorter name instead of
6522 const function pointer.
6523 (fmtwarn_n): New function.
6524 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6525 appropriate, get rid of all the fmtstr temporaries, move conditionals
6526 with G_() wrapped string literals directly into fmtwarn arguments,
6527 cast dir.len to (int), formatting fixes.
6529 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6531 * doc/invoke.texi: Remove "Cilk Plus" references.
6533 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6534 Richard Biener <rguenther@suse.de>
6537 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6538 for error or warning attributes if CALL_FROM_THUNK_P is set.
6541 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6544 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6545 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6548 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6549 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6551 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6552 instead of -1U in last predictors element's probability member.
6554 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6557 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6558 arguments if they are comparable.
6560 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6562 PR tree-optimization/84634
6563 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6564 masks and masked_loop_p with a single loop_masks, making sure it's
6565 null for bb vectorization.
6567 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6569 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6570 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6571 loop->force_vectorize to check whether there is no alias.
6573 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6576 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6578 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6580 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6581 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6582 instead of a loop around prev_real_insn.
6583 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6587 * config/i386/i386.c (ix86_print_operand): Use conditional
6588 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6591 2018-03-02 Richard Biener <rguenther@suse.de>
6593 PR tree-optimization/84427
6594 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6595 (bitmap_set_subtract_values): Rewrite to handle multiple
6598 (prune_clobbered_mems): Likewise.
6599 (phi_translate): Take edge instead of pred/phiblock.
6600 (phi_translate_1): Likewise.
6601 (phi_translate_set): Likewise. Insert all translated
6602 exprs for a value into the set, keeping possibly multiple
6603 expressions per value.
6604 (compute_antic_aux): Adjust for phi_translate changes.
6605 When intersecting union the expressions and prune those
6606 not in the final value set, keeping possibly multiple
6607 expressions per value. Do not use value-insertion
6608 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6609 all expressions. Add verification that the value-sets
6610 only shrink during iteration.
6611 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6612 (do_pre_regular_insertion): Likewise.
6613 (do_pre_partial_partial_insertion): Likewise.
6615 2018-03-02 Richard Biener <rguenther@suse.de>
6618 * config/darwin.c (saved_debug_info_level): New static global.
6619 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6620 (darwin_asm_lto_end): Restore debug info generation settings.
6622 2018-03-01 Martin Liska <mliska@suse.cz>
6625 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6628 2018-03-01 Richard Biener <rguenther@suse.de>
6631 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6632 type annotation with LTO when debug was disabled at compile-time.
6634 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6636 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6638 (mips_final_postscan_insn): Likewise.
6640 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6642 PR rtl-optimization/84528
6643 * alias.c (init_alias_target): Add commentary.
6644 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6645 a unique base value if the frame pointer is not eliminated
6646 to the stack pointer.
6648 2018-03-01 Tom de Vries <tom@codesourcery.com>
6650 PR rtl-optimization/83327
6651 * lra-int.h (hard_regs_spilled_into): Declare.
6652 * lra.c (hard_regs_spilled_into): Define.
6653 (init_reg_info): Init hard_regs_spilled_into.
6654 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6655 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6656 (process_bb_lives): Handle hard_regs_spilled_into.
6657 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6658 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6660 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6662 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6663 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6664 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6665 * config/rs6000/aix72.h: New file.
6667 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6669 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6670 instead of warning_at with conditional singular and plural messages
6674 * stor-layout.c (update_alignment_for_field): For
6675 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6676 && !DECL_PACKED (field), do the alignment update, just use
6677 only desired_align instead of MAX (type_align, desired_align)
6679 (place_field): Don't do known_align < desired_align handling
6680 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6681 is non-NULL, instead do it after rli->prev_field handling and
6682 only if not within a bitfield word. For DECL_PACKED (field)
6683 use type_align of BITS_PER_UNIT.
6685 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6687 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6688 superfluous parentheses and trailing spaces.
6690 2018-02-28 Richard Biener <rguenther@suse.de>
6692 PR tree-optimization/84584
6693 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6694 SCoPs with fake exit edge.
6696 2018-02-28 Martin Liska <mliska@suse.cz>
6699 * timevar.c (timer::print): Fix format to properly print 100%
6702 2018-02-28 Richard Biener <rguenther@suse.de>
6705 * genmatch.c (capture_info::walk_match): Do not mark
6706 captured expressions without operands as expr_p given
6707 they act more like predicates and should be subject to
6708 "lost tail" side-effect preserving.
6710 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6712 PR rtl-optimization/81611
6713 * auto-inc-dec.c (attempt_change): Move dead note from
6714 mem_insn if it's the next use of regno
6715 (find_address): Take address use of reg holding
6716 non-incremented value. Add parm to limit search to the named
6718 (merge_in_block): Attempt to use a mem insn that is the next
6719 use of the original regno.
6721 2018-02-27 Martin Sebor <msebor@redhat.com>
6724 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6725 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6727 2018-02-27 Martin Sebor <msebor@redhat.com>
6729 PR translation/84207
6730 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6731 n argument to unsigned HOST_WIDE_INT.
6732 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6733 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6734 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6735 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6737 2018-02-27 Richard Biener <rguenther@suse.de>
6739 PR tree-optimization/84512
6740 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6741 Do not use the estimate returned from record_stmt_cost for
6742 the scalar iteration cost but sum properly using add_stmt_cost.
6744 2018-02-27 Richard Biener <rguenther@suse.de>
6746 PR tree-optimization/84466
6747 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6748 Adjust last change to less strictly validate use operands.
6750 2018-02-27 Martin Liska <mliska@suse.cz>
6752 PR gcov-profile/84548
6753 * gcov.c (process_file): Allow partial overlap and consider it
6754 also as group functions.
6755 (output_lines): Properly calculate range of lines for a group.
6757 2018-02-27 Martin Liska <mliska@suse.cz>
6759 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6760 'ggc' suffixes. Change first column width.
6761 (timer::print): Fix formatting of the column.
6763 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6765 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6766 preserve inline entry blocks for the sake of debug inline
6767 entry point markers alone.
6768 (remove_unused_locals): Suggest in comments a better place to
6769 force the preservation of inline entry blocks that are
6770 otherwise unused, but do not preserve them.
6772 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6774 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6776 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6779 * config/i386/constraints.md (Bs): Replace
6780 ix86_indirect_branch_register with
6781 TARGET_INDIRECT_BRANCH_REGISTER.
6783 * config/i386/i386.md (indirect_jump): Likewise.
6784 (tablejump): Likewise.
6785 (*sibcall_memory): Likewise.
6786 (*sibcall_value_memory): Likewise.
6787 Peepholes of indirect call and jump via memory: Likewise.
6788 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6789 (*sibcall_value_GOT_32): Likewise.
6790 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6791 (GOT_memory_operand): Likewise.
6792 (call_insn_operand): Likewise.
6793 (sibcall_insn_operand): Likewise.
6794 (GOT32_symbol_operand): Likewise.
6795 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6797 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6799 PR rtl-optimization/83496
6800 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6801 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6802 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6803 redundant insn, if any.
6804 (relax_delay_slots): Likewise.
6805 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6807 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6809 PR tree-optimization/83965
6810 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6811 that grouped statements are part of a reduction chain. Return
6812 true if the statement is not marked as a reduction itself but
6814 (vect_recog_dot_prod_pattern): Don't check whether the statement
6815 is part of a group here.
6816 (vect_recog_sad_pattern): Likewise.
6817 (vect_recog_widen_sum_pattern): Likewise.
6819 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6822 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6824 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6827 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6829 (ix86_output_indirect_function_return): New prototype.
6830 (ix86_split_simple_return_pop_internal): Likewise.
6831 * config/i386/i386.c (indirect_return_via_cx): New.
6832 (indirect_return_via_cx_bnd): Likewise.
6833 (indirect_thunk_name): Handle return va CX_REG.
6834 (output_indirect_thunk_function): Create alias for
6835 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6836 (ix86_output_indirect_jmp): Remove the bool argument.
6837 (ix86_output_indirect_function_return): New function.
6838 (ix86_split_simple_return_pop_internal): Likewise.
6839 * config/i386/i386.md (*indirect_jump): Don't pass false
6840 to ix86_output_indirect_jmp.
6841 (*tablejump_1): Likewise.
6842 (simple_return_pop_internal): Change it to define_insn_and_split.
6843 Call ix86_split_simple_return_pop_internal to split it for
6845 (simple_return_indirect_internal): Call
6846 ix86_output_indirect_function_return instead of
6847 ix86_output_indirect_jmp.
6849 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6852 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6853 memset and value initialization afterwards.
6855 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6857 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6859 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6862 * common/config/aarch64/aarch64-common.c
6863 (aarch_option_optimization_table[]): Switch
6864 off fomit-frame-pointer
6866 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6867 Chung-Ju Wu <jasonwucj@gmail.com>
6869 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6871 (store_multiple): Ditto.
6873 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6875 * config.gcc: Add --with-cpu support for nds32 target.
6876 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6877 * config/nds32/nds32.opt: Add -mcpu= option.
6879 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6881 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6882 isel=yes): Warn for these deprecated options.
6884 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6886 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6887 ISA_2_5_MASKS_EMBEDDED.
6889 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6891 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6892 p->max as pointers rather than using iterative_hash_expr.
6894 2018-02-23 Carl Love <cel@us.ibm.com>
6896 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6897 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6899 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6900 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6901 P8V_BUILTIN_VEC_VUNSIGNED2.
6903 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6906 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6907 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6909 * lra-constraints.c (curr_insn_transform): Set up
6910 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6912 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6913 * lra-eliminations.c (spill_pseudos): Ditto.
6914 (process_insn_for_elimination): Ditto.
6915 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6916 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6917 LRA_NON_CLOBBERED_ALT.
6919 2018-02-22 Martin Sebor <msebor@redhat.com>
6921 PR tree-optimization/84480
6922 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6923 to maybe_diag_stxncpy_trunc. Call it.
6924 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6925 from gimple_fold_builtin_strcpy. Print inlining stack.
6926 (handle_builtin_stxncpy): Print inlining stack.
6927 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6929 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6932 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6933 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6934 and -fcheck-pointer-bounds are used together.
6935 (indirect_thunk_prefix): New enum.
6936 (indirect_thunk_need_prefix): New function.
6937 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6938 "_nt" instead of "_bnd" for NOTRACK prefix.
6939 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6940 (output_indirect_thunk_function): Likewise.
6942 (ix86_code_end): Update output_indirect_thunk_function calls.
6943 (ix86_output_indirect_branch_via_reg): Replace
6944 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6945 (ix86_output_indirect_branch_via_push): Likewise.
6946 (ix86_output_function_return): Likewise.
6947 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6948 incompatible with -fcf-protection=branch and
6949 -fcheck-pointer-bounds.
6951 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6954 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6955 Change gcc_assert call to output_operand_lossage.
6957 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6959 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6961 2018-02-22 DJ Delorie <dj@redhat.com>
6962 Sebastian Perta <sebastian.perta@renesas.com>
6963 Oleg Endo <olegendo@gcc.gnu.org>
6965 * config/rx/rx.c (rx_rtx_costs): New function.
6966 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6968 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6970 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6972 2018-02-22 Martin Liska <mliska@suse.cz>
6975 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6976 Add "native" as a possible value.
6978 2018-02-22 Martin Liska <mliska@suse.cz>
6981 * config/i386/i386.c (ix86_option_override_internal):
6982 Add "native" as a possible value for -march and -mtune.
6984 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6987 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6988 to all type variants.
6990 PR tree-optimization/84503
6991 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6992 width as info->bitpos + info->bitsize - start.
6993 (merged_store_group::merge_overlapping): Simplify width computation.
6994 (check_no_overlap): New function.
6995 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6996 start + width and last_order of the group, fail if check_no_overlap
6998 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6999 to group if check_no_overlap fails.
7001 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7003 * config/rs6000/altivec.md: Delete contraint arguments to
7004 define_expand, define_split, and define_peephole2, and in
7005 define_insn_and_split if always unused.
7006 * config/rs6000/darwin.md: Ditto.
7007 * config/rs6000/dfp.md: Ditto.
7008 * config/rs6000/rs6000.md: Ditto.
7009 * config/rs6000/sync.md: Ditto.
7010 * config/rs6000/vector.md: Ditto.
7011 * config/rs6000/vsx.md: Ditto.
7013 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7015 * config/rs6000/altivec.md: Write output control strings as braced
7016 blocks instead of double-quoted strings.
7017 * config/rs6000/darwin.md: Ditto.
7018 * config/rs6000/rs6000.md: Ditto.
7019 * config/rs6000/vector.md: Ditto.
7020 * config/rs6000/vsx.md: Ditto.
7022 2018-02-21 Jason Merrill <jason@redhat.com>
7024 PR c++/84314 - ICE with templates and fastcall attribute.
7025 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7027 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7029 * ipa-cp.c (determine_versionability): Fix comment typos.
7031 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7034 * ipa-cp.c (determine_versionability): Do not version functions caling
7037 2018-02-21 Martin Liska <mliska@suse.cz>
7040 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7041 Add "native" as a possible value.
7042 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7043 the macro when native cpu detection is available.
7045 2018-02-21 Martin Liska <mliska@suse.cz>
7048 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7049 Add "native" as a possible value.
7050 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7051 when native cpu detection is available.
7053 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7054 Martin Sebor <msebor@redhat.com>
7056 PR tree-optimization/84478
7057 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7059 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7060 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7061 support which is conservatively correct, for 2 only stay conservative
7062 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7063 argument to the 2 argument get_range_strlen, adjust 6 arg
7064 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7066 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7067 (gimple_fold_builtin_strlen): Pass true as last argument to
7070 2018-02-20 Martin Sebor <msebor@redhat.com>
7073 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7074 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7075 (builtin_memref::builtin_memref): Factor out parts into
7076 set_base_and_offset and call it.
7078 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7081 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7082 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7083 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7084 search at the associated MODE_INT.
7086 2018-02-20 Jeff Law <law@redhat.com>
7089 PR tree-optimization/81592
7091 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7092 for range data rather than using global data.
7093 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7094 range data rather than using global data.
7095 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7096 pass it to children as needed.
7097 (struct directive::fmtresult): Similarly.
7098 (struct directive::set_width): Similarly.
7099 (struct directive::set_precision): Similarly.
7100 (format_integer, format_directive, parse_directive): Similarly.
7101 (format_none): Accept unnamed vr_values parameter.
7102 (format_percent, format_floating, format_character): Similarly.
7103 (format_string, format_plain): Similarly.
7104 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7105 the EVRP range analyzer for range data rather than using global data.
7106 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7107 gimple-ssa-evrp-analyze.h
7108 (class sprintf_dom_walker): Add after_dom_children member function.
7109 Add evrp_range_analyzer member.
7110 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7111 range analyzer as needed.
7112 (sprintf_dom_walker::after_dom_children): New member function.
7113 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7115 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7116 (evrp_range_analyzer::pop_to_marker): Likewise.
7118 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7120 PR tree-optimization/84419
7121 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7122 with the required type if its current type is compatible but
7125 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7128 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7129 after vectorization.
7131 2018-02-20 Martin Liska <mliska@suse.cz>
7134 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7135 possible values if we don't have a hint.
7137 2018-02-20 Martin Liska <mliska@suse.cz>
7141 * final.c (shorten_branches): Build align_tab array with one
7143 * opts.c (finish_options): Add alignment option limit check.
7144 (MAX_CODE_ALIGN): Likewise.
7145 (MAX_CODE_ALIGN_VALUE): Likewise.
7146 * doc/invoke.texi: Document maximum allowed option value for
7147 all -falign-* options.
7149 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7152 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7153 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7154 * var-tracking.c (emit_note_insn_var_location): Remove all references
7155 to NOTE_INSN_CALL_ARG_LOCATION.
7156 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7157 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7158 Use copy_rtx_if_shared.
7159 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7160 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7161 (dwarf2out_var_location): Remove handling of
7162 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7164 * final.c (final_scan_insn): Remove all references to
7165 NOTE_INSN_CALL_ARG_LOCATION.
7166 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7167 before dumping final insns.
7168 * except.c (emit_note_eh_region_end): Remove all references to
7169 NOTE_INSN_CALL_ARG_LOCATION.
7170 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7171 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7172 * config/arc/arc.c (hwloop_optimize): Likewise.
7173 * config/arm/arm.c (create_fix_barrier): Likewise.
7174 * config/s390/s390.c (s390_chunkify_start): Likewise.
7175 * config/sh/sh.c (find_barrier): Likewise.
7176 * config/i386/i386.c (rest_of_insert_endbranch,
7177 ix86_seh_fixup_eh_fallthru): Likewise.
7178 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7179 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7180 * config/frv/frv.c (frv_function_prologue): Likewise.
7181 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7183 (note_outside_basic_block_p): Remove all references to
7184 NOTE_INSN_CALL_ARG_LOCATION.
7185 * gengtype.c (adjust_field_rtx_def): Likewise.
7186 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7188 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7189 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7192 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7195 PR tree-optimization/84452
7196 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7197 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7200 2018-02-19 Martin Liska <mliska@suse.cz>
7203 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7205 2018-02-19 Martin Liska <mliska@suse.cz>
7206 Richard Sandiford <richard.sandiford@linaro.org>
7208 PR tree-optimization/82491
7209 * gimple-fold.c (get_base_constructor): Make earlier bail out
7212 2018-02-19 Carl Love <cel@us.ibm.com>
7214 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7215 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7217 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7218 P8V_BUILTIN_VEC_NEG.
7220 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7222 * config/rl78/rl78.md (movdf): New define expand.
7224 2018-02-19 Martin Liska <mliska@suse.cz>
7227 * doc/invoke.texi: Fix typo.
7228 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7230 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7232 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7233 handle rs6000_single_float and rs6000_double_float specially for
7236 2018-02-16 Jeff Law <law@redhat.com>
7238 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7241 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7244 * ipa-inline.c (inline_small_functions): Fix a typo.
7246 2018-02-16 Nathan Sidwell <nathan@acm.org>
7248 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7250 2018-02-16 Carl Love <cel@us.ibm.com>
7252 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7253 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7254 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7255 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7256 expansion to P8V_BUILTIN_VEC_FLOAT2.
7258 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7260 PR rtl-optimization/70023
7261 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7262 src_regno into account.
7264 2018-02-16 Carl Love <cel@us.ibm.com>
7266 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7267 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7268 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7269 * config/rs6000/rs6000.c: Remove case statements for
7270 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7271 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7272 and P9V_BUILTIN_VEC_VINSERT4B.
7273 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7274 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7275 * config/rs6000/vsx.md:
7276 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7279 2018-02-16 Carl Love <cel@us.ibm.com>
7281 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7283 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7285 * config/rs6000/rs6000-c.c: Add the definitions for
7286 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7287 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7288 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7289 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7290 definition for insert4b and define insn *insert3b_internal.
7291 * doc/extend.texi: Add documentation for vec_extract4b.
7293 2018-02-16 Nathan Sidwell <nathan@acm.org>
7295 * doc/extend.texi (Backwards Compatibility): Mention friend
7296 injection. Note for-scope is deprecated.
7297 * doc/invoke.texi (-ffriend-injection): Deprecate.
7299 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7301 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7302 that moved to I2, also allow destinations that are a paradoxical
7303 subreg (instead of a normal reg).
7305 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7308 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7311 2018-02-16 Richard Biener <rguenther@suse.de>
7313 PR tree-optimization/84037
7314 PR tree-optimization/84016
7316 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7317 Adjust vec_construct for the fact we need additional higher latency
7318 128bit inserts for AVX256 and AVX512 vector builds.
7319 (ix86_add_stmt_cost): Scale vector construction cost for
7322 2018-02-16 Richard Biener <rguenther@suse.de>
7324 PR tree-optimization/84417
7325 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7326 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7327 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7329 2018-02-16 Martin Liska <mliska@suse.cz>
7332 * internal-fn.def (ASAN_CHECK): Set proper flags.
7333 (ASAN_MARK): Likewise.
7335 2018-02-16 Julia Koval <julia.koval@intel.com>
7337 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7338 from PTA_CANNONLAKE.
7340 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7343 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7344 Use ++iter rather than iter++ for std::list iterators.
7345 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7346 defer deleting them until all nodes in the forest are processed. Do
7347 free even leaf nodes. Change to_process into auto_vec.
7350 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7351 * vec.h (vec_default_construct): Use memset instead of placement new
7352 if BROKEN_VALUE_INITIALIZATION is defined.
7353 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7354 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7357 PR rtl-optimization/83723
7358 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7359 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7360 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7362 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7364 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7366 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7368 PR rtl-optimization/81443
7369 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7370 from inner REGs to paradoxical SUBREGs.
7372 2018-02-16 Richard Biener <rguenther@suse.de>
7374 PR tree-optimization/84399
7375 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7376 For operands we can analyze at their definition make sure we can
7377 analyze them at each use as well.
7379 2018-02-16 Richard Biener <rguenther@suse.de>
7381 PR tree-optimization/84190
7382 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7383 volatile accesses if the decl isn't volatile.
7385 2018-02-15 Jason Merrill <jason@redhat.com>
7387 PR c++/84314 - ICE with templates and fastcall attribute.
7388 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7389 TYPE_CANONICAL on an existing type.
7391 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7393 PR tree-optimization/84383
7394 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7395 dstoff nor call operand_equal_p if dstbase is NULL.
7397 PR tree-optimization/84334
7398 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7399 also a CONSTANT_CLASS_P, punt.
7401 2018-02-14 Jim Wilson <jimw@sifive.com>
7403 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7404 first SMALL_OPERAND check. New local min_second_step. Move assert
7405 to where locals are set. Add TARGET_RVC support.
7406 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7408 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7410 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7412 2018-02-14 Martin Sebor <msebor@redhat.com>
7414 PR tree-optimization/83698
7415 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7416 arrays constrain the offset range to their bounds.
7417 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7418 (builtin_access::overlap): Avoid setting the size of overlap if it's
7420 (maybe_diag_overlap): Also consider arrays when deciding what values
7421 of offsets to include in diagnostics.
7423 2018-02-14 Martin Sebor <msebor@redhat.com>
7426 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7427 that correspond to the kind of a declaration.
7429 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7432 * config/pa/pa.md: Load address of PIC label using the linkage table
7433 if the label is nonlocal.
7435 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7437 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7438 warning message if user requests -maltivec=be.
7439 * doc/invoke.texi: Document deprecation of -maltivec=be.
7441 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7444 * config/rs6000/rs6000-c.c: Update definitions for
7445 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7446 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7448 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7451 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7452 add _get_ssp intrinsics. Remove argument from
7453 __builtin_ia32_rdssp[d|q].
7454 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7455 * config/i386/i386-builtin.def: Remove argument from
7456 __builtin_ia32_rdssp[d|q].
7457 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7458 ix86_expand_special_args_builtin for _rdssp[d|q].
7459 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7460 Clear register before usage.
7461 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7462 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7464 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7466 PR tree-optimization/84357
7467 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7468 operand 1 of an ARRAY_REF too.
7470 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7473 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7474 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7476 (set_of_reg): New struct.
7477 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7478 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7479 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7481 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7482 Split into bitclr, bitset, bitinvert patterns if appropriate.
7483 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7484 use rx_fuse_in_memory_bitop.
7485 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7486 to named insn, correct maximum insn length.
7488 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7491 * machmode.def: Define a complex mode for PARTIAL_INT.
7492 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7494 * doc/rtl.texi: Document CSPImode.
7495 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7497 (msp430_hard_regno_nregs_with_padding): Likewise.
7499 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7502 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7504 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7506 PR rtl-optimization/84169
7507 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7508 we generated a parallel as new i3 and we split that to new i2 and i3
7509 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7510 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7511 those to i2, not i1. Partially rewrite this scan code.
7513 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7516 * stor-layout.c (place_field): For variable length fields, adjust
7517 offset_align afterwards not just based on the field's alignment,
7518 but also on the size.
7521 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7522 of exps and logs in the use_exp2 case.
7524 2018-02-13 Jeff Law <law@redhat.com>
7526 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7529 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7532 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7536 * common.opt (gas-loc-support, gas-locview-support): New.
7537 (ginline-points, ginternal-reset-location-views): New.
7538 * doc/invoke.texi: Document them. Use @itemx where intended.
7539 (gvariable-location-views): Adjust.
7540 * target.def (reset_location_view): New.
7541 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7542 (TARGET_RESET_LOCATION_VIEW): New.
7543 * doc/tm.texi: Rebuilt.
7544 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7545 (dwarf2out_default_as_locview_support): New.
7546 (output_asm_line_debug_info): Use option variables.
7547 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7548 (output_loc_list): Likewise.
7549 (add_high_low_attributes): Check option variables.
7550 Don't output entry view attribute in strict mode.
7551 (gen_inlined_subroutine_die): Check option variables.
7552 (dwarf2out_inline_entry): Likewise.
7553 (init_sections_and_labels): Likewise.
7554 (dwarf2out_early_finish): Likewise.
7555 (maybe_reset_location_view): New, from...
7556 (dwarf2out_var_location): ... here. Call it.
7557 * debug.h (dwarf2out_default_as_loc_support): Declare.
7558 (dwarf2out_default_as_locview_support): Declare.
7559 * hooks.c (hook_int_rtx_insn_0): New.
7560 * hooks.h (hook_int_rtx_insn_0): Declare.
7561 * toplev.c (process_options): Take -gas-loc-support and
7562 -gas-locview-support from dwarf2out. Enable
7563 -gvariable-location-views by default only with locview
7564 assembler support. Enable -ginternal-reset-location-views by
7565 default only if the target defines the corresponding hook.
7566 Enable -ginline-points by default if location views are
7567 enabled; force it disabled if statement frontiers are
7569 * tree-inline.c (expand_call_inline): Check option variables.
7570 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7572 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7574 PR tree-optimization/84321
7575 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7576 handling. Also check whether the anti-range contains any values
7577 that satisfy the mask; switch to a VR_RANGE if not.
7579 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7582 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7584 2018-02-13 Martin Jambor <mjambor@suse.cz>
7587 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7588 of call statements, also set location of a load to a temporary.
7590 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7592 * config/rl78/rl78.c (add_vector_labels): New function.
7593 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7594 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7595 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7596 which checks that no arguments are passed.
7597 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7598 * doc/extend.texi: Documentation for the new attribute.
7600 2018-02-13 Andreas Schwab <schwab@suse.de>
7602 * config/riscv/linux.h (CPP_SPEC): Define.
7604 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7607 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7608 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7609 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7610 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7611 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7612 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7613 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7614 temporarily for AES and PCLMUL builtins.
7616 PR tree-optimization/84339
7617 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7618 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7622 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7623 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7624 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7626 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7628 * omp-simd-clone.h: New file.
7629 * omp-simd-clone.c: Include omp-simd-clone.h.
7630 (expand_simd_clones): No longer static.
7631 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7632 cgraph.h and omp-simd-clone.h.
7633 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7634 (vect_recog_widen_shift_pattern): Formatting fix.
7635 (vect_pattern_recog_1): Don't check optab for calls.
7638 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7639 operands[2] into a REG before using gen_lowpart on it.
7641 2018-02-12 Jeff Law <law@redhat.com>
7644 * config/sh/sh.c (find_barrier): Consider a sibling call
7647 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7648 successfully back substituting a reg.
7650 2018-02-12 Richard Biener <rguenther@suse.de>
7652 PR tree-optimization/84037
7653 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7654 parameter, move visited init to caller.
7655 (vect_slp_analyze_operations): Separate cost from validity
7656 check, initialize visited once for all instances.
7657 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7659 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7661 (vect_model_promotion_demotion_cost): Likewise.
7662 (vectorizable_bswap): Guard cost modeling with !slp_node
7663 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7665 (vectorizable_call): Likewise.
7666 (vectorizable_conversion): Likewise.
7667 (vectorizable_assignment): Likewise.
7668 (vectorizable_shift): Likewise.
7669 (vectorizable_operation): Likewise.
7670 (vectorizable_store): Likewise.
7671 (vectorizable_load): Likewise.
7672 (vectorizable_condition): Likewise.
7673 (vectorizable_comparison): Likewise.
7675 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7678 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7679 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7680 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7682 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7685 * params.def (inline-min-speedup): Increase from 8 to 15.
7686 (max-inline-insns-auto): Decrease from 40 to 30.
7687 * ipa-split.c (consider_split): Add some buffer for function to
7688 be considered inlining candidate.
7689 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7692 2018-02-12 Richard Biener <rguenther@suse.de>
7694 PR tree-optimization/84037
7695 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7696 matched stmts if we cannot swap the non-matched ones.
7698 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7700 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7701 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7702 _mm_maskz_scalef_round_ss): New intrinsics.
7703 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7704 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7705 __builtin_ia32_scalefss_round): Remove.
7706 (__builtin_ia32_scalefsd_mask_round,
7707 __builtin_ia32_scalefss_mask_round): New intrinsics.
7708 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7709 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7710 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7711 "<round_constraint>")): Changed to ...
7712 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7713 "<round_scalar_constraint>")): ... this.
7714 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7715 %0, %1, %2<round_op3>}"): Changed to ...
7716 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7717 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7718 %2<round_scalar_mask_op3>}"): ... this.
7719 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7721 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7723 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7724 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7725 (_mm_maskz_sqrt_round_ss): New intrinsics.
7726 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7727 (__builtin_ia32_sqrtsd_mask_round)
7728 (__builtin_ia32_sqrtss_mask_round): New builtins.
7729 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7730 (__builtin_ia32_sqrtss_round): Remove.
7731 (__builtin_ia32_sqrtsd_mask_round)
7732 (__builtin_ia32_sqrtss_mask_round): New builtins.
7733 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7734 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7735 ((match_operand:VF_128 1 "vector_operand"
7736 "xBm,<round_constraint>")): Changed to ...
7737 ((match_operand:VF_128 1 "vector_operand"
7738 "xBm,<round_scalar_constraint>")): ... this.
7739 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7740 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7741 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7742 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7743 %<iptr>1<round_scalar_mask_op3>}): ... this.
7744 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7745 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7747 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7750 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7751 Cast vec_cmpeq result to correct type.
7752 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7753 Cast vec_cmpgt result to correct type.
7755 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7757 * final.c (final_scan_insn_1): Renamed from...
7758 (final_scan_insn): ... this. New wrapper, to recover
7759 seen from the outermost call in recursive ones.
7760 * config/sparc/sparc.c (output_return): Drop seen from call.
7761 (output_sibcall): Likewise.
7762 * config/visium/visium.c (output_branch): Likewise.
7764 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7766 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7769 2018-02-10 Alan Modra <amodra@gmail.com>
7772 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7773 Specify LR as an input.
7775 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7778 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7779 remove_member_access_dummy_vars): New functions.
7780 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7781 lower_omp_1, execute_lower_omp): Use them.
7783 PR rtl-optimization/84308
7784 * shrink-wrap.c (spread_components): Release todo vector.
7786 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7788 PR rtl-optimization/57193
7789 * ira-color.c (struct allocno_color_data): Add member
7790 conflict_allocno_hard_prefs.
7791 (update_conflict_allocno_hard_prefs): New.
7792 (bucket_allocno_compare_func): Add a preference based on
7793 conflict_allocno_hard_prefs.
7794 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7795 (color_allocnos): Remove a dead code. Initiate
7796 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7798 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7801 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7802 constraint from =wa to wa. Avoid a subreg on the output operand,
7803 instead use a pseudo and subreg it in a move.
7804 (p9_xxbrd_<mode>): Changed to ...
7805 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7806 (p9_xxbrd_v2df): New expander.
7807 (p9_xxbrw_<mode>): Changed to ...
7808 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7809 (p9_xxbrw_v4sf): New expander.
7811 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7813 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7815 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7818 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7819 multiply in 32-bit mode.
7820 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7821 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7824 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7826 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7827 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7828 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7829 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7831 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7834 * dwarf2out.c (is_trivial_indirect_ref): New function.
7835 (dwarf2out_late_global_decl): Do not generate a location
7836 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7837 and that are not defined in the current unit.
7839 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7841 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7842 instead of a libcall for UNORDERED.
7844 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7847 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7848 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7850 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7853 * config/s390/s390.c (s390_set_current_function): Invoke
7854 s390_indirect_branch_settings also if fndecl didn't change.
7856 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7858 * config/rs6000/rs6000.md (blockage): Set length to zero.
7860 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7862 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7864 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7867 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7868 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7872 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7873 PARALLEL incoming that failed vt_get_decl_and_offset check.
7876 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7877 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7879 (get_variable_section): For decls in named .bss* sections pass true as
7880 second argument to bss_initializer_p.
7882 2018-02-09 Marek Polacek <polacek@redhat.com>
7883 Jakub Jelinek <jakub@redhat.com>
7886 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7887 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7888 Sync some changes from cxx_fold_indirect_ref.
7890 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7892 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7894 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7895 (BLOCK_INLINE_ENTRY_LABEL): New.
7896 (dwarf2out_var_location): Disregard inline entry markers.
7897 (inline_entry_data): New struct.
7898 (inline_entry_data_hasher): New hashtable type.
7899 (inline_entry_data_hasher::hash): New.
7900 (inline_entry_data_hasher::equal): New.
7901 (inline_entry_data_table): New variable.
7902 (add_high_low_attributes): Add DW_AT_entry_pc and
7903 DW_AT_GNU_entry_view attributes if a pending entry is found
7904 in inline_entry_data_table. Add old entry_pc attribute only
7905 if debug nonbinding markers are disabled.
7906 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7907 markers are enabled.
7908 (block_within_block_p, dwarf2out_inline_entry): New.
7909 (dwarf2out_finish): Check that no entries remained in
7910 inline_entry_data_table.
7911 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7912 (final_scan_insn, notice_source_line): Likewise.
7913 (rest_of_clean_state): Skip inline entry markers.
7914 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7916 * gimple.c (gimple_build_debug_inline_entry): New.
7917 * gimple.h (enum gimple_debug_subcode): Add
7918 GIMPLE_DEBUG_INLINE_ENTRY.
7919 (gimple_build_debug_inline_entry): Declare.
7920 (gimple_debug_inline_entry_p): New.
7921 (gimple_debug_nonbind_marker_p): Adjust.
7922 * insn-notes.def (INLINE_ENTRY): New.
7923 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7924 inline entry marker notes.
7925 (print_insn): Likewise.
7926 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7927 (INSN_DEBUG_MARKER_KIND): Likewise.
7928 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7929 * tree-inline.c (expand_call_inline): Build and insert
7930 debug_inline_entry stmt.
7931 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7932 inline entry blocks early, if nonbind markers are enabled.
7933 (dump_scope_block): Dump fragment info.
7934 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7935 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7936 (gimple_build_debug_inline_entry): New.
7937 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7938 Enable/disable inline entry points too.
7939 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7940 (DEBUG_INSN): Describe inline entry markers.
7942 * common.opt (gvariable-location-views): New.
7943 (gvariable-location-views=incompat5): New.
7944 * config.in: Rebuilt.
7945 * configure: Rebuilt.
7946 * configure.ac: Test assembler for view support.
7947 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7948 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7949 * dwarf2out.c (var_loc_view): New typedef.
7950 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7951 (dwarf2out_locviews_in_attribute): New.
7952 (dwarf2out_locviews_in_loclist): New.
7953 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7954 (enum dw_line_info_opcode): Add LI_adv_address.
7955 (struct dw_line_info_table): Add view.
7956 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7957 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7958 (zero_view_p): New variable.
7959 (ZERO_VIEW_P): New macro.
7960 (output_asm_line_debug_info): New.
7961 (struct var_loc_node): Add view.
7962 (add_AT_view_list, AT_loc_list): New.
7963 (add_var_loc_to_decl): Add view param. Test it against last.
7964 (new_loc_list): Add view params. Record them.
7965 (AT_loc_list_ptr): Handle loc and view lists.
7966 (view_list_to_loc_list_val_node): New.
7967 (print_dw_val): Handle dw_val_class_view_list.
7968 (size_of_die): Likewise.
7969 (value_format): Likewise.
7970 (loc_list_has_views): New.
7971 (gen_llsym): Set vl_symbol too.
7972 (maybe_gen_llsym, skip_loc_list_entry): New.
7973 (dwarf2out_maybe_output_loclist_view_pair): New.
7974 (output_loc_list): Output view list or entries too.
7975 (output_view_list_offset): New.
7976 (output_die): Handle dw_val_class_view_list.
7977 (output_dwarf_version): New.
7978 (output_compilation_unit_header): Use it.
7979 (output_skeleton_debug_sections): Likewise.
7980 (output_rnglists, output_line_info): Likewise.
7981 (output_pubnames, output_aranges): Update version comments.
7982 (output_one_line_info_table): Output view numbers in asm comments.
7983 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7984 Call maybe_gen_llsym.
7985 (loc_list_from_tree_1): Adjust.
7986 (add_AT_location_description): Create view list attribute if
7987 needed, check it's absent otherwise.
7988 (convert_cfa_to_fb_loc_list): Adjust.
7989 (maybe_emit_file): Call output_asm_line_debug_info for test.
7990 (dwarf2out_var_location): Reset views as needed. Precompute
7991 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7992 attribute. Set view.
7993 (new_line_info_table): Reset next view.
7994 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7995 (dwarf2out_source_line): Likewise. Output view resets and labels to
7996 the assembler, or select appropriate line info opcodes.
7997 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7998 (optimize_string_length): Catch it. Adjust.
7999 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8000 dw_val_class_view_list, and remove it if no longer needed.
8001 (hash_loc_list): Hash view numbers.
8002 (loc_list_hasher::equal): Compare them.
8003 (optimize_location_lists): Check whether a view list symbol is
8004 needed, and whether the locview attribute is present, and
8005 whether they match. Remove the locview attribute if no longer
8007 (index_location_lists): Call skip_loc_list_entry for test.
8008 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8009 Use output_dwarf_version.
8010 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8011 (struct dw_val_node): Add val_view_list.
8012 * final.c (SEEN_NEXT_VIEW): New.
8013 (set_next_view_needed): New.
8014 (clear_next_view_needed): New.
8015 (maybe_output_next_view): New.
8016 (final_start_function): Rename to...
8017 (final_start_function_1): ... this. Take pointer to FIRST,
8018 add SEEN parameter. Emit param bindings in the initial view.
8019 (final_start_function): Reintroduce SEEN-less interface.
8020 (final): Rename to...
8021 (final_1): ... this. Take SEEN parameter. Output final pending
8022 next view at the end.
8023 (final): Reintroduce seen-less interface.
8024 (final_scan_insn): Output pending next view before switching
8025 sections or ending a block. Mark the next view as needed when
8026 outputting variable locations. Notify debug backend of section
8027 changes, and of location view changes.
8028 (rest_of_handle_final): Adjust.
8029 * toplev.c (process_options): Autodetect value for debug variable
8030 location views option. Warn on incompat5 without -gdwarf-5.
8031 * doc/invoke.texi (gvariable-location-views): New.
8032 (gvariable-location-views=incompat5): New.
8033 (gno-variable-location-views): New.
8035 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8037 PR tree-optimization/84136
8038 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8039 that the result of find_edge is non-NULL.
8041 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8044 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8045 storing integer register in SImode. Fix cost of 256 and 512
8046 byte aligned SSE register store.
8048 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8050 * config/i386/i386.c (ix86_multiplication_cost): Fix
8051 multiplication cost for TARGET_AVX512DQ.
8053 2018-02-08 Marek Polacek <polacek@redhat.com>
8055 PR tree-optimization/84238
8056 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8059 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8061 PR tree-optimization/84265
8062 * tree-vect-stmts.c (vectorizable_store): Don't treat
8063 VMAT_CONTIGUOUS accesses as grouped.
8064 (vectorizable_load): Likewise.
8066 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8068 PR tree-optimization/81635
8069 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8070 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8071 (test_round_for_mask): New functions.
8072 (wide_int_cc_tests): Call test_round_for_mask.
8073 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8074 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8075 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8076 range returned by get_range_info.
8078 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8081 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8082 * symtab.c: Include builtins.h
8083 (symtab_node::output_to_lto_symbol_table_p): Move here
8084 from lto-streamer-out.c:output_symbol_p.
8085 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8086 (output_symbol_p): Move all logic to symtab.c
8087 (produce_symtab): Update.
8089 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8091 * config/s390/s390-opts.h (enum indirect_branch): Define.
8092 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8093 (s390_indirect_branch_via_thunk)
8094 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8095 (enum s390_indirect_branch_type): Define.
8096 * config/s390/s390.c (struct s390_frame_layout, struct
8097 machine_function): Remove.
8098 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8099 (indirect_branch_table_label_no, indirect_branch_table_name):
8101 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8102 (enum s390_indirect_branch_option): Define.
8103 (s390_return_addr_from_memory): New function.
8104 (s390_handle_string_attribute): New function.
8105 (s390_attribute_table): Add new attribute handler.
8106 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8107 (s390_indirect_branch_via_thunk): New function.
8108 (s390_indirect_branch_via_inline_thunk): New function.
8109 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8110 sibling call optimization for non z10 compiles.
8111 (s390_emit_call): Force indirect branch target to be a single
8112 register. Add r1 clobber for non-z10 compiles.
8113 (s390_emit_epilogue): Emit return jump via return_use expander.
8114 (s390_reorg): Handle JUMP_INSNs as execute targets.
8115 (s390_option_override_internal): Perform validity checks for the
8116 new command line options.
8117 (s390_indirect_branch_attrvalue): New function.
8118 (s390_indirect_branch_settings): New function.
8119 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8120 (s390_output_indirect_thunk_function): New function.
8121 (s390_code_end): Implement target hook.
8122 (s390_case_values_threshold): Implement target hook.
8123 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8125 * config/s390/s390.h (struct s390_frame_layout)
8126 (struct machine_function): Move here from s390.c.
8127 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8128 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8129 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8130 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8131 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8132 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8133 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8134 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8135 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8136 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8137 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8138 (mnemonic attribute): Add values which aren't recognized
8140 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8141 pattern for branch conversion. Fix mnemonic attribute.
8142 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8143 indirect branch via thunk if requested.
8144 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8145 ("*indirect_jump"): Disable for branch conversion using out of
8147 ("indirect_jump_via_thunk<mode>_z10")
8148 ("indirect_jump_via_thunk<mode>")
8149 ("indirect_jump_via_inlinethunk<mode>_z10")
8150 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8151 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8152 ("casesi_jump_via_inlinethunk<mode>_z10")
8153 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8154 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8155 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8156 ("*indirect2_jump"): Disable for branch conversion.
8157 ("casesi_jump"): Turn into expander and expand patterns for branch
8159 ("return_use"): New expander.
8160 ("*return"): Emit return via thunk and rename it to ...
8161 ("*return<mode>"): ... this one.
8162 * config/s390/s390.opt: Add new options and and enum for the
8165 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8167 * lra-constraints.c (match_reload): Unconditionally use
8168 gen_lowpart_SUBREG, rather than selecting between that
8169 and equivalent gen_rtx_SUBREG code.
8171 2018-02-08 Richard Biener <rguenther@suse.de>
8173 PR tree-optimization/84233
8174 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8175 changed flag instead of boguously re-using phi_inserted.
8177 2018-02-08 Martin Jambor <mjambor@suse.cz>
8179 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8180 static local variables.
8182 2018-02-08 Richard Biener <rguenther@suse.de>
8184 PR tree-optimization/84278
8185 * tree-vect-stmts.c (vectorizable_store): When looking for
8186 smaller vector types to perform grouped strided loads/stores
8187 make sure the mode is supported by the target.
8188 (vectorizable_load): Likewise.
8190 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8192 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8193 Increase LDP/STP opportunities by adding adjacent callee-saves.
8195 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8197 PR rtl-optimization/84068
8198 PR rtl-optimization/83459
8199 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8201 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8203 PR tree-optimization/84224
8204 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8205 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8208 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8211 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8212 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8213 count, remove one USE.
8215 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8217 * doc/install.texi (Configuration): Document the
8218 --with-long-double-format={ibm,ieee} PowerPC configuration
8222 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8223 Convert from define_expand to be define_insn_and_split. Rework
8224 float/double/_Float128 conversions to QI/HI/SImode to work with
8225 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8226 conversions to QI/HImode types did a store and then a load to
8227 truncate the value. For conversions to VSX registers, don't split
8228 the insn, instead emit the code directly. Use the code iterator
8229 any_fix to combine signed and unsigned conversions.
8230 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8231 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8232 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8233 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8234 (fix_<mode>di2_hw): Likewise.
8235 (fixuns_<mode>di2_hw): Likewise.
8236 (fix_<mode>si2_hw): Likewise.
8237 (fixuns_<mode>si2_hw): Likewise.
8238 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8239 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8240 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8241 fix<uns>_trunc<SFDF:mode>si2_p8.
8242 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8244 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8245 (fix<uns>_<mode>_mem): Likewise.
8246 (fctiw<u>z_<mode>_mem): Likewise.
8247 (fix<uns>_<mode>_mem): Likewise.
8248 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8249 the register allocator from doing a direct move to the GPRs to do
8250 a store, and instead use the ISA 3.0 store byte/half-word from
8251 vector register instruction. For IEEE 128-bit floating point,
8252 also optimize stores of 32-bit ints.
8253 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8255 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8257 * genextract.c (push_pathstr_operand): New function to support
8259 (walk_rtx): Call push_pathstr_operand.
8260 (print_path): Support [a-zA-Z].
8262 2018-02-07 Richard Biener <rguenther@suse.de>
8264 PR tree-optimization/84037
8265 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8266 (cse_and_gimplify_to_preheader): Declare.
8267 (vect_get_place_in_interleaving_chain): Likewise.
8268 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8270 (_loop_vec_info::~_loop_vec_info): Delete it.
8271 (cse_and_gimplify_to_preheader): New function.
8272 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8273 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8274 (vectorizable_load): Likewise. For grouped stores always base
8275 the IV on the first element.
8276 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8277 condition before gimplifying.
8279 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8281 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8282 *DIV_EXPR and *MOD_EXPR.
8284 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8287 * config/i386/i386.c (ix86_option_override_internal): Mask out
8288 the CF_SET bit when checking -fcf-protection.
8290 2018-02-07 Tom de Vries <tom@codesourcery.com>
8293 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8296 2018-02-07 Richard Biener <rguenther@suse.de>
8298 PR tree-optimization/84204
8299 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8302 PR tree-optimization/84205
8303 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8304 special-case isl_ast_op_zdiv_r.
8306 PR tree-optimization/84223
8307 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8308 Only add conditions from within the region.
8309 (gather_bbs::after_dom_children): Adjust.
8311 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8314 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8315 * config/avr/avr.md: Only post-reload split REG-REG moves if
8316 either register is GENERAL_REG_P.
8318 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8320 PR tree-optimization/84235
8321 * tree-ssa-scopedtables.c
8322 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8323 if the subtraction is performed in floating point type where NaNs are
8324 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8325 build 1. Formatting fix.
8327 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8330 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8331 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8332 and skip it regardless of bb boundaries. Use CALL_P macro,
8333 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8334 unnecessarily, formatting fix.
8336 2018-02-06 Michael Collison <michael.collison@arm.com>
8338 * config/arm/thumb2.md:
8339 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8340 (*thumb_mov_notscc): Ditto.
8342 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8345 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8346 unsigned_fix, not "s".
8348 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8350 * configure.ac (gcc_fn_eh_frame_ro): New function.
8351 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8352 correct .eh_frame permissions.
8353 * configure: Regenerate.
8355 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8357 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8360 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8362 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8363 Display warning message for -mno-speculate-indirect-jumps.
8365 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8367 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8369 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8371 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8373 PR tree-optimization/84225
8374 * tree-eh.c (find_trapping_overflow): Only call
8375 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8377 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8380 * config/i386/i386.c: Reimplement the check of possible options
8381 -mibt/-mshstk conbination. Change error messages.
8382 * doc/invoke.texi: Fix a typo: remove extra '='.
8384 2018-02-06 Marek Polacek <polacek@redhat.com>
8386 PR tree-optimization/84228
8387 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8389 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8392 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8393 emitted arch directives.
8394 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8395 __ARM_FEATURE_COPROC before changing architectures.
8397 2018-02-06 Richard Biener <rguenther@suse.de>
8399 * config/i386/i386.c (print_reg): Fix typo.
8400 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8402 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8404 * configure: Regenerate.
8406 2018-02-05 Martin Sebor <msebor@redhat.com>
8408 PR tree-optimization/83369
8409 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8412 2018-02-05 Martin Liska <mliska@suse.cz>
8414 * doc/invoke.texi: Cherry-pick upstream r323995.
8416 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8418 * ira.c (ira_init_register_move_cost): Adjust comment.
8420 2018-02-05 Martin Liska <mliska@suse.cz>
8422 PR gcov-profile/84137
8423 * doc/gcov.texi: Fix typo in documentation.
8425 2018-02-05 Martin Liska <mliska@suse.cz>
8427 PR gcov-profile/83879
8428 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8429 using dlopen functionality.
8431 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8433 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8434 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8435 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8436 _mm_maskz_range_round_ss): New intrinsics.
8437 (__builtin_ia32_rangesd128_round)
8438 (__builtin_ia32_rangess128_round): Remove.
8439 (__builtin_ia32_rangesd128_mask_round,
8440 __builtin_ia32_rangess128_mask_round): New builtins.
8441 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8442 __builtin_ia32_rangess128_round): Remove.
8443 (__builtin_ia32_rangesd128_mask_round,
8444 __builtin_ia32_rangess128_mask_round): New builtins.
8445 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8446 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8447 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8448 "<round_saeonly_constraint>")): Changed to ...
8449 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8450 "<round_saeonly_scalar_constraint>")): ... this.
8451 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8452 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8453 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8454 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8455 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8457 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8459 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8461 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8462 Remove all values except native, 8540 and 8548.
8464 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8466 * config/i386/i386.c (ix86_output_function_return): Pass
8467 INVALID_REGNUM, instead of -1, as invalid register number to
8468 indirect_thunk_name and output_indirect_thunk.
8470 2018-02-02 Julia Koval <julia.koval@intel.com>
8472 * config.gcc: Add -march=icelake.
8473 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8474 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8475 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8476 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8477 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8478 (processor_target_table): Add icelake.
8479 (ix86_option_override_internal): Handle new PTAs.
8480 (get_builtin_code_for_version): Handle icelake.
8481 (M_INTEL_COREI7_ICELAKE): New.
8482 (fold_builtin_cpu): Handle icelake.
8483 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8484 * doc/invoke.texi: Add -march=icelake.
8486 2018-02-02 Julia Koval <julia.koval@intel.com>
8488 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8489 to wide_int_bitmask.
8490 * wide-int-bitmask.h: New.
8492 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8495 * config/i386/i386.md: Replace Pmode with word_mode in
8496 builtin_setjmp_setup and builtin_longjmp to support x32.
8498 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8502 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8504 (rs6000_supported_cpu_names): New static variable.
8505 (linux_cpu_translation_table): Likewise.
8506 (elf_platform) <cpu>: Define new static variable and use it.
8507 Translate kernel AT_PLATFORM name to canonical name if needed.
8508 Error if platform name is unknown.
8510 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8513 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8515 2018-02-01 Jeff Law <law@redhat.com>
8518 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8519 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8520 the scratch if RELEASE_VIA_POP is false.
8521 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8522 If we have to save a temporary register, decrement SIZE appropriately.
8523 Pass new arguments to release_scratch_register_on_entry.
8524 (ix86_adjust_stack_and_probe): Likewise.
8525 (ix86_emit_probe_stack_range): Pass new arguments to
8526 release_scratch_register_on_entry.
8528 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8530 PR rtl-optimization/84157
8531 * combine.c (change_zero_ext): Use REG_P predicate in
8532 front of HARD_REGISTER_P predicate.
8534 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8536 * config/avr/avr.c (avr_option_override): Move disabling of
8537 -fdelete-null-pointer-checks to...
8538 * common/config/avr/avr-common.c (avr_option_optimization_table):
8541 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8543 PR tree-optimization/81635
8544 * tree-data-ref.c (split_constant_offset_1): For types that
8545 wrap on overflow, try to use range info to prove that wrapping
8548 2018-02-01 Renlin Li <renlin.li@arm.com>
8551 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8553 (aarch64_register_move_cost): Likewise.
8554 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8556 (REG_CLASS_NAMES): Likewise.
8557 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8558 TAILCALL_ADDR_REGS. Remove IP registers.
8559 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8561 2018-02-01 Richard Biener <rguenther@suse.de>
8563 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8564 for specifying RPO order and allow NULL for that.
8565 * domwalk.c (dom_walker::dom_walker): Likewise.
8566 (dom_walker::walk): Handle NULL RPO order.
8567 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8569 (rewrite_update_dom_walker): Likewise.
8570 (mark_def_dom_walker): Likewise.
8572 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8574 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8575 (aarch64_maybe_expand_sve_subreg_move): Declare.
8576 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8577 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8579 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8580 that are semantically a reverse operation.
8581 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8582 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8583 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8585 (aarch64_can_change_mode_class): For big-endian, forbid changes
8586 between two SVE modes if they have different element sizes.
8588 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8590 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8591 the TImode handling for big-endian targets.
8593 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8595 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8596 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8598 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8599 Remove BSWAP handing for big-endian targets and use the form of
8600 LD1RQ appropariate for the mode.
8602 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8604 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8605 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8608 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8611 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8612 check for operands that need to go through aarch64_sve_reload_be.
8614 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8616 PR tree-optimization/81661
8617 PR tree-optimization/84117
8618 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8619 * tree-eh.c: Include gimplify.h.
8620 (find_trapping_overflow, replace_trapping_overflow,
8621 rewrite_to_non_trapping_overflow): New functions.
8622 * tree-vect-loop.c: Include tree-eh.h.
8623 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8624 * tree-data-ref.c: Include tree-eh.h.
8625 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8627 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8629 PR rtl-optimization/84123
8630 * combine.c (change_zero_ext): Check if hard register satisfies
8631 can_change_dest_mode before calling gen_lowpart_SUBREG.
8633 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8636 * ira.c (ira_init_register_move_cost): Remove assert.
8638 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8640 PR rtl-optimization/84071
8641 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8642 * doc/tm.texi: Regenerate.
8644 2018-01-31 Richard Biener <rguenther@suse.de>
8646 PR tree-optimization/84132
8647 * tree-data-ref.c (analyze_miv_subscript): Properly
8648 check whether evolution_function_is_affine_multivariate_p
8649 before calling gcd_of_steps_may_divide_p.
8651 2018-01-31 Julia Koval <julia.koval@intel.com>
8654 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8655 * config/i386/i386.md (rdpid_rex64) New.
8656 (rdpid): Make 32bit only.
8658 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8661 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8662 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8664 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8667 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8669 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8671 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8673 PR rtl-optimization/84071
8674 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8675 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8677 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8679 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8680 (arc_attribute_table): Add 'aux' attribute.
8681 (arc_in_small_data_p): Consider aux like variables.
8682 (arc_is_aux_reg_p): New function.
8683 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8684 (arc_get_aux_arg): New function.
8685 (prepare_move_operands): Handle aux-register access.
8686 (arc_handle_aux_attribute): New function.
8687 * doc/extend.texi (ARC Variable attributes): Add subsection.
8689 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8691 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8692 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8693 (arc_attribute_table): Add 'uncached' attribute.
8694 (arc_print_operand): Print '.di' flag for uncached memory
8696 (arc_in_small_data_p): Do not consider for small data the uncached
8698 (arc_is_uncached_mem_p): New function.
8699 * config/arc/predicates.md (compact_store_memory_operand): Check
8700 for uncached memory accesses.
8701 (nonvol_nonimm_operand): Likewise.
8702 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8704 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8707 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8708 falign-loops=): Add Optimization flag.
8710 2018-01-30 Jeff Law <law@redhat.com>
8713 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8714 INT_REGISTERS_SAVED. Check it prior to calling
8715 get_scratch_register_on_entry.
8716 (ix86_adjust_stack_and_probe): Similarly.
8717 (ix86_emit_probe_stack_range): Similarly.
8718 (ix86_expand_prologue): Corresponding changes.
8720 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8723 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8724 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8726 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8729 * lra-constraints.c (curr_insn_transform): Process AND in the
8732 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8734 PR rtl-optimization/83986
8735 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8736 dependence against last_pending_memory_flush in addition to
8739 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8741 PR tree-optimization/81611
8742 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8745 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8748 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8751 2018-01-30 Richard Biener <rguenther@suse.de>
8752 Jakub Jelinek <jakub@redhat.com>
8754 PR tree-optimization/84111
8755 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8756 inner loops added during recursion, as they don't have up-to-date
8759 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8762 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8763 (can_inline_edge_by_limits_p): ... here.
8764 (can_early_inline_edge_p, check_callers,
8765 update_caller_keys, update_callee_keys, recursive_inlining,
8766 add_new_edges_to_heap, speculation_useful_p,
8767 inline_small_functions,
8768 inline_small_functions, flatten_function,
8769 inline_to_all_callers_1): Update.
8771 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8773 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8776 2018-01-30 Richard Biener <rguenther@suse.de>
8778 PR tree-optimization/83008
8779 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8780 invariant and constant vector uses in stmts when they need
8783 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8786 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8787 * configure: Regenerate.
8789 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8791 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8793 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8794 Use gen_rtx_REG rather than gen_lowpart.
8796 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8798 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8799 rather than 0 when creating partial subregs.
8801 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8803 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8806 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8809 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8810 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8811 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8812 flags. This restores the settings used before the 2017-07-24.
8813 Turning off pre increment/decrement/modify allows IVOPTS to
8814 optimize DF/SF loops where the index is an int.
8816 2018-01-29 Richard Biener <rguenther@suse.de>
8817 Kelvin Nilsen <kelvin@gcc.gnu.org>
8820 * tree-vect-stmts.c (vectorizable_call): Don't call
8821 targetm.vectorize_builtin_md_vectorized_function if callee is
8824 2018-01-22 Carl Love <cel@us.ibm.com>
8826 * doc/extend.tex: Fix typo in second arg in
8827 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8829 2018-01-29 Richard Biener <rguenther@suse.de>
8831 PR tree-optimization/84086
8832 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8833 (flush_ssaname_freelist): When SSA names were released reset
8834 the SCEV hash table.
8836 2018-01-29 Richard Biener <rguenther@suse.de>
8838 PR tree-optimization/84057
8839 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8840 removed paths when removing edges.
8842 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8844 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8845 -mfunction-return=@var{choice}.
8847 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8850 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8852 (layout::print_source_line): Likewise.
8853 (test_get_line_width_without_trailing_whitespace): Add test cases.
8855 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8858 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8861 2018-01-26 Jim Wilson <jimw@sifive.com>
8863 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8865 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8868 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8870 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8871 and CMP + SUB-immediate -> SUBS.
8873 2018-01-26 Martin Sebor <msebor@redhat.com>
8875 PR tree-optimization/83896
8876 * tree-ssa-strlen.c (get_string_len): Rename...
8877 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8878 Avoid assuming length is constant.
8879 (handle_char_store): Use HOST_WIDE_INT for string length.
8881 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8884 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8885 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8887 2018-01-26 Richard Biener <rguenther@suse.de>
8889 PR rtl-optimization/84003
8890 * dse.c (record_store): Only record redundant stores when
8891 the earlier store aliases at least all accesses the later one does.
8893 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8895 PR rtl-optimization/83985
8896 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8897 REG_CFA_RESTORE insns.
8898 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8899 REG_CFA_RESTORE insns here.
8902 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8903 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8905 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8907 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8908 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8909 (arc_init): Likewise.
8910 (arc_override_options): Likewise.
8911 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8913 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8915 * config/arc/arc.h (TARGET_DBNZ): Define.
8916 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8917 properly set the tune attribute.
8918 (dbnz): Use TARGET_DBNZ guard.
8919 * config/arc/arc.opt (mtune): Add core3 option.
8921 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8923 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8924 recognize new pic like addresses.
8925 (arc_delegitimize_address): Clean up.
8927 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8929 * config/arc/arc-arches.def: Option mrf16 valid for all
8931 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8932 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8933 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8934 * config/arc/arc-tables.opt: Regenerate.
8935 * config/arc/arc.c (arc_conditional_register_usage): Handle
8936 reduced register file case.
8937 (arc_file_start): Set must have build attributes.
8938 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8940 * config/arc/arc.opt (mrf16): Add new option.
8941 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8942 * config/arc/genmultilib.awk: Handle new mrf16 option.
8943 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8944 * config/arc/t-multilib: Regenerate.
8945 * doc/invoke.texi (ARC Options): Document mrf16 option.
8947 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8949 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8950 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8951 (arc_attribute_table): Add 'secure_call' attribute.
8952 (arc_print_operand): Print secure call operand.
8953 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8955 (arc_is_secure_call_p): New function. * config/arc/arc.md
8956 (call_i): Add support for sjli instruction.
8957 (call_value_i): Likewise.
8958 * config/arc/constraints.md (Csc): New constraint.
8960 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8961 John Eric Martin <John.Martin@emmicro-us.com>
8963 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8964 * config/arc/arc.c (_arc_jli_section): New struct.
8965 (arc_jli_section): New type.
8966 (rc_jli_sections): New static variable.
8967 (arc_handle_jli_attribute): New function.
8968 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8969 (arc_file_end): New function.
8970 (TARGET_ASM_FILE_END): Define.
8971 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8972 (arc_add_jli_section): New function.
8973 (jli_call_scan): Likewise.
8974 (arc_reorg): Call jli_call_scan.
8975 (arc_output_addsi): Remove 'S' from printing asm operand.
8976 (arc_is_jli_call_p): New function.
8977 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8979 (movhi_insn): Likewise.
8980 (movsi_insn): Likewise.
8981 (movsi_set_cc_insn): Likewise.
8982 (loadqi_update): Likewise.
8983 (load_zeroextendqisi_update): Likewise.
8984 (load_signextendqisi_update): Likewise.
8985 (loadhi_update): Likewise.
8986 (load_zeroextendhisi_update): Likewise.
8987 (load_signextendhisi_update): Likewise.
8988 (loadsi_update): Likewise.
8989 (loadsf_update): Likewise.
8990 (movsicc_insn): Likewise.
8991 (bset_insn): Likewise.
8992 (bxor_insn): Likewise.
8993 (bclr_insn): Likewise.
8994 (bmsk_insn): Likewise.
8995 (bicsi3_insn): Likewise.
8996 (cmpsi_cc_c_insn): Likewise.
8997 (movsi_ne): Likewise.
8998 (movsi_cond_exec): Likewise.
8999 (clrsbsi2): Likewise.
9010 (call_i): Remove 'S' asm letter, add jli instruction.
9011 (call_value_i): Likewise.
9012 * config/arc/arc.op (mjli-always): New option.
9013 * config/arc/constraints.md (Cji): New constraint.
9014 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9016 (subsf3_fpx): Likewise.
9017 (mulsf3_fpx): Likewise.
9018 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9020 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9022 * doc/invoke.texi (ARC): Document mjli-always option.
9024 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9026 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9027 avoid addition with 0 and use incw and decw where possible.
9029 2018-01-26 Richard Biener <rguenther@suse.de>
9031 PR tree-optimization/81082
9032 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9033 association if it requires casting to unsigned.
9034 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9035 from fold_plusminus_mult_expr to catch important cases late when
9036 range info is available.
9038 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9040 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9041 * configure.ac (hidden_linkonce): New test.
9042 * configure: Regenerate.
9043 * config.in: Regenerate.
9045 2018-01-26 Julia Koval <julia.koval@intel.com>
9047 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9048 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9049 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9050 _mm_mask_bitshuffle_epi64_mask): Fix type.
9051 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9052 USI_FTYPE_V4DI_V4DI_USI): Remove.
9053 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9054 __builtin_ia32_vpshufbitqmb256_mask,
9055 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9056 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9057 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9059 2018-01-26 Alan Modra <amodra@gmail.com>
9062 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9063 UNSPEC_VBPERMQ. Sort other unspecs.
9065 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9067 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9069 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9072 * predict.c (drop_profile): Do not push/pop cfun; update also
9074 (handle_missing_profiles): Fix logic looking for zero profiles.
9076 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9079 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9080 on functions with #pragma omp declare simd or functions with simd
9082 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9083 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9084 Remove trailing \n from warning_at calls.
9086 2018-01-25 Tom de Vries <tom@codesourcery.com>
9089 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9090 for neutered workers.
9092 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9095 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9096 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9098 2018-01-24 Jeff Law <law@redhat.com>
9101 * i386.c (get_probe_interval): Move to earlier point.
9102 (ix86_compute_frame_layout): If -fstack-clash-protection and
9103 the frame is larger than the probe interval, then use pushes
9104 to save registers rather than reg->mem moves.
9105 (ix86_expand_prologue): Remove conditional for int_registers_saved
9108 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9111 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9112 min/max for never referenced object.
9114 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9117 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9119 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9120 attributes from DECL_ATTRIBUTES (decl) without affecting
9121 DECL_ATTRIBUTES (current_function_decl).
9122 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9123 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9125 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9127 PR tree-optimization/83979
9128 * fold-const.c (fold_comparison): Use constant_boolean_node
9129 instead of boolean_{true,false}_node.
9131 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9133 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9136 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9138 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9139 Simplify the clause that sets the length attribute.
9140 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9141 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9142 clause that sets the length attribute.
9143 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9145 2018-01-24 Tom de Vries <tom@codesourcery.com>
9148 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9149 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9150 Add strict parameter.
9151 (prevent_branch_around_nothing): Insert dummy insn between branch to
9152 label and label with no ptx insn inbetween.
9153 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9155 2018-01-24 Tom de Vries <tom@codesourcery.com>
9158 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9159 for neutered threads in warp.
9160 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9162 2018-01-24 Richard Biener <rguenther@suse.de>
9164 PR tree-optimization/83176
9165 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9168 2018-01-24 Richard Biener <rguenther@suse.de>
9170 PR tree-optimization/82819
9171 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9172 code generating pluses that are no-ops in the target precision.
9174 2018-01-24 Richard Biener <rguenther@suse.de>
9177 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9179 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9181 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9182 to merge probabilities.
9183 * predict.c (probably_never_executed): Also mark as cold functions
9184 with global 0 profile and guessed local profile.
9185 * profile-count.c (profile_probability::combine_with_count): New
9187 * profile-count.h (profile_probability::operator*,
9188 profile_probability::operator*=, profile_probability::operator/,
9189 profile_probability::operator/=): Reduce precision to adjusted
9190 and set value to guessed on contradictory divisions.
9191 (profile_probability::combine_with_freq): Remove.
9192 (profile_probability::combine_wiht_count): Declare.
9193 (profile_count::force_nonzero):: Set to adjusted.
9194 (profile_count::probability_in):: Set quality to adjusted.
9195 * tree-ssa-tail-merge.c (replace_block_by): Use
9198 2018-01-23 Andrew Waterman <andrew@sifive.com>
9199 Jim Wilson <jimw@sifive.com>
9201 * config/riscv/riscv.c (riscv_stack_boundary): New.
9202 (riscv_option_override): Set riscv_stack_boundary. Handle
9203 riscv_preferred_stack_boundary_arg.
9204 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9205 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9206 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9207 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9208 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9209 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9211 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9214 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9215 of struct ix86_frame.
9216 (ix86_expand_epilogue): Likewise. Add a local variable for
9217 the reg_save_offset field in struct ix86_frame.
9219 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9221 PR tree-optimization/82604
9222 * tree-loop-distribution.c (enum partition_kind): New enum item
9223 PKIND_PARTIAL_MEMSET.
9224 (partition_builtin_p): Support above new enum item.
9225 (generate_code_for_partition): Ditto.
9226 (compute_access_range): Differentiate cases that equality can be
9227 proven at all loops, the innermost loops or no loops.
9228 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9229 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9230 (finalize_partitions, distribute_loop): Don't fuse partition of
9231 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9232 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9235 2018-01-23 Martin Liska <mliska@suse.cz>
9237 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9238 order to ignore the predictor.
9239 (PRED_POLYMORPHIC_CALL): Likewise.
9240 (PRED_RECURSIVE_CALL): Likewise.
9242 2018-01-23 Martin Liska <mliska@suse.cz>
9244 * tree-profile.c (tree_profiling): Print function header to
9245 aware reader which function we are working on.
9246 * value-prof.c (gimple_find_values_to_profile): Do not print
9247 not interesting value histograms.
9249 2018-01-23 Martin Liska <mliska@suse.cz>
9251 * profile-count.h (enum profile_quality): Add
9252 profile_uninitialized as the first value. Do not number values
9253 as they are zero based.
9254 (profile_count::verify): Update sanity check.
9255 (profile_probability::verify): Likewise.
9257 2018-01-23 Nathan Sidwell <nathan@acm.org>
9259 * doc/invoke.texi (ffor-scope): Deprecate.
9261 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9263 PR tree-optimization/83510
9264 * domwalk.c (set_all_edges_as_executable): New function.
9265 (dom_walker::dom_walker): Convert bool param
9266 "skip_unreachable_blocks" to enum reachability. Move setup of
9267 edge flags to set_all_edges_as_executable and only do it when
9268 reachability is REACHABLE_BLOCKS.
9269 * domwalk.h (enum dom_walker::reachability): New enum.
9270 (dom_walker::dom_walker): Convert bool param
9271 "skip_unreachable_blocks" to enum reachability.
9272 (set_all_edges_as_executable): New decl.
9273 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9274 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9276 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9277 but converting true to REACHABLE_BLOCKS.
9278 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9280 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9281 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9282 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9284 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9285 if check_all_array_refs will be called.
9287 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9289 * tree.c (selftest::test_location_wrappers): Add more test
9292 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9294 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9295 (selftest::test_bit_in_range): Likewise.
9297 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9300 * doc/sourcebuild.texi (vect_float): Say that the selector
9301 only describes the situation when -funsafe-math-optimizations is on.
9302 (vect_float_strict): Document.
9304 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9306 PR tree-optimization/83965
9307 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9308 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9309 instead of checking only for a reduction.
9310 (vect_recog_widen_sum_pattern): Likewise.
9312 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9314 * predict.c (probably_never_executed): Only use precise profile info.
9315 (compute_function_frequency): Skip after inlining hack since we now
9316 have quality checking.
9318 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9320 * profile-count.h (profile_probability::very_unlikely,
9321 profile_probability::unlikely, profile_probability::even): Set
9322 precision to guessed.
9324 2018-01-23 Richard Biener <rguenther@suse.de>
9326 PR tree-optimization/83963
9327 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9328 Properly terminate dominator walk when crossing the exit edge not
9329 when visiting its source block.
9331 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9334 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9335 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9337 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9339 PR tree-optimization/83957
9340 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9341 semicolon after for body surrounded by braces.
9343 PR tree-optimization/83081
9344 * profile-count.h (profile_probability::split): New method.
9345 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9346 Use profile_probability::split.
9347 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9348 when splitting a single conditional jump into 2.
9350 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9352 PR tree-optimization/69452
9353 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9356 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9358 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9359 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9360 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9362 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9364 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9365 * config/rl78/rl78.md (movdi): New define_expand.
9366 * config/rl78/rl78.c (rl78_split_movdi): New function.
9368 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9371 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9373 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9374 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9375 128-bit to produce an UNSPEC move to get the double word with the
9376 signbit and then a shift directly to do signbit.
9377 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9378 implementation with a new version that just does either a direct
9379 move or a regular move. Move memory interface to separate insns.
9380 Move insns so they are next to the expander.
9381 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9382 with signbit move. Split big and little endian case.
9383 (signbit<mode>2_dm_mem_le): Likewise.
9384 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9385 (signbit<mode>2_dm2): Likewise.
9387 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9389 * config/rl78/rl78.md (anddi3): New define_expand.
9391 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9393 * config/rl78/rl78.md (umindi3): New define_expand.
9395 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9397 * config/rl78/rl78.md (smindi3): New define_expand.
9399 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9401 * config/rl78/rl78.md (smaxdi3): New define_expand.
9403 2018-01-22 Carl Love <cel@us.ibm.com>
9405 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9406 LVX_V1TI): Add macro expansion.
9407 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9408 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9409 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9410 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9411 Change check to determine if the instruction is a byte reversing
9412 entry. Fix typo in comment.
9413 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9414 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9415 Add def_builtin calls for new builtins.
9416 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9417 Add define_insn expansion.
9419 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9421 * config/rl78/rl78.md (umaxdi3): New define_expand.
9423 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9425 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9426 for non-QImode registers.
9428 2018-01-22 Richard Biener <rguenther@suse.de>
9430 PR tree-optimization/83963
9431 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9432 including the loop exit block.
9433 (scop_detection::merge_sese): Likewise.
9434 (scop_detection::add_scop): Do it here instead.
9436 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9438 * doc/sourcebuild.texi (arm_softfloat): Document.
9440 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9443 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9444 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9445 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9447 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9448 David Edelsohn <dje.gcc@gmail.com>
9451 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9452 Change "crset eq" to "crset 2".
9453 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9454 (*call_indirect_aix<mode>_nospec): Likewise.
9455 (*call_value_indirect_aix<mode>_nospec): Likewise.
9456 (*call_indirect_elfv2<mode>_nospec): Likewise.
9457 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9458 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9459 change assembly output from . to $.
9460 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9461 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9462 (*tablejump<mode>_internal1_nospec): Likewise.
9464 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9467 * config/sh/sh_optimize_sett_clrt.cc:
9468 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9470 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9472 PR tree-optimization/83940
9473 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9474 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9475 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9476 use it to pass back the definition type.
9477 (vect_check_store_rhs): Likewise rhs_dt_out.
9478 (vect_build_gather_load_calls): Add a mask_dt argument and use
9479 it instead of a call to vect_is_simple_use.
9480 (vectorizable_store): Update calls to vect_check_load_store_mask
9481 and vect_check_store_rhs. Use the dt returned by the latter instead
9482 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9483 instead of calls to vect_is_simple_use. Pass the scalar rather
9484 than the vector operand to vect_is_simple_use when handling
9485 second and subsequent copies of an rhs value.
9486 (vectorizable_load): Update calls to vect_check_load_store_mask
9487 and vect_build_gather_load_calls. Use the cached mask_dt and
9488 gs_info.offset_dt instead of calls to vect_is_simple_use.
9490 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9493 * tree-emutls.c: Include gimplify.h.
9494 (lower_emutls_2): New function.
9495 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9496 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9497 it before further processing.
9500 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9501 UINTVAL (trueop1) instead of INTVAL (op1).
9503 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9507 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9508 INCOMING_FRAME_SP_OFFSET if not defined.
9509 (scan_trace): Add ENTRY argument. If true and
9510 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9511 emit a note to adjust the CFA offset.
9512 (create_cfi_notes): Adjust scan_trace callers.
9513 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9514 INCOMING_FRAME_SP_OFFSET in the CIE.
9515 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9516 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9518 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9519 * doc/tm.texi: Regenerated.
9521 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9523 PR rtl-optimization/83147
9524 * lra-constraints.c (remove_inheritance_pseudos): Use
9525 lra_substitute_pseudo_within_insn.
9527 2018-01-19 Tom de Vries <tom@codesourcery.com>
9528 Cesar Philippidis <cesar@codesourcery.com>
9531 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9533 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9536 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9537 spaces for function labels.
9539 2018-01-19 Martin Liska <mliska@suse.cz>
9541 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9542 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9543 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9544 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9545 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9546 (PRED_CONST_RETURN): Change from 69 to 65.
9547 (PRED_NULL_RETURN): Change from 91 to 71.
9548 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9549 (PRED_LOOP_GUARD): Change from 66 to 73.
9551 2018-01-19 Martin Liska <mliska@suse.cz>
9553 * predict.c (predict_insn_def): Add new assert.
9554 (struct branch_predictor): Change type to signed integer.
9555 (test_prediction_value_range): Amend test to cover
9557 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9558 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9559 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9560 (PRED_LOOP_IV_COMPARE): Likewise.
9561 * predict.h (PROB_UNINITIALIZED): Define new constant.
9563 2018-01-19 Martin Liska <mliska@suse.cz>
9565 * predict.c (dump_prediction): Add new format for
9566 analyze_brprob.py script which is enabled with -details
9568 * profile-count.h (precise_p): New function.
9570 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9572 PR tree-optimization/83922
9573 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9574 there are no statements that need masking.
9575 (vect_active_double_reduction_p): New function.
9576 (vect_analyze_loop_operations): Use it when handling phis that
9577 are not in the loop header.
9579 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9581 PR tree-optimization/83914
9582 * tree-vect-loop.c (vectorizable_induction): Don't convert
9583 init_expr or apply the peeling adjustment for inductions
9584 that are nested within the vectorized loop.
9586 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9588 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9591 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9595 * function.h (gimplify_parameters): Add gimple_seq * argument.
9596 * function.c: Include gimple.h and options.h.
9597 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9598 for the added local temporaries if needed.
9599 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9600 if there are any parameter cleanups, wrap whole body into a
9601 try/finally with the cleanups.
9603 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9606 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9607 Use GET_MODE_CLASS for scalar floating point.
9609 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9613 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9614 Fix call of call_cgraph_insertion_hooks.
9616 2018-01-18 Martin Sebor <msebor@redhat.com>
9618 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9620 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9623 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9626 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9629 * common.opt: (-ffile-prefix-map): New option.
9630 * opts.c (common_handle_option): Defer it.
9631 * opts-global.c (handle_common_deferred_options): Handle it.
9632 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9633 * file-prefix-map.h: New file.
9634 (remap_debug_filename, add_debug_prefix_map): ...here.
9635 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9636 * final.c (debug_prefix_map, add_debug_prefix_map
9637 remap_debug_filename): Move to...
9638 * file-prefix-map.c: New file.
9639 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9640 generalize, get rid of alloca(), use strrchr() instead of strchr().
9641 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9642 Implement in terms of add_prefix_map().
9643 (remap_macro_filename, remap_debug_filename): Implement in term of
9645 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9646 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9647 * dbxout.c: Include file-prefix-map.h.
9648 * varasm.c: Likewise.
9649 * vmsdbgout.c: Likewise.
9650 * xcoffout.c: Likewise.
9651 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9652 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9653 * doc/invoke.texi (-ffile-prefix-map): Document.
9654 (-fdebug-prefix-map): Update description.
9656 2018-01-18 Martin Liska <mliska@suse.cz>
9658 * config/i386/i386.c (indirect_thunk_name): Document that also
9660 (output_indirect_thunk): Document why both instructions
9661 (pause and lfence) are generated.
9663 2018-01-18 Richard Biener <rguenther@suse.de>
9665 PR tree-optimization/83887
9666 * graphite-scop-detection.c
9667 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9668 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9669 (scop_detection::merge_sese): Re-implement with a flood-fill
9670 algorithm that properly finds a SESE region if it exists.
9672 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9675 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9676 pointer_diff optimizations use view_convert instead of convert.
9678 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9680 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9681 Generate different code for -mno-speculate-indirect-jumps.
9682 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9683 (*call_indirect_aix<mode>): Disable for
9684 -mno-speculate-indirect-jumps.
9685 (*call_indirect_aix<mode>_nospec): New define_insn.
9686 (*call_value_indirect_aix<mode>): Disable for
9687 -mno-speculate-indirect-jumps.
9688 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9689 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9690 -mno-speculate-indirect-jumps.
9691 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9693 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9695 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9696 long double type, set the flags for noting the default long double
9697 type, even if we don't pass or return a long double type.
9699 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9702 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9705 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9707 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9708 support for merge[hl].
9709 (fold_mergehl_helper): New helper function.
9710 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9711 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9712 (altivec_vmrglw_direct): Add xxmrglw insn.
9714 2018-01-17 Andrew Waterman <andrew@sifive.com>
9716 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9717 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9719 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9722 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9723 call the lto_location_cache before reading the
9724 DECL_SOURCE_LOCATION of the types.
9726 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9727 Richard Sandiford <richard.sandiford@linaro.org>
9729 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9730 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9731 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9732 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9733 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9735 * config/aarch64/constraints.md (aarch64_movti_operand):
9737 * config/aarch64/predicates.md (Uti): Add new constraint.
9739 2018-01-17 Carl Love <cel@us.ibm.com>
9741 * config/rs6000/vsx.md (define_expand xl_len_r,
9742 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9743 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9745 (define_expand, define_insn): Move the shift left from the
9746 define_insn to the define_expand for lxvl and stxvl instructions.
9747 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9748 and XL_LEN_R definitions to PURE.
9750 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9752 * config/i386/i386.c (indirect_thunk_name): Declare regno
9753 as unsigned int. Compare regno with INVALID_REGNUM.
9754 (output_indirect_thunk): Ditto.
9755 (output_indirect_thunk_function): Ditto.
9756 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9757 in the call to output_indirect_thunk_function.
9759 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9762 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9763 rather than the size of inner_type to determine the stack slot size
9764 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9766 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9769 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9772 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9774 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9775 endian Linux systems to optionally enable multilibs for selecting
9776 the long double type if the user configured an explicit type.
9777 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9778 have no long double multilibs if not defined.
9779 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9780 warn if the user used -mabi={ieee,ibm}longdouble and we built
9781 multilibs for long double.
9782 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9783 appropriate multilib option.
9784 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9786 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9787 for building long double multilibs.
9788 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9790 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9792 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9795 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9797 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9800 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9803 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9806 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9808 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9809 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9811 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9813 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9814 different rtl trees depending on TARGET_64BIT.
9815 (rs6000_gen_lvx): Likewise.
9817 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9819 * config/visium/visium.md (nop): Tweak comment.
9820 (hazard_nop): Likewise.
9822 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9824 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9825 -mspeculate-indirect-jumps.
9826 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9827 for -mno-speculate-indirect-jumps.
9828 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9829 (*call_value_indirect_elfv2<mode>): Disable for
9830 -mno-speculate-indirect-jumps.
9831 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9832 (indirect_jump): Emit different RTL for
9833 -mno-speculate-indirect-jumps.
9834 (*indirect_jump<mode>): Disable for
9835 -mno-speculate-indirect-jumps.
9836 (*indirect_jump<mode>_nospec): New define_insn.
9837 (tablejump): Emit different RTL for
9838 -mno-speculate-indirect-jumps.
9839 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9840 (tablejumpsi_nospec): New define_expand.
9841 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9842 (tablejumpdi_nospec): New define_expand.
9843 (*tablejump<mode>_internal1): Disable for
9844 -mno-speculate-indirect-jumps.
9845 (*tablejump<mode>_internal1_nospec): New define_insn.
9846 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9849 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9851 * caller-save.c (insert_save): Drop unnecessary parameter. All
9854 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9855 Richard Biener <rguenth@suse.de>
9858 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9859 return early, inline manually is_gimple_sizepos. Make sure if we
9860 call gimplify_expr we don't end up with a gimple constant.
9861 * tree.c (variably_modified_type_p): Don't return true for
9862 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9863 * gimplify.h (is_gimple_sizepos): Remove.
9865 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9867 PR tree-optimization/83857
9868 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9869 vectorizable_live_operation for pure SLP statements.
9870 (vectorizable_live_operation): Handle PHIs.
9872 2018-01-16 Richard Biener <rguenther@suse.de>
9874 PR tree-optimization/83867
9875 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9876 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9878 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9881 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9882 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9883 If off is not INTEGER_CST, issue a may not be aligned warning
9884 rather than isn't aligned. Use isn%'t rather than isn't.
9885 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9887 <case MULT_EXPR>: Improve the case when bottom and one of the
9888 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9889 operand, in that case check if the other operand is multiple of
9890 bottom divided by the INTEGER_CST operand.
9892 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9895 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9896 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9897 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9898 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9899 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9900 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9901 (pa_function_arg_size): New function.
9903 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9905 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9906 in a separate statement.
9908 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9910 PR tree-optimization/83847
9911 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9912 group gathers and scatters.
9914 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9916 PR rtl-optimization/86620
9917 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9919 PR rtl-optimization/83213
9920 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9921 to last if both are JUMP_INSNs.
9923 PR tree-optimization/83843
9924 * gimple-ssa-store-merging.c
9925 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9926 store_immediate_info for bswap/nop orig_stores.
9928 2018-01-15 Andrew Waterman <andrew@sifive.com>
9930 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9932 <UDIV>: Increase cost if !TARGET_DIV.
9934 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9936 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9937 (define_attr "cr_logical_3op"): New.
9938 (cceq_ior_compare): Adjust.
9939 (cceq_ior_compare_complement): Adjust.
9940 (*cceq_rev_compare): Adjust.
9941 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9942 (is_cracked_insn): Adjust.
9943 (insn_must_be_first_in_group): Adjust.
9944 * config/rs6000/40x.md: Adjust.
9945 * config/rs6000/440.md: Adjust.
9946 * config/rs6000/476.md: Adjust.
9947 * config/rs6000/601.md: Adjust.
9948 * config/rs6000/603.md: Adjust.
9949 * config/rs6000/6xx.md: Adjust.
9950 * config/rs6000/7450.md: Adjust.
9951 * config/rs6000/7xx.md: Adjust.
9952 * config/rs6000/8540.md: Adjust.
9953 * config/rs6000/cell.md: Adjust.
9954 * config/rs6000/e300c2c3.md: Adjust.
9955 * config/rs6000/e500mc.md: Adjust.
9956 * config/rs6000/e500mc64.md: Adjust.
9957 * config/rs6000/e5500.md: Adjust.
9958 * config/rs6000/e6500.md: Adjust.
9959 * config/rs6000/mpc.md: Adjust.
9960 * config/rs6000/power4.md: Adjust.
9961 * config/rs6000/power5.md: Adjust.
9962 * config/rs6000/power6.md: Adjust.
9963 * config/rs6000/power7.md: Adjust.
9964 * config/rs6000/power8.md: Adjust.
9965 * config/rs6000/power9.md: Adjust.
9966 * config/rs6000/rs64.md: Adjust.
9967 * config/rs6000/titan.md: Adjust.
9969 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9971 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9972 ix86_indirect_branch_register logic.
9974 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9976 * config/i386/constraints.md (Bs): Update
9977 ix86_indirect_branch_register check. Don't check
9978 ix86_indirect_branch_register with GOT_memory_operand.
9980 * config/i386/predicates.md (GOT_memory_operand): Don't check
9981 ix86_indirect_branch_register here.
9982 (GOT32_symbol_operand): Likewise.
9984 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9986 * config/i386/predicates.md (constant_call_address_operand):
9987 Rewrite ix86_indirect_branch_register logic.
9988 (sibcall_insn_operand): Likewise.
9990 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9992 * config/i386/constraints.md (Bs): Replace
9993 ix86_indirect_branch_thunk_register with
9994 ix86_indirect_branch_register.
9996 * config/i386/i386.md (indirect_jump): Likewise.
9997 (tablejump): Likewise.
9998 (*sibcall_memory): Likewise.
9999 (*sibcall_value_memory): Likewise.
10000 Peepholes of indirect call and jump via memory: Likewise.
10001 * config/i386/i386.opt: Likewise.
10002 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10003 (GOT_memory_operand): Likewise.
10004 (call_insn_operand): Likewise.
10005 (sibcall_insn_operand): Likewise.
10006 (GOT32_symbol_operand): Likewise.
10008 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10010 PR middle-end/83837
10011 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10012 type rather than type addr's type points to.
10013 (expand_omp_atomic_mutex): Likewise.
10014 (expand_omp_atomic): Likewise.
10016 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10019 * config/i386/i386.c (output_indirect_thunk_function): Use
10020 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10021 for __x86_return_thunk.
10023 2018-01-15 Richard Biener <rguenther@suse.de>
10025 PR middle-end/83850
10026 * expmed.c (extract_bit_field_1): Fix typo.
10028 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10031 * config/arm/iterators.md (VF): New mode iterator.
10032 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10033 Remove integer-related logic from pattern.
10034 (neon_vabd<mode>_3): Likewise.
10036 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10038 PR middle-end/82694
10039 * common.opt (fstrict-overflow): No longer an alias.
10040 (fwrapv-pointer): New option.
10041 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10042 also for pointer types based on flag_wrapv_pointer.
10043 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10044 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10045 opts->x_flag_wrapv got set.
10046 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10047 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10048 POINTER_TYPE_OVERFLOW_UNDEFINED.
10049 * match.pd: Likewise in address comparison pattern.
10050 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10052 2018-01-15 Richard Biener <rguenther@suse.de>
10055 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10056 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10057 Reset type names to their identifier if their TYPE_DECL doesn't
10058 have linkage (and thus is used for ODR and devirt).
10059 (save_debug_info_for_decl): Remove.
10060 (save_debug_info_for_type): Likewise.
10061 (add_tree_to_fld_list): Adjust.
10062 * tree-pretty-print.c (dump_generic_node): Make dumping of
10063 type names more robust.
10065 2018-01-15 Richard Biener <rguenther@suse.de>
10067 * BASE-VER: Bump to 8.0.1.
10069 2018-01-14 Martin Sebor <msebor@redhat.com>
10072 * builtins.c (check_access): Avoid warning when the no-warning bit
10075 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10077 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10078 * ira-color (allocno_hard_regs_compare): Likewise.
10080 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10083 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10084 Use .pushsection/.popsection.
10086 2018-01-14 Martin Sebor <msebor@redhat.com>
10089 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10091 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10093 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10094 entry from extra_headers.
10095 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10096 extra_headers, make the list bitwise identical to the i?86-*-* one.
10098 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10100 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10101 -mcmodel=large with -mindirect-branch=thunk,
10102 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10103 -mfunction-return=thunk-extern.
10104 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10105 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10106 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10108 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10110 * config/i386/i386.c (print_reg): Print the name of the full
10111 integer register without '%'.
10112 (ix86_print_operand): Handle 'V'.
10113 * doc/extend.texi: Document 'V' modifier.
10115 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10117 * config/i386/constraints.md (Bs): Disallow memory operand for
10118 -mindirect-branch-register.
10120 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10121 (GOT_memory_operand): Likewise.
10122 (call_insn_operand): Likewise.
10123 (sibcall_insn_operand): Likewise.
10124 (GOT32_symbol_operand): Likewise.
10125 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10126 for -mindirect-branch-register.
10127 (tablejump): Likewise.
10128 (*sibcall_memory): Likewise.
10129 (*sibcall_value_memory): Likewise.
10130 Disallow peepholes of indirect call and jump via memory for
10131 -mindirect-branch-register.
10132 (*call_pop): Replace m with Bw.
10133 (*call_value_pop): Likewise.
10134 (*sibcall_pop_memory): Replace m with Bs.
10135 * config/i386/i386.opt (mindirect-branch-register): New option.
10136 * doc/invoke.texi: Document -mindirect-branch-register option.
10138 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10140 * config/i386/i386-protos.h (ix86_output_function_return): New.
10141 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10142 set function_return_type.
10143 (indirect_thunk_name): Add ret_p to indicate thunk for function
10145 (output_indirect_thunk_function): Pass false to
10146 indirect_thunk_name.
10147 (ix86_output_indirect_branch_via_reg): Likewise.
10148 (ix86_output_indirect_branch_via_push): Likewise.
10149 (output_indirect_thunk_function): Create alias for function
10150 return thunk if regno < 0.
10151 (ix86_output_function_return): New function.
10152 (ix86_handle_fndecl_attribute): Handle function_return.
10153 (ix86_attribute_table): Add function_return.
10154 * config/i386/i386.h (machine_function): Add
10155 function_return_type.
10156 * config/i386/i386.md (simple_return_internal): Use
10157 ix86_output_function_return.
10158 (simple_return_internal_long): Likewise.
10159 * config/i386/i386.opt (mfunction-return=): New option.
10160 (indirect_branch): Mention -mfunction-return=.
10161 * doc/extend.texi: Document function_return function attribute.
10162 * doc/invoke.texi: Document -mfunction-return= option.
10164 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10166 * config/i386/i386-opts.h (indirect_branch): New.
10167 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10168 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10169 with local indirect jump when converting indirect call and jump.
10170 (ix86_set_indirect_branch_type): New.
10171 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10172 (indirectlabelno): New.
10173 (indirect_thunk_needed): Likewise.
10174 (indirect_thunk_bnd_needed): Likewise.
10175 (indirect_thunks_used): Likewise.
10176 (indirect_thunks_bnd_used): Likewise.
10177 (INDIRECT_LABEL): Likewise.
10178 (indirect_thunk_name): Likewise.
10179 (output_indirect_thunk): Likewise.
10180 (output_indirect_thunk_function): Likewise.
10181 (ix86_output_indirect_branch_via_reg): Likewise.
10182 (ix86_output_indirect_branch_via_push): Likewise.
10183 (ix86_output_indirect_branch): Likewise.
10184 (ix86_output_indirect_jmp): Likewise.
10185 (ix86_code_end): Call output_indirect_thunk_function if needed.
10186 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10188 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10189 (ix86_attribute_table): Add indirect_branch.
10190 * config/i386/i386.h (machine_function): Add indirect_branch_type
10191 and has_local_indirect_jump.
10192 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10194 (tablejump): Likewise.
10195 (*indirect_jump): Use ix86_output_indirect_jmp.
10196 (*tablejump_1): Likewise.
10197 (simple_return_indirect_internal): Likewise.
10198 * config/i386/i386.opt (mindirect-branch=): New option.
10199 (indirect_branch): New.
10202 (thunk-inline): Likewise.
10203 (thunk-extern): Likewise.
10204 * doc/extend.texi: Document indirect_branch function attribute.
10205 * doc/invoke.texi: Document -mindirect-branch= option.
10207 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10210 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10212 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10214 * ipa-inline.c (want_inline_small_function_p): Return false if
10215 inlining has already failed with CIF_FINAL_ERROR.
10216 (update_caller_keys): Call want_inline_small_function_p before
10218 (update_callee_keys): Likewise.
10220 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10222 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10224 (rs6000_quadword_masked_address_p): Likewise.
10225 (quad_aligned_load_p): Likewise.
10226 (quad_aligned_store_p): Likewise.
10227 (const_load_sequence_p): Add comment to describe the outer-most loop.
10228 (mimic_memory_attributes_and_flags): New function.
10229 (rs6000_gen_stvx): Likewise.
10230 (replace_swapped_aligned_store): Likewise.
10231 (rs6000_gen_lvx): Likewise.
10232 (replace_swapped_aligned_load): Likewise.
10233 (replace_swapped_load_constant): Capitalize argument name in
10234 comment describing this function.
10235 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10236 and stores that access quad-word aligned addresses and replace
10237 with stvx or lvx instructions when appropriate.
10238 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10239 New function prototype.
10240 (rs6000_quadword_masked_address_p): Likewise.
10241 (rs6000_gen_lvx): Likewise.
10242 (rs6000_gen_stvx): Likewise.
10243 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10244 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10245 when memory address is aligned.
10246 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10247 this split to select lvx instruction when memory address is aligned.
10248 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10249 instruction when memory address is aligned.
10250 (*vsx_le_perm_load_v16qi): Likewise.
10251 (four unnamed splitters): Modify to select the stvx instruction
10252 when memory is aligned.
10254 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10256 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10257 which appears in the queue multiple times.
10259 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10260 Alan Hayward <alan.hayward@arm.com>
10261 David Sherwood <david.sherwood@arm.com>
10263 * tree-vectorizer.h (vec_lower_bound): New structure.
10264 (_loop_vec_info): Add check_nonzero and lower_bounds.
10265 (LOOP_VINFO_CHECK_NONZERO): New macro.
10266 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10267 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10268 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10269 fields. Make seg_len the distance travelled, not including the
10271 (dr_direction_indicator): Declare.
10272 (dr_zero_step_indicator): Likewise.
10273 (dr_known_forward_stride_p): Likewise.
10274 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10276 (runtime_alias_check_p): Allow runtime alias checks with
10278 (operator ==): Compare access_size and align.
10279 (prune_runtime_alias_test_list): Rework for new distinction between
10280 the access_size and seg_len.
10281 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10283 (get_segment_min_max): New function.
10284 (create_intersect_range_checks): Use it.
10285 (dr_step_indicator): New function.
10286 (dr_direction_indicator): Likewise.
10287 (dr_zero_step_indicator): Likewise.
10288 (dr_known_forward_stride_p): Likewise.
10289 * tree-loop-distribution.c (data_ref_segment_size): Return
10290 DR_STEP * (niters - 1).
10291 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10293 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10294 (vect_preserves_scalar_order_p): New function, split out from...
10295 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10296 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10297 (vect_vfa_access_size): New function.
10298 (vect_vfa_align): Likewise.
10299 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10300 (dump_lower_bound): New function.
10301 (vect_check_lower_bound): Likewise.
10302 (vect_small_gap_p): Likewise.
10303 (vectorizable_with_step_bound_p): Likewise.
10304 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10305 depencies if the vectorization factor is 1. Convert the checks
10306 for nonzero steps into checks on the bounds of DR_STEP. Try using
10307 a bunds check for variable steps if the minimum required step is
10308 relatively small. Update calls to the dr_with_seg_len
10309 constructor and to vect_compile_time_alias.
10310 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10312 (vect_loop_versioning): Call it.
10313 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10315 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10317 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10318 Alan Hayward <alan.hayward@arm.com>
10319 David Sherwood <david.sherwood@arm.com>
10321 * doc/sourcebuild.texi (vect_scatter_store): Document.
10322 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10324 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10326 * genopinit.c (main): Add supports_vec_scatter_store and
10327 supports_vec_scatter_store_cached to target_optabs.
10328 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10329 IFN_MASK_SCATTER_STORE.
10330 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10332 * internal-fn.h (internal_store_fn_p): Declare.
10333 (internal_fn_stored_value_index): Likewise.
10334 * internal-fn.c (scatter_store_direct): New macro.
10335 (expand_scatter_store_optab_fn): New function.
10336 (direct_scatter_store_optab_supported_p): New macro.
10337 (internal_store_fn_p): New function.
10338 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10339 IFN_MASK_SCATTER_STORE.
10340 (internal_fn_mask_index): Likewise.
10341 (internal_fn_stored_value_index): New function.
10342 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10343 for scatter stores.
10344 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10345 * optabs-query.c (supports_vec_scatter_store_p): New function.
10346 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10347 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10348 true for scatter stores.
10349 (vect_gather_scatter_fn_p): Handle scatter stores too.
10350 (vect_check_gather_scatter): Consider using scatter stores if
10351 supports_vec_scatter_store_p.
10352 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10353 scatter stores too.
10354 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10355 internal_fn_stored_value_index.
10356 (check_load_store_masking): Handle scatter stores too.
10357 (vect_get_store_rhs): Make public.
10358 (vectorizable_call): Use internal_store_fn_p.
10359 (vectorizable_store): Handle scatter store internal functions.
10360 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10361 when deciding whether the end of the group has been reached.
10362 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10363 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10364 (mask_scatter_store<mode>): New insns.
10366 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10367 Alan Hayward <alan.hayward@arm.com>
10368 David Sherwood <david.sherwood@arm.com>
10370 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10371 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10372 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10374 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10375 Use vect_truncate_gather_scatter_offset if we can't treat the
10376 operation as a normal gather load or scatter store.
10377 (get_group_load_store_type): Take the gather_scatter_info
10378 as argument. Try using a gather load or scatter store for
10379 single-element groups.
10380 (get_load_store_type): Update calls to get_group_load_store_type
10381 and vect_use_strided_gather_scatters_p.
10383 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10384 Alan Hayward <alan.hayward@arm.com>
10385 David Sherwood <david.sherwood@arm.com>
10387 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10388 optional tree argument.
10389 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10391 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10392 but continue to use the current value as a fallback.
10393 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10394 to compare the updates.
10395 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10396 (get_load_store_type): Use it when handling a strided access.
10397 (vect_get_strided_load_store_ops): New function.
10398 (vect_get_data_ptr_increment): Likewise.
10399 (vectorizable_load): Handle strided gather loads. Always pass
10400 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10402 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10403 Alan Hayward <alan.hayward@arm.com>
10404 David Sherwood <david.sherwood@arm.com>
10406 * doc/md.texi (gather_load@var{m}): Document.
10407 (mask_gather_load@var{m}): Likewise.
10408 * genopinit.c (main): Add supports_vec_gather_load and
10409 supports_vec_gather_load_cached to target_optabs.
10410 * optabs-tree.c (init_tree_optimization_optabs): Use
10411 ggc_cleared_alloc to allocate target_optabs.
10412 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10413 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10415 * internal-fn.h (internal_load_fn_p): Declare.
10416 (internal_gather_scatter_fn_p): Likewise.
10417 (internal_fn_mask_index): Likewise.
10418 (internal_gather_scatter_fn_supported_p): Likewise.
10419 * internal-fn.c (gather_load_direct): New macro.
10420 (expand_gather_load_optab_fn): New function.
10421 (direct_gather_load_optab_supported_p): New macro.
10422 (direct_internal_fn_optab): New function.
10423 (internal_load_fn_p): Likewise.
10424 (internal_gather_scatter_fn_p): Likewise.
10425 (internal_fn_mask_index): Likewise.
10426 (internal_gather_scatter_fn_supported_p): Likewise.
10427 * optabs-query.c (supports_at_least_one_mode_p): New function.
10428 (supports_vec_gather_load_p): Likewise.
10429 * optabs-query.h (supports_vec_gather_load_p): Declare.
10430 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10431 and memory_type field.
10432 (NUM_PATTERNS): Bump to 15.
10433 * tree-vect-data-refs.c: Include internal-fn.h.
10434 (vect_gather_scatter_fn_p): New function.
10435 (vect_describe_gather_scatter_call): Likewise.
10436 (vect_check_gather_scatter): Try using internal functions for
10437 gather loads. Recognize existing calls to a gather load function.
10438 (vect_analyze_data_refs): Consider using gather loads if
10439 supports_vec_gather_load_p.
10440 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10441 (vect_get_gather_scatter_offset_type): Likewise.
10442 (vect_convert_mask_for_vectype): Likewise.
10443 (vect_add_conversion_to_patterm): Likewise.
10444 (vect_try_gather_scatter_pattern): Likewise.
10445 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10446 (vect_vect_recog_func_ptrs): Add it.
10447 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10448 internal_fn_mask_index and internal_gather_scatter_fn_p.
10449 (check_load_store_masking): Take the gather_scatter_info as an
10450 argument and handle gather loads.
10451 (vect_get_gather_scatter_ops): New function.
10452 (vectorizable_call): Check internal_load_fn_p.
10453 (vectorizable_load): Likewise. Handle gather load internal
10455 (vectorizable_store): Update call to check_load_store_masking.
10456 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10457 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10458 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10459 (aarch64_gather_scale_operand_d): New predicates.
10460 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10461 (mask_gather_load<mode>): New insns.
10463 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10464 Alan Hayward <alan.hayward@arm.com>
10465 David Sherwood <david.sherwood@arm.com>
10467 * optabs.def (fold_left_plus_optab): New optab.
10468 * doc/md.texi (fold_left_plus_@var{m}): Document.
10469 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10470 * internal-fn.c (fold_left_direct): Define.
10471 (expand_fold_left_optab_fn): Likewise.
10472 (direct_fold_left_optab_supported_p): Likewise.
10473 * fold-const-call.c (fold_const_fold_left): New function.
10474 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10475 * tree-parloops.c (valid_reduction_p): New function.
10476 (gather_scalar_reductions): Use it.
10477 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10478 (vect_finish_replace_stmt): Declare.
10479 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10480 (needs_fold_left_reduction_p): New function, split out from...
10481 (vect_is_simple_reduction): ...here. Accept reductions that
10482 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10483 (vect_force_simple_reduction): Also store the reduction type in
10484 the assignment's STMT_VINFO_REDUC_TYPE.
10485 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10486 (merge_with_identity): New function.
10487 (vect_expand_fold_left): Likewise.
10488 (vectorize_fold_left_reduction): Likewise.
10489 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10490 scalar phi in place for it. Check for target support and reject
10491 cases that would reassociate the operation. Defer the transform
10492 phase to vectorize_fold_left_reduction.
10493 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10494 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10495 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10499 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10500 call to ifc_temp_var.
10502 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10503 Alan Hayward <alan.hayward@arm.com>
10504 David Sherwood <david.sherwood@arm.com>
10506 * target.def (legitimize_address_displacement): Take the original
10507 offset as a poly_int.
10508 * targhooks.h (default_legitimize_address_displacement): Update
10510 * targhooks.c (default_legitimize_address_displacement): Likewise.
10511 * doc/tm.texi: Regenerate.
10512 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10513 as an argument, moving assert of ad->disp == ad->disp_term to...
10514 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10515 Try calling targetm.legitimize_address_displacement before expanding
10516 the address rather than afterwards, and adjust for the new interface.
10517 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10518 Match the new hook interface. Handle SVE addresses.
10519 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10520 new hook interface.
10522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10524 * Makefile.in (OBJS): Add early-remat.o.
10525 * target.def (select_early_remat_modes): New hook.
10526 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10527 * doc/tm.texi: Regenerate.
10528 * targhooks.h (default_select_early_remat_modes): Declare.
10529 * targhooks.c (default_select_early_remat_modes): New function.
10530 * timevar.def (TV_EARLY_REMAT): New timevar.
10531 * passes.def (pass_early_remat): New pass.
10532 * tree-pass.h (make_pass_early_remat): Declare.
10533 * early-remat.c: New file.
10534 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10536 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10538 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10539 Alan Hayward <alan.hayward@arm.com>
10540 David Sherwood <david.sherwood@arm.com>
10542 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10543 vfm1 with a bound_epilog parameter.
10544 (vect_do_peeling): Update calls accordingly, and move the prologue
10545 call earlier in the function. Treat the base bound_epilog as 0 for
10546 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10547 this base when peeling for gaps.
10548 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10549 with fully-masked loops.
10550 (vect_estimate_min_profitable_iters): Handle the single peeled
10551 iteration in that case.
10553 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10554 Alan Hayward <alan.hayward@arm.com>
10555 David Sherwood <david.sherwood@arm.com>
10557 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10558 single-element interleaving even if the size is not a power of 2.
10559 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10560 accesses for single-element interleaving if the group size is
10563 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10564 Alan Hayward <alan.hayward@arm.com>
10565 David Sherwood <david.sherwood@arm.com>
10567 * doc/md.texi (fold_extract_last_@var{m}): Document.
10568 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10569 * optabs.def (fold_extract_last_optab): New optab.
10570 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10571 * internal-fn.c (fold_extract_direct): New macro.
10572 (expand_fold_extract_optab_fn): Likewise.
10573 (direct_fold_extract_optab_supported_p): Likewise.
10574 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10575 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10576 EXTRACT_LAST_REDUCTION.
10577 (get_initial_def_for_reduction): Do not create an initial vector
10578 for EXTRACT_LAST_REDUCTION reductions.
10579 (vectorizable_reduction): Leave the scalar phi in place for
10580 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10581 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10582 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10583 transform phase to vectorizable_condition.
10584 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10586 (vect_finish_stmt_generation): ...here.
10587 (vect_finish_replace_stmt): New function.
10588 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10589 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10591 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10593 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10594 Alan Hayward <alan.hayward@arm.com>
10595 David Sherwood <david.sherwood@arm.com>
10597 * doc/md.texi (extract_last_@var{m}): Document.
10598 * optabs.def (extract_last_optab): New optab.
10599 * internal-fn.def (EXTRACT_LAST): New internal function.
10600 * internal-fn.c (cond_unary_direct): New macro.
10601 (expand_cond_unary_optab_fn): Likewise.
10602 (direct_cond_unary_optab_supported_p): Likewise.
10603 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10604 loops using EXTRACT_LAST.
10605 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10606 (extract_last_<mode>): ...this optab.
10607 (vec_extract<mode><Vel>): Update accordingly.
10609 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10610 Alan Hayward <alan.hayward@arm.com>
10611 David Sherwood <david.sherwood@arm.com>
10613 * target.def (empty_mask_is_expensive): New hook.
10614 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10615 * doc/tm.texi: Regenerate.
10616 * targhooks.h (default_empty_mask_is_expensive): Declare.
10617 * targhooks.c (default_empty_mask_is_expensive): New function.
10618 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10619 if the target says that empty masks are expensive.
10620 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10622 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10624 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10625 Alan Hayward <alan.hayward@arm.com>
10626 David Sherwood <david.sherwood@arm.com>
10628 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10629 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10630 (vect_use_loop_mask_for_alignment_p): New function.
10631 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10632 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10633 niters_skip argument. Make sure that the first niters_skip elements
10634 of the first iteration are inactive.
10635 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10636 Update call to vect_set_loop_masks_directly.
10637 (get_misalign_in_elems): New function, split out from...
10638 (vect_gen_prolog_loop_niters): ...here.
10639 (vect_update_init_of_dr): Take a code argument that specifies whether
10640 the adjustment should be added or subtracted.
10641 (vect_update_init_of_drs): Likewise.
10642 (vect_prepare_for_masked_peels): New function.
10643 (vect_do_peeling): Skip prologue peeling if we're using a mask
10644 instead. Update call to vect_update_inits_of_drs.
10645 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10647 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10648 alignment. Do not include the number of peeled iterations in
10649 the minimum threshold in that case.
10650 (vectorizable_induction): Adjust the start value down by
10651 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10652 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10653 Take the number of skipped iterations into account when calculating
10655 * tree-vect-stmts.c (vect_gen_while_not): New function.
10657 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10658 Alan Hayward <alan.hayward@arm.com>
10659 David Sherwood <david.sherwood@arm.com>
10661 * doc/sourcebuild.texi (vect_fully_masked): Document.
10662 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10663 default value to 0.
10664 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10666 (vect_analyze_loop_2): ...here. Don't check the vectorization
10667 factor against the number of loop iterations if the loop is
10670 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10671 Alan Hayward <alan.hayward@arm.com>
10672 David Sherwood <david.sherwood@arm.com>
10674 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10675 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10676 (dump_groups): Update accordingly.
10677 (iv_use::mem_type): New member variable.
10678 (address_p): New function.
10679 (record_use): Add a mem_type argument and initialize the new
10681 (record_group_use): Add a mem_type argument. Use address_p.
10682 Remove obsolete null checks of base_object. Update call to record_use.
10683 (find_interesting_uses_op): Update call to record_group_use.
10684 (find_interesting_uses_cond): Likewise.
10685 (find_interesting_uses_address): Likewise.
10686 (get_mem_type_for_internal_fn): New function.
10687 (find_address_like_use): Likewise.
10688 (find_interesting_uses_stmt): Try find_address_like_use before
10689 calling find_interesting_uses_op.
10690 (addr_offset_valid_p): Use the iv mem_type field as the type
10691 of the addressed memory.
10692 (add_autoinc_candidates): Likewise.
10693 (get_address_cost): Likewise.
10694 (split_small_address_groups_p): Use address_p.
10695 (split_address_groups): Likewise.
10696 (add_iv_candidate_for_use): Likewise.
10697 (autoinc_possible_for_pair): Likewise.
10698 (rewrite_groups): Likewise.
10699 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10700 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10701 (get_alias_ptr_type_for_ptr_address): New function.
10702 (rewrite_use_address): Rewrite address uses in calls that were
10703 identified by find_address_like_use.
10705 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10706 Alan Hayward <alan.hayward@arm.com>
10707 David Sherwood <david.sherwood@arm.com>
10709 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10711 * gimple-expr.h (is_gimple_addressable: Likewise.
10712 * gimple-expr.c (is_gimple_address): Likewise.
10713 * internal-fn.c (expand_call_mem_ref): New function.
10714 (expand_mask_load_optab_fn): Use it.
10715 (expand_mask_store_optab_fn): Likewise.
10717 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10718 Alan Hayward <alan.hayward@arm.com>
10719 David Sherwood <david.sherwood@arm.com>
10721 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10722 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10723 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10724 (cond_umax@var{mode}): Document.
10725 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10726 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10727 (cond_umin_optab, cond_umax_optab): New optabs.
10728 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10729 (COND_IOR, COND_XOR): New internal functions.
10730 * internal-fn.h (get_conditional_internal_fn): Declare.
10731 * internal-fn.c (cond_binary_direct): New macro.
10732 (expand_cond_binary_optab_fn): Likewise.
10733 (direct_cond_binary_optab_supported_p): Likewise.
10734 (get_conditional_internal_fn): New function.
10735 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10736 Cope with reduction statements that are vectorized as calls rather
10738 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10739 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10740 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10741 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10742 (UNSPEC_COND_EOR): New unspecs.
10743 (optab): Add mappings for them.
10744 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10745 (sve_int_op, sve_fp_op): New int attributes.
10747 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10748 Alan Hayward <alan.hayward@arm.com>
10749 David Sherwood <david.sherwood@arm.com>
10751 * optabs.def (while_ult_optab): New optab.
10752 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10753 * internal-fn.def (WHILE_ULT): New internal function.
10754 * internal-fn.h (direct_internal_fn_supported_p): New override
10755 that takes two types as argument.
10756 * internal-fn.c (while_direct): New macro.
10757 (expand_while_optab_fn): New function.
10758 (convert_optab_supported_p): Likewise.
10759 (direct_while_optab_supported_p): New macro.
10760 * wide-int.h (wi::udiv_ceil): New function.
10761 * tree-vectorizer.h (rgroup_masks): New structure.
10762 (vec_loop_masks): New typedef.
10763 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10764 and fully_masked_p.
10765 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10766 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10767 (vect_max_vf): New function.
10768 (slpeel_make_loop_iterate_ntimes): Delete.
10769 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10770 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10771 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10772 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10773 internal-fn.h, stor-layout.h and optabs-query.h.
10774 (vect_set_loop_mask): New function.
10775 (add_preheader_seq): Likewise.
10776 (add_header_seq): Likewise.
10777 (interleave_supported_p): Likewise.
10778 (vect_maybe_permute_loop_masks): Likewise.
10779 (vect_set_loop_masks_directly): Likewise.
10780 (vect_set_loop_condition_masked): Likewise.
10781 (vect_set_loop_condition_unmasked): New function, split out from
10782 slpeel_make_loop_iterate_ntimes.
10783 (slpeel_make_loop_iterate_ntimes): Rename to..
10784 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10785 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10786 (vect_do_peeling): Update call accordingly.
10787 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10789 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10790 mask_compare_type, can_fully_mask_p and fully_masked_p.
10791 (release_vec_loop_masks): New function.
10792 (_loop_vec_info): Use it to free the loop masks.
10793 (can_produce_all_loop_masks_p): New function.
10794 (vect_get_max_nscalars_per_iter): Likewise.
10795 (vect_verify_full_masking): Likewise.
10796 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10797 retries, and free the mask rgroups before retrying. Check loop-wide
10798 reasons for disallowing fully-masked loops. Make the final decision
10799 about whether use a fully-masked loop or not.
10800 (vect_estimate_min_profitable_iters): Do not assume that peeling
10801 for the number of iterations will be needed for fully-masked loops.
10802 (vectorizable_reduction): Disable fully-masked loops.
10803 (vectorizable_live_operation): Likewise.
10804 (vect_halve_mask_nunits): New function.
10805 (vect_double_mask_nunits): Likewise.
10806 (vect_record_loop_mask): Likewise.
10807 (vect_get_loop_mask): Likewise.
10808 (vect_transform_loop): Handle the case in which the final loop
10809 iteration might handle a partial vector. Call vect_set_loop_condition
10810 instead of slpeel_make_loop_iterate_ntimes.
10811 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10812 (check_load_store_masking): New function.
10813 (prepare_load_store_mask): Likewise.
10814 (vectorizable_store): Handle fully-masked loops.
10815 (vectorizable_load): Likewise.
10816 (supportable_widening_operation): Use vect_halve_mask_nunits for
10818 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10819 (vect_gen_while): New function.
10820 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10821 (aarch64_uqdec<mode>): New insn.
10823 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10824 Alan Hayward <alan.hayward@arm.com>
10825 David Sherwood <david.sherwood@arm.com>
10827 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10828 (reduc_xor_scal_optab): New optabs.
10829 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10830 (reduc_xor_scal_@var{m}): Document.
10831 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10832 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10833 internal functions.
10834 * fold-const-call.c (fold_const_call): Handle them.
10835 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10836 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10837 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10838 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10839 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10840 (UNSPEC_XORV): New unspecs.
10841 (optab): Add entries for them.
10842 (BITWISEV): New int iterator.
10843 (bit_reduc_op): New int attributes.
10845 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10846 Alan Hayward <alan.hayward@arm.com>
10847 David Sherwood <david.sherwood@arm.com>
10849 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10850 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10851 * optabs.def (vec_shl_insert_optab): New optab.
10852 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10853 (duplicate_and_interleave): Likewise.
10854 * tree-vect-loop.c: Include internal-fn.h.
10855 (neutral_op_for_slp_reduction): New function, split out from
10856 get_initial_defs_for_reduction.
10857 (get_initial_def_for_reduction): Handle option 2 for variable-length
10858 vectors by loading the neutral value into a vector and then shifting
10859 the initial value into element 0.
10860 (get_initial_defs_for_reduction): Replace the code argument with
10861 the neutral value calculated by neutral_op_for_slp_reduction.
10862 Use gimple_build_vector for constant-length vectors.
10863 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10864 but the first group_size elements have a neutral value.
10865 Use duplicate_and_interleave otherwise.
10866 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10867 Update call to get_initial_defs_for_reduction. Handle SLP
10868 reductions for variable-length vectors by creating one vector
10869 result for each scalar result, with the elements associated
10870 with other scalar results stubbed out with the neutral value.
10871 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10872 Require IFN_VEC_SHL_INSERT for double reductions on
10873 variable-length vectors, or SLP reductions that have
10874 a neutral value. Require can_duplicate_and_interleave_p
10875 support for variable-length unchained SLP reductions if there
10876 is no neutral value, such as for MIN/MAX reductions. Also require
10877 the number of vector elements to be a multiple of the number of
10878 SLP statements when doing variable-length unchained SLP reductions.
10879 Update call to vect_create_epilog_for_reduction.
10880 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10881 and remove initial values.
10882 (duplicate_and_interleave): Make public.
10883 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10884 * config/aarch64/aarch64-sve.md (vec_shl_insert_<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 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10891 (can_duplicate_and_interleave_p): New function.
10892 (vect_get_and_check_slp_defs): Take the vector of statements
10893 rather than just the current one. Remove excess parentheses.
10894 Restriction rejectinon of vect_constant_def and vect_external_def
10895 for variable-length vectors to boolean types, or types for which
10896 can_duplicate_and_interleave_p is false.
10897 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10898 (duplicate_and_interleave): New function.
10899 (vect_get_constant_vectors): Use gimple_build_vector for
10900 constant-length vectors and suitable variable-length constant
10901 vectors. Use duplicate_and_interleave for other variable-length
10902 vectors. Don't defer the update when inserting new statements.
10904 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10905 Alan Hayward <alan.hayward@arm.com>
10906 David Sherwood <david.sherwood@arm.com>
10908 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10909 min_profitable_iters doesn't go negative.
10911 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10912 Alan Hayward <alan.hayward@arm.com>
10913 David Sherwood <david.sherwood@arm.com>
10915 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10916 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10917 * optabs.def (vec_mask_load_lanes_optab): New optab.
10918 (vec_mask_store_lanes_optab): Likewise.
10919 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10920 (MASK_STORE_LANES): Likewise.
10921 * internal-fn.c (mask_load_lanes_direct): New macro.
10922 (mask_store_lanes_direct): Likewise.
10923 (expand_mask_load_optab_fn): Handle masked operations.
10924 (expand_mask_load_lanes_optab_fn): New macro.
10925 (expand_mask_store_optab_fn): Handle masked operations.
10926 (expand_mask_store_lanes_optab_fn): New macro.
10927 (direct_mask_load_lanes_optab_supported_p): Likewise.
10928 (direct_mask_store_lanes_optab_supported_p): Likewise.
10929 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10931 (vect_load_lanes_supported): Likewise.
10932 * tree-vect-data-refs.c (strip_conversion): New function.
10933 (can_group_stmts_p): Likewise.
10934 (vect_analyze_data_ref_accesses): Use it instead of checking
10935 for a pair of assignments.
10936 (vect_store_lanes_supported): Take a masked_p parameter.
10937 (vect_load_lanes_supported): Likewise.
10938 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10939 vect_store_lanes_supported and vect_load_lanes_supported.
10940 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10941 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10942 parameter. Don't allow gaps for masked accesses.
10943 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10944 and vect_load_lanes_supported.
10945 (get_load_store_type): Take a masked_p parameter and update
10946 call to get_group_load_store_type.
10947 (vectorizable_store): Update call to get_load_store_type.
10948 Handle IFN_MASK_STORE_LANES.
10949 (vectorizable_load): Update call to get_load_store_type.
10950 Handle IFN_MASK_LOAD_LANES.
10952 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10953 Alan Hayward <alan.hayward@arm.com>
10954 David Sherwood <david.sherwood@arm.com>
10956 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10958 * config/aarch64/aarch64-protos.h
10959 (aarch64_sve_struct_memory_operand_p): Declare.
10960 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10961 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10962 (VPRED, vpred): Handle SVE structure modes.
10963 * config/aarch64/constraints.md (Utx): New constraint.
10964 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10965 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10966 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10967 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10968 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10969 structure modes. Split into pieces after RA.
10970 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10971 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10973 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10974 SVE structure modes.
10975 (aarch64_classify_address): Likewise.
10976 (sizetochar): Move earlier in file.
10977 (aarch64_print_operand): Handle SVE register lists.
10978 (aarch64_array_mode): New function.
10979 (aarch64_sve_struct_memory_operand_p): Likewise.
10980 (TARGET_ARRAY_MODE): Redefine.
10982 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10983 Alan Hayward <alan.hayward@arm.com>
10984 David Sherwood <david.sherwood@arm.com>
10986 * target.def (array_mode): New target hook.
10987 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10988 * doc/tm.texi: Regenerate.
10989 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10990 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10991 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10992 targetm.array_mode.
10993 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10996 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10997 Alan Hayward <alan.hayward@arm.com>
10998 David Sherwood <david.sherwood@arm.com>
11000 * fold-const.c (fold_binary_loc): Check the argument types
11001 rather than the result type when testing for a vector operation.
11003 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11005 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11006 * doc/tm.texi: Regenerate.
11008 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11009 Alan Hayward <alan.hayward@arm.com>
11010 David Sherwood <david.sherwood@arm.com>
11012 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11013 (sve): Document new AArch64 extension.
11014 * doc/md.texi (w): Extend the description of the AArch64
11015 constraint to include SVE vectors.
11016 (Upl, Upa): Document new AArch64 predicate constraints.
11017 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11019 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11020 (msve-vector-bits=): New option.
11021 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11022 SVE when these are disabled.
11023 (sve): New extension.
11024 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11025 modes. Adjust their number of units based on aarch64_sve_vg.
11026 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11027 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11028 aarch64_addr_query_type.
11029 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11030 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11031 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11032 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11033 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11034 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11035 (aarch64_simd_imm_zero_p): Delete.
11036 (aarch64_check_zero_based_sve_index_immediate): Declare.
11037 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11038 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11039 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11040 (aarch64_sve_float_mul_immediate_p): Likewise.
11041 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11042 rather than an rtx.
11043 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11044 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11045 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11046 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11047 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11048 (aarch64_regmode_natural_size): Likewise.
11049 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11050 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11052 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11053 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11054 for VG and the SVE predicate registers.
11055 (V_ALIASES): Add a "z"-prefixed alias.
11056 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11057 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11058 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11059 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11060 (REG_CLASS_NAMES): Add entries for them.
11061 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11062 and the predicate registers.
11063 (aarch64_sve_vg): Declare.
11064 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11065 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11066 (REGMODE_NATURAL_SIZE): Define.
11067 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11069 * config/aarch64/aarch64.c: Include cfgrtl.h.
11070 (simd_immediate_info): Add a constructor for series vectors,
11071 and an associated step field.
11072 (aarch64_sve_vg): New variable.
11073 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11074 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11075 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11076 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11077 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11078 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11079 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11080 (aarch64_get_mask_mode): New functions.
11081 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11082 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11083 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11084 predicate modes and predicate registers. Explicitly restrict
11085 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11086 to store a vector mode if it is recognized by
11087 aarch64_classify_vector_mode.
11088 (aarch64_regmode_natural_size): New function.
11089 (aarch64_hard_regno_caller_save_mode): Return the original mode
11091 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11092 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11093 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11094 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11096 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11097 does not overlap dest if the function is frame-related. Handle
11099 (aarch64_split_add_offset): New function.
11100 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11101 them aarch64_add_offset.
11102 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11103 and update call to aarch64_sub_sp.
11104 (aarch64_add_cfa_expression): New function.
11105 (aarch64_expand_prologue): Pass extra temporary registers to the
11106 functions above. Handle the case in which we need to emit new
11107 DW_CFA_expressions for registers that were originally saved
11108 relative to the stack pointer, but now have to be expressed
11109 relative to the frame pointer.
11110 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11112 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11113 IP0 and IP1 values for SVE frames.
11114 (aarch64_expand_vec_series): New function.
11115 (aarch64_expand_sve_widened_duplicate): Likewise.
11116 (aarch64_expand_sve_const_vector): Likewise.
11117 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11118 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11119 into the register, rather than emitting a SET directly.
11120 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11121 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11122 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11123 (offset_9bit_signed_scaled_p): New functions.
11124 (aarch64_replicate_bitmask_imm): New function.
11125 (aarch64_bitmask_imm): Use it.
11126 (aarch64_cannot_force_const_mem): Reject expressions involving
11127 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11128 (aarch64_classify_index): Handle SVE indices, by requiring
11129 a plain register index with a scale that matches the element size.
11130 (aarch64_classify_address): Handle SVE addresses. Assert that
11131 the mode of the address is VOIDmode or an integer mode.
11132 Update call to aarch64_classify_symbol.
11133 (aarch64_classify_symbolic_expression): Update call to
11134 aarch64_classify_symbol.
11135 (aarch64_const_vec_all_in_range_p): New function.
11136 (aarch64_print_vector_float_operand): Likewise.
11137 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11138 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11139 and the FP immediates 1.0 and 0.5.
11140 (aarch64_print_address_internal): Handle SVE addresses.
11141 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11142 (aarch64_regno_regclass): Handle predicate registers.
11143 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11145 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11146 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11147 (aarch64_convert_sve_vector_bits): New function.
11148 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11149 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11150 rather than an rtx.
11151 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11152 Handle SVE vector and predicate modes. Accept VL-based constants
11153 that need only one temporary register, and VL offsets that require
11154 no temporary registers.
11155 (aarch64_conditional_register_usage): Mark the predicate registers
11156 as fixed if SVE isn't available.
11157 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11158 Return true for SVE vector and predicate modes.
11159 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11160 rather than an unsigned int. Handle SVE modes.
11161 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11163 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11165 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11166 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11167 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11168 (aarch64_sve_float_mul_immediate_p): New functions.
11169 (aarch64_sve_valid_immediate): New function.
11170 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11171 Explicitly reject structure modes. Check for INDEX constants.
11172 Handle PTRUE and PFALSE constants.
11173 (aarch64_check_zero_based_sve_index_immediate): New function.
11174 (aarch64_simd_imm_zero_p): Delete.
11175 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11176 vector modes. Accept constants in the range of CNT[BHWD].
11177 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11178 ask for an Advanced SIMD mode.
11179 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11180 (aarch64_simd_vector_alignment): Handle SVE predicates.
11181 (aarch64_vectorize_preferred_vector_alignment): New function.
11182 (aarch64_simd_vector_alignment_reachable): Use it instead of
11184 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11185 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11187 (MAX_VECT_LEN): Delete.
11188 (expand_vec_perm_d): Add a vec_flags field.
11189 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11190 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11191 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11193 (aarch64_evpc_rev): Rename to...
11194 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11195 (aarch64_evpc_rev_global): New function.
11196 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11197 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11199 (aarch64_evpc_sve_tbl): New function.
11200 (aarch64_expand_vec_perm_const_1): Update after rename of
11201 aarch64_evpc_rev. Handle SVE permutes too, trying
11202 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11203 than aarch64_evpc_tbl.
11204 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11205 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11206 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11207 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11208 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11209 (aarch64_expand_sve_vcond): New functions.
11210 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11211 of aarch64_vector_mode_p.
11212 (aarch64_dwarf_poly_indeterminate_value): New function.
11213 (aarch64_compute_pressure_classes): Likewise.
11214 (aarch64_can_change_mode_class): Likewise.
11215 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11216 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11217 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11218 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11219 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11220 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11221 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11222 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11224 (Dn, Dl, Dr): Accept const as well as const_vector.
11225 (Dz): Likewise. Compare against CONST0_RTX.
11226 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11227 of "vector" where appropriate.
11228 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11229 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11230 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11231 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11232 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11233 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11234 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11235 (v_int_equiv): Extend to SVE modes.
11236 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11238 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11239 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11240 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11241 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11242 (SVE_COND_FP_CMP): New int iterators.
11243 (perm_hilo): Handle the new unpack unspecs.
11244 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11246 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11247 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11248 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11249 (aarch64_equality_operator, aarch64_constant_vector_operand)
11250 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11251 (aarch64_sve_nonimmediate_operand): Likewise.
11252 (aarch64_sve_general_operand): Likewise.
11253 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11254 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11255 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11256 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11257 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11258 (aarch64_sve_float_arith_immediate): Likewise.
11259 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11260 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11261 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11262 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11263 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11264 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11265 (aarch64_sve_float_arith_operand): Likewise.
11266 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11267 (aarch64_sve_float_mul_operand): Likewise.
11268 (aarch64_sve_vec_perm_operand): Likewise.
11269 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11270 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11271 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11272 as well as const_vector.
11273 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11274 in file. Use CONST0_RTX and CONSTM1_RTX.
11275 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11276 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11277 Use aarch64_simd_imm_zero.
11278 * config/aarch64/aarch64-sve.md: New file.
11279 * config/aarch64/aarch64.md: Include it.
11280 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11281 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11282 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11283 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11284 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11285 (sve): New attribute.
11286 (enabled): Disable instructions with the sve attribute unless
11288 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11289 aarch64_expand_mov_immediate.
11290 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11291 CNT[BHSD] immediates.
11292 (movti): Split CONST_POLY_INT moves into two halves.
11293 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11294 Split additions that need a temporary here if the destination
11295 is the stack pointer.
11296 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11297 (*add<mode>3_poly_1): New instruction.
11298 (set_clobber_cc): New expander.
11300 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11302 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11303 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11304 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11305 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11306 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11307 Change innermode from fixed_mode_size to machine_mode.
11308 (simplify_subreg): Update call accordingly. Handle a constant-sized
11309 subreg of a variable-length CONST_VECTOR.
11311 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11312 Alan Hayward <alan.hayward@arm.com>
11313 David Sherwood <david.sherwood@arm.com>
11315 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11316 (add_offset_to_base): New function, split out from...
11317 (create_mem_ref): ...here. When handling a scale other than 1,
11318 check first whether the address is valid without the offset.
11319 Add it into the base if so, leaving the index and scale as-is.
11321 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11324 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11325 fold_for_warn before checking if arg2 is INTEGER_CST.
11327 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11329 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11330 (store_multiple_operation): Delete.
11331 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11332 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11333 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11334 guarded by TARGET_STRING.
11335 (rs6000_output_load_multiple): Delete.
11336 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11337 OPTION_MASK_STRING / TARGET_STRING handling.
11338 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11339 (const rs6000_opt_masks) <"string">: Change mask to 0.
11340 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11341 (MASK_STRING): Delete.
11342 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11344 (load_multiple): Delete.
11351 (store_multiple): Delete.
11358 (movmemsi_8reg): Delete.
11359 (corresponding unnamed define_insn): Delete.
11360 (movmemsi_6reg): Delete.
11361 (corresponding unnamed define_insn): Delete.
11362 (movmemsi_4reg): Delete.
11363 (corresponding unnamed define_insn): Delete.
11364 (movmemsi_2reg): Delete.
11365 (corresponding unnamed define_insn): Delete.
11366 (movmemsi_1reg): Delete.
11367 (corresponding unnamed define_insn): Delete.
11368 * config/rs6000/rs6000.opt (mno-string): New.
11369 (mstring): Replace by deprecation warning stub.
11370 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11372 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11374 * regrename.c (regrename_do_replace): If replacing the same
11375 reg multiple times, try to reuse last created gen_raw_REG.
11378 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11379 main to workaround a bug in GDB.
11381 2018-01-12 Tom de Vries <tom@codesourcery.com>
11384 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11386 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11388 PR rtl-optimization/80481
11389 * ira-color.c (get_cap_member): New function.
11390 (allocnos_conflict_by_live_ranges_p): Use it.
11391 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11392 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11394 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11397 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11398 (*saddl_se_1): Ditto.
11399 (*ssubsi_1): Ditto.
11400 (*ssubl_se_1): Ditto.
11402 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11404 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11405 rather than wi::to_widest for DR_INITs.
11406 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11407 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11408 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11410 (vect_analyze_group_access_1): Note that here.
11412 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11414 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11415 polynomial type sizes.
11417 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11419 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11420 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11421 (gimple_add_tmp_var): Likewise.
11423 2018-01-12 Martin Liska <mliska@suse.cz>
11425 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11426 (gimple_alloc_sizes): Likewise.
11427 (dump_gimple_statistics): Use PRIu64 in printf format.
11428 * gimple.h: Change uint64_t to int.
11430 2018-01-12 Martin Liska <mliska@suse.cz>
11432 * tree-core.h: Use uint64_t instead of int.
11433 * tree.c (tree_node_counts): Likewise.
11434 (tree_node_sizes): Likewise.
11435 (dump_tree_statistics): Use PRIu64 in printf format.
11437 2018-01-12 Martin Liska <mliska@suse.cz>
11439 * Makefile.in: As qsort_chk is implemented in vec.c, add
11440 vec.o to linkage of gencfn-macros.
11441 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11442 passing the info to record_node_allocation_statistics.
11443 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11445 * ggc-common.c (struct ggc_usage): Add operator== and use
11446 it in operator< and compare function.
11447 * mem-stats.h (struct mem_usage): Likewise.
11448 * vec.c (struct vec_usage): Remove operator< and compare
11449 function. Can be simply inherited.
11451 2018-01-12 Martin Jambor <mjambor@suse.cz>
11454 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11455 * tree-ssa-math-opts.c: Include domwalk.h.
11456 (convert_mult_to_fma_1): New function.
11457 (fma_transformation_info): New type.
11458 (fma_deferring_state): Likewise.
11459 (cancel_fma_deferring): New function.
11460 (result_of_phi): Likewise.
11461 (last_fma_candidate_feeds_initial_phi): Likewise.
11462 (convert_mult_to_fma): Added deferring logic, split actual
11463 transformation to convert_mult_to_fma_1.
11464 (math_opts_dom_walker): New type.
11465 (math_opts_dom_walker::after_dom_children): New method, body moved
11466 here from pass_optimize_widening_mul::execute, added deferring logic
11468 (pass_optimize_widening_mul::execute): Moved most of code to
11469 math_opts_dom_walker::after_dom_children.
11470 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11471 * config/i386/i386.c (ix86_option_override_internal): Added
11472 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11474 2018-01-12 Richard Biener <rguenther@suse.de>
11477 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11478 inline instance vars.
11480 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11483 * config/rx/rx.c (rx_is_restricted_memory_address):
11484 Handle SUBREG case.
11486 2018-01-12 Richard Biener <rguenther@suse.de>
11488 PR tree-optimization/80846
11489 * target.def (split_reduction): New target hook.
11490 * targhooks.c (default_split_reduction): New function.
11491 * targhooks.h (default_split_reduction): Declare.
11492 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11493 target requests first reduce vectors by combining low and high
11495 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11496 (get_vectype_for_scalar_type_and_size): Export.
11497 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11498 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11499 * doc/tm.texi: Regenerate.
11500 * config/i386/i386.c (ix86_split_reduction): Implement
11501 TARGET_VECTORIZE_SPLIT_REDUCTION.
11503 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11506 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11507 in PIC mode except for TARGET_VXWORKS_RTP.
11508 * config/sparc/sparc.c: Include cfgrtl.h.
11509 (TARGET_INIT_PIC_REG): Define.
11510 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11511 (sparc_pic_register_p): New predicate.
11512 (sparc_legitimate_address_p): Use it.
11513 (sparc_legitimize_pic_address): Likewise.
11514 (sparc_delegitimize_address): Likewise.
11515 (sparc_mode_dependent_address_p): Likewise.
11516 (gen_load_pcrel_sym): Remove 4th parameter.
11517 (load_got_register): Adjust call to above. Remove obsolete stuff.
11518 (sparc_expand_prologue): Do not call load_got_register here.
11519 (sparc_flat_expand_prologue): Likewise.
11520 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11521 (sparc_use_pseudo_pic_reg): New function.
11522 (sparc_init_pic_reg): Likewise.
11523 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11524 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11526 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11528 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11529 Add item for branch_cost.
11531 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11533 PR rtl-optimization/83565
11534 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11535 not extend the result to a larger mode for rotate operations.
11536 (num_sign_bit_copies1): Likewise.
11538 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11541 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11543 Use values-Xc.o for -pedantic.
11544 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11546 2018-01-12 Martin Liska <mliska@suse.cz>
11549 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11551 (possible_polymorphic_call_targets): Use it.
11552 (ipa_devirt): Likewise.
11554 2018-01-12 Martin Liska <mliska@suse.cz>
11556 * profile-count.h (enum profile_quality): Use 0 as invalid
11557 enum value of profile_quality.
11559 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11561 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11562 -mext-string options.
11564 2018-01-12 Richard Biener <rguenther@suse.de>
11566 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11567 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11568 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11570 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11572 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11574 * configure.ac (--with-long-double-format): Add support for the
11575 configuration option to change the default long double format on
11577 * config.gcc (powerpc*-linux*-*): Likewise.
11578 * configure: Regenerate.
11579 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11580 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11581 used without modification.
11583 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11585 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11586 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11587 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11588 MISC_BUILTIN_SPEC_BARRIER.
11589 (rs6000_init_builtins): Likewise.
11590 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11592 (speculation_barrier): New define_insn.
11593 * doc/extend.texi: Document __builtin_speculation_barrier.
11595 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11598 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11599 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11600 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11602 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11603 integral modes instead of "ss" and "sd".
11604 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11605 vectors with 32-bit and 64-bit elements.
11606 (vecdupssescalarmodesuffix): New mode attribute.
11607 (vec_dup<mode>): Use it.
11609 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11612 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11613 frame if argument is passed on stack.
11615 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11618 * ree.c (combine_reaching_defs): Optimize also
11619 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11620 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11622 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11624 PR middle-end/83189
11625 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11627 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11629 PR middle-end/83718
11630 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11631 after they are computed.
11633 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11635 PR tree-optimization/83695
11636 * gimple-loop-linterchange.cc
11637 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11638 reset cached scev information after interchange.
11639 (pass_linterchange::execute): Remove call to scev_reset_htab.
11641 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11643 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11644 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11645 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11646 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11647 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11648 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11649 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11650 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11651 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11652 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11653 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11654 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11655 (V_lane_reg): Likewise.
11656 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11658 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11659 (vfmal_lane_low<mode>_intrinsic,
11660 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11661 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11662 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11663 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11664 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11665 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11667 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11669 * config/arm/arm-cpus.in (fp16fml): New feature.
11670 (ALL_SIMD): Add fp16fml.
11671 (armv8.2-a): Add fp16fml as an option.
11672 (armv8.3-a): Likewise.
11673 (armv8.4-a): Add fp16fml as part of fp16.
11674 * config/arm/arm.h (TARGET_FP16FML): Define.
11675 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11677 * config/arm/arm-modes.def (V2HF): Define.
11678 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11679 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11680 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11681 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11682 vfmsl_low, vfmsl_high): New set of builtins.
11683 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11684 (vfml_op): New code attribute.
11685 (VFMLHALVES): New int iterator.
11686 (VFML, VFMLSEL): New mode attributes.
11687 (V_reg): Define mapping for V2HF.
11688 (V_hi, V_lo): New mode attributes.
11689 (VF_constraint): Likewise.
11690 (vfml_half, vfml_half_selector): New int attributes.
11691 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11693 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11694 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11696 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11697 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11698 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11699 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11701 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11702 Document new effective target and option set.
11704 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11706 * config/arm/arm-cpus.in (armv8_4): New feature.
11707 (ARMv8_4a): New fgroup.
11708 (armv8.4-a): New arch.
11709 * config/arm/arm-tables.opt: Regenerate.
11710 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11711 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11712 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11713 Add matching rules for -march=armv8.4-a and extensions.
11714 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11716 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11719 * config/rx/rx.md (BW): New mode attribute.
11720 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11722 2018-01-11 Richard Biener <rguenther@suse.de>
11724 PR tree-optimization/83435
11725 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11726 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11727 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11729 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11730 Alan Hayward <alan.hayward@arm.com>
11731 David Sherwood <david.sherwood@arm.com>
11733 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11735 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11736 (aarch64_print_address_internal): Use it to check for a zero offset.
11738 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11739 Alan Hayward <alan.hayward@arm.com>
11740 David Sherwood <david.sherwood@arm.com>
11742 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11743 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11744 Return a poly_int64 rather than a HOST_WIDE_INT.
11745 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11746 rather than a HOST_WIDE_INT.
11747 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11748 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11749 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11750 final_offset from HOST_WIDE_INT to poly_int64.
11751 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11752 to_constant when getting the number of units in an Advanced SIMD
11754 (aarch64_builtin_vectorized_function): Check for a constant number
11756 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11758 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11759 attribute instead of GET_MODE_NUNITS.
11760 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11761 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11762 GET_MODE_SIZE for fixed-size registers.
11763 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11764 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11765 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11766 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11767 (aarch64_print_operand, aarch64_print_address_internal)
11768 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11769 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11770 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11771 Handle polynomial GET_MODE_SIZE.
11772 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11773 wider than SImode without modification.
11774 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11775 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11776 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11777 passing and returning SVE modes.
11778 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11779 rather than GEN_INT.
11780 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11781 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11782 (aarch64_allocate_and_probe_stack_space): Likewise.
11783 (aarch64_layout_frame): Cope with polynomial offsets.
11784 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11785 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11786 polynomial offsets.
11787 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11788 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11789 poly_int64 rather than a HOST_WIDE_INT.
11790 (aarch64_get_separate_components, aarch64_process_components)
11791 (aarch64_expand_prologue, aarch64_expand_epilogue)
11792 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11793 (aarch64_anchor_offset): New function, split out from...
11794 (aarch64_legitimize_address): ...here.
11795 (aarch64_builtin_vectorization_cost): Handle polynomial
11796 TYPE_VECTOR_SUBPARTS.
11797 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11799 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11800 number of elements from the PARALLEL rather than the mode.
11801 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11802 rather than GET_MODE_BITSIZE.
11803 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11804 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11805 (aarch64_expand_vec_perm_const_1): Handle polynomial
11806 d->perm.length () and d->perm elements.
11807 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11808 Apply to_constant to d->perm elements.
11809 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11810 polynomial CONST_VECTOR_NUNITS.
11811 (aarch64_move_pointer): Take amount as a poly_int64 rather
11813 (aarch64_progress_pointer): Avoid temporary variable.
11814 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11815 the mode attribute instead of GET_MODE.
11817 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11818 Alan Hayward <alan.hayward@arm.com>
11819 David Sherwood <david.sherwood@arm.com>
11821 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11822 x exists before using it.
11823 (aarch64_add_constant_internal): Rename to...
11824 (aarch64_add_offset_1): ...this. Replace regnum with separate
11825 src and dest rtxes. Handle the case in which they're different,
11826 including when the offset is zero. Replace scratchreg with an rtx.
11827 Use 2 additions if there is no spare register into which we can
11828 move a 16-bit constant.
11829 (aarch64_add_constant): Delete.
11830 (aarch64_add_offset): Replace reg with separate src and dest
11831 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11832 Use aarch64_add_offset_1.
11833 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11834 an rtx rather than an int. Take the delta as a poly_int64
11835 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11836 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11837 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11838 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11839 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11840 and aarch64_add_sp.
11841 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11842 aarch64_add_constant.
11844 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11846 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11847 Use scalar_float_mode.
11849 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11851 * config/aarch64/aarch64-simd.md
11852 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11853 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11854 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11855 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11856 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11857 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11858 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11859 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11860 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11861 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11863 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11866 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11867 targ_options->x_arm_arch_string is non NULL.
11869 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11871 * config/aarch64/aarch64.h
11872 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11874 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11877 * expmed.c (emit_store_flag_force): Swap if const op0
11878 and change VOIDmode to mode of op0.
11880 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11882 PR rtl-optimization/83761
11883 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11884 than bytes to mode_for_size.
11886 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11888 PR middle-end/83189
11889 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11890 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11893 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11895 PR middle-end/83575
11896 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11897 when in layout mode.
11898 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11899 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11902 2018-01-10 Michael Collison <michael.collison@arm.com>
11904 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11905 * config/aarch64/aarch64-option-extension.def: Add
11906 AARCH64_OPT_EXTENSION of 'fp16fml'.
11907 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11908 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11909 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11910 * config/aarch64/constraints.md (Ui7): New constraint.
11911 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11912 (VFMLA_SEL_W): Ditto.
11915 (VFMLA16_LOW): New int iterator.
11916 (VFMLA16_HIGH): Ditto.
11917 (UNSPEC_FMLAL): New unspec.
11918 (UNSPEC_FMLSL): Ditto.
11919 (UNSPEC_FMLAL2): Ditto.
11920 (UNSPEC_FMLSL2): Ditto.
11921 (f16mac): New code attribute.
11922 * config/aarch64/aarch64-simd-builtins.def
11923 (aarch64_fmlal_lowv2sf): Ditto.
11924 (aarch64_fmlsl_lowv2sf): Ditto.
11925 (aarch64_fmlalq_lowv4sf): Ditto.
11926 (aarch64_fmlslq_lowv4sf): Ditto.
11927 (aarch64_fmlal_highv2sf): Ditto.
11928 (aarch64_fmlsl_highv2sf): Ditto.
11929 (aarch64_fmlalq_highv4sf): Ditto.
11930 (aarch64_fmlslq_highv4sf): Ditto.
11931 (aarch64_fmlal_lane_lowv2sf): Ditto.
11932 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11933 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11934 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11935 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11936 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11937 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11938 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11939 (aarch64_fmlal_lane_highv2sf): Ditto.
11940 (aarch64_fmlsl_lane_highv2sf): Ditto.
11941 (aarch64_fmlal_laneq_highv2sf): Ditto.
11942 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11943 (aarch64_fmlalq_lane_highv4sf): Ditto.
11944 (aarch64_fmlsl_lane_highv4sf): Ditto.
11945 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11946 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11947 * config/aarch64/aarch64-simd.md:
11948 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11949 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11950 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11951 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11952 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11953 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11954 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11955 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11956 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11957 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11958 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11959 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11960 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11961 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11962 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11963 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11964 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11965 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11966 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11967 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11968 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11969 (vfmlsl_low_u32): Ditto.
11970 (vfmlalq_low_u32): Ditto.
11971 (vfmlslq_low_u32): Ditto.
11972 (vfmlal_high_u32): Ditto.
11973 (vfmlsl_high_u32): Ditto.
11974 (vfmlalq_high_u32): Ditto.
11975 (vfmlslq_high_u32): Ditto.
11976 (vfmlal_lane_low_u32): Ditto.
11977 (vfmlsl_lane_low_u32): Ditto.
11978 (vfmlal_laneq_low_u32): Ditto.
11979 (vfmlsl_laneq_low_u32): Ditto.
11980 (vfmlalq_lane_low_u32): Ditto.
11981 (vfmlslq_lane_low_u32): Ditto.
11982 (vfmlalq_laneq_low_u32): Ditto.
11983 (vfmlslq_laneq_low_u32): Ditto.
11984 (vfmlal_lane_high_u32): Ditto.
11985 (vfmlsl_lane_high_u32): Ditto.
11986 (vfmlal_laneq_high_u32): Ditto.
11987 (vfmlsl_laneq_high_u32): Ditto.
11988 (vfmlalq_lane_high_u32): Ditto.
11989 (vfmlslq_lane_high_u32): Ditto.
11990 (vfmlalq_laneq_high_u32): Ditto.
11991 (vfmlslq_laneq_high_u32): Ditto.
11992 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11993 (AARCH64_FL_FOR_ARCH8_4): New.
11994 (AARCH64_ISA_F16FML): New ISA flag.
11995 (TARGET_F16FML): New feature flag for fp16fml.
11996 (doc/invoke.texi): Document new fp16fml option.
11998 2018-01-10 Michael Collison <michael.collison@arm.com>
12000 * config/aarch64/aarch64-builtins.c:
12001 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12002 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12003 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12004 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12005 (AARCH64_ISA_SHA3): New ISA flag.
12006 (TARGET_SHA3): New feature flag for sha3.
12007 * config/aarch64/iterators.md (sha512_op): New int attribute.
12008 (CRYPTO_SHA512): New int iterator.
12009 (UNSPEC_SHA512H): New unspec.
12010 (UNSPEC_SHA512H2): Ditto.
12011 (UNSPEC_SHA512SU0): Ditto.
12012 (UNSPEC_SHA512SU1): Ditto.
12013 * config/aarch64/aarch64-simd-builtins.def
12014 (aarch64_crypto_sha512hqv2di): New builtin.
12015 (aarch64_crypto_sha512h2qv2di): Ditto.
12016 (aarch64_crypto_sha512su0qv2di): Ditto.
12017 (aarch64_crypto_sha512su1qv2di): Ditto.
12018 (aarch64_eor3qv8hi): Ditto.
12019 (aarch64_rax1qv2di): Ditto.
12020 (aarch64_xarqv2di): Ditto.
12021 (aarch64_bcaxqv8hi): Ditto.
12022 * config/aarch64/aarch64-simd.md:
12023 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12024 (aarch64_crypto_sha512su0qv2di): Ditto.
12025 (aarch64_crypto_sha512su1qv2di): Ditto.
12026 (aarch64_eor3qv8hi): Ditto.
12027 (aarch64_rax1qv2di): Ditto.
12028 (aarch64_xarqv2di): Ditto.
12029 (aarch64_bcaxqv8hi): Ditto.
12030 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12031 (vsha512h2q_u64): Ditto.
12032 (vsha512su0q_u64): Ditto.
12033 (vsha512su1q_u64): Ditto.
12034 (veor3q_u16): Ditto.
12035 (vrax1q_u64): Ditto.
12036 (vxarq_u64): Ditto.
12037 (vbcaxq_u16): Ditto.
12038 * config/arm/types.md (crypto_sha512): New type attribute.
12039 (crypto_sha3): Ditto.
12040 (doc/invoke.texi): Document new sha3 option.
12042 2018-01-10 Michael Collison <michael.collison@arm.com>
12044 * config/aarch64/aarch64-builtins.c:
12045 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12046 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12047 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12048 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12049 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12050 (AARCH64_ISA_SM4): New ISA flag.
12051 (TARGET_SM4): New feature flag for sm4.
12052 * config/aarch64/aarch64-simd-builtins.def
12053 (aarch64_sm3ss1qv4si): Ditto.
12054 (aarch64_sm3tt1aq4si): Ditto.
12055 (aarch64_sm3tt1bq4si): Ditto.
12056 (aarch64_sm3tt2aq4si): Ditto.
12057 (aarch64_sm3tt2bq4si): Ditto.
12058 (aarch64_sm3partw1qv4si): Ditto.
12059 (aarch64_sm3partw2qv4si): Ditto.
12060 (aarch64_sm4eqv4si): Ditto.
12061 (aarch64_sm4ekeyqv4si): Ditto.
12062 * config/aarch64/aarch64-simd.md:
12063 (aarch64_sm3ss1qv4si): Ditto.
12064 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12065 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12066 (aarch64_sm4eqv4si): Ditto.
12067 (aarch64_sm4ekeyqv4si): Ditto.
12068 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12069 (sm3part_op): Ditto.
12070 (CRYPTO_SM3TT): Ditto.
12071 (CRYPTO_SM3PART): Ditto.
12072 (UNSPEC_SM3SS1): New unspec.
12073 (UNSPEC_SM3TT1A): Ditto.
12074 (UNSPEC_SM3TT1B): Ditto.
12075 (UNSPEC_SM3TT2A): Ditto.
12076 (UNSPEC_SM3TT2B): Ditto.
12077 (UNSPEC_SM3PARTW1): Ditto.
12078 (UNSPEC_SM3PARTW2): Ditto.
12079 (UNSPEC_SM4E): Ditto.
12080 (UNSPEC_SM4EKEY): Ditto.
12081 * config/aarch64/constraints.md (Ui2): New constraint.
12082 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12083 * config/arm/types.md (crypto_sm3): New type attribute.
12084 (crypto_sm4): Ditto.
12085 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12086 (vsm3tt1aq_u32): Ditto.
12087 (vsm3tt1bq_u32): Ditto.
12088 (vsm3tt2aq_u32): Ditto.
12089 (vsm3tt2bq_u32): Ditto.
12090 (vsm3partw1q_u32): Ditto.
12091 (vsm3partw2q_u32): Ditto.
12092 (vsm4eq_u32): Ditto.
12093 (vsm4ekeyq_u32): Ditto.
12094 (doc/invoke.texi): Document new sm4 option.
12096 2018-01-10 Michael Collison <michael.collison@arm.com>
12098 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12099 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12100 (AARCH64_FL_FOR_ARCH8_4): New.
12101 (AARCH64_FL_V8_4): New flag.
12102 (doc/invoke.texi): Document new armv8.4-a option.
12104 2018-01-10 Michael Collison <michael.collison@arm.com>
12106 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12107 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12108 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12109 * config/aarch64/aarch64-option-extension.def: Add
12110 AARCH64_OPT_EXTENSION of 'sha2'.
12111 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12112 (crypto): Disable sha2 and aes if crypto disabled.
12113 (crypto): Enable aes and sha2 if enabled.
12114 (simd): Disable sha2 and aes if simd disabled.
12115 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12117 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12118 (TARGET_SHA2): New feature flag for sha2.
12119 (TARGET_AES): New feature flag for aes.
12120 * config/aarch64/aarch64-simd.md:
12121 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12122 conditional on TARGET_AES.
12123 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12124 (aarch64_crypto_sha1hsi): Make pattern conditional
12126 (aarch64_crypto_sha1hv4si): Ditto.
12127 (aarch64_be_crypto_sha1hv4si): Ditto.
12128 (aarch64_crypto_sha1su1v4si): Ditto.
12129 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12130 (aarch64_crypto_sha1su0v4si): Ditto.
12131 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12132 (aarch64_crypto_sha256su0v4si): Ditto.
12133 (aarch64_crypto_sha256su1v4si): Ditto.
12134 (doc/invoke.texi): Document new aes and sha2 options.
12136 2018-01-10 Martin Sebor <msebor@redhat.com>
12138 PR tree-optimization/83781
12139 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12142 2018-01-11 Martin Sebor <msebor@gmail.com>
12143 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12145 PR tree-optimization/83501
12146 PR tree-optimization/81703
12148 * tree-ssa-strlen.c (get_string_cst): Rename...
12149 (get_string_len): ...to this. Handle global constants.
12150 (handle_char_store): Adjust.
12152 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12153 Jim Wilson <jimw@sifive.com>
12155 * config/riscv/riscv-protos.h (riscv_output_return): New.
12156 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12157 (riscv_attribute_table, riscv_output_return),
12158 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12159 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12160 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12162 (riscv_expand_prologue): Add early return for naked function.
12163 (riscv_expand_epilogue): Likewise.
12164 (riscv_function_ok_for_sibcall): Return false for naked function.
12165 (riscv_set_current_function): New.
12166 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12167 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12168 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12169 * doc/extend.texi (RISC-V Function Attributes): New.
12171 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12173 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12174 check for 128-bit long double before checking TCmode.
12175 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12176 128-bit long doubles before checking TFmode or TCmode.
12177 (FLOAT128_IBM_P): Likewise.
12179 2018-01-10 Martin Sebor <msebor@redhat.com>
12181 PR tree-optimization/83671
12182 * builtins.c (c_strlen): Unconditionally return zero for the empty
12184 Use -Warray-bounds for warnings.
12185 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12186 for non-constant array indices with COMPONENT_REF, arrays of
12187 arrays, and pointers to arrays.
12188 (gimple_fold_builtin_strlen): Determine and set length range for
12189 non-constant character arrays.
12191 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12193 PR middle-end/81897
12194 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12197 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12199 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12201 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12204 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12205 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12206 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12207 indexed_or_indirect_operand predicate.
12208 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12209 (*vsx_le_perm_load_v8hi): Likewise.
12210 (*vsx_le_perm_load_v16qi): Likewise.
12211 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12212 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12213 (*vsx_le_perm_store_v8hi): Likewise.
12214 (*vsx_le_perm_store_v16qi): Likewise.
12215 (eight unnamed splitters): Likewise.
12217 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12219 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12220 * config/rs6000/emmintrin.h: Likewise.
12221 * config/rs6000/mmintrin.h: Likewise.
12222 * config/rs6000/xmmintrin.h: Likewise.
12224 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12227 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12229 * tree.c (tree_nop_conversion): Return true for location wrapper
12231 (maybe_wrap_with_location): New function.
12232 (selftest::check_strip_nops): New function.
12233 (selftest::test_location_wrappers): New function.
12234 (selftest::tree_c_tests): Call it.
12235 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12236 (maybe_wrap_with_location): New decl.
12237 (EXPR_LOCATION_WRAPPER_P): New macro.
12238 (location_wrapper_p): New inline function.
12239 (tree_strip_any_location_wrapper): New inline function.
12241 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12244 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12245 stack_realign_offset for the largest alignment of stack slot
12247 (ix86_find_max_used_stack_alignment): New function.
12248 (ix86_finalize_stack_frame_flags): Use it. Set
12249 max_used_stack_alignment if we don't realign stack.
12250 * config/i386/i386.h (machine_function): Add
12251 max_used_stack_alignment.
12253 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12255 * config/arm/arm.opt (-mbranch-cost): New option.
12256 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12259 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12262 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12263 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12265 2018-01-10 Richard Biener <rguenther@suse.de>
12268 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12269 early out so it also covers the case where we have a non-NULL
12272 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12274 PR tree-optimization/83753
12275 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12276 for non-strided grouped accesses if the number of elements is 1.
12278 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12281 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12282 * i386.h (TARGET_USE_GATHER): Define.
12283 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12285 2018-01-10 Martin Liska <mliska@suse.cz>
12288 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12289 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12291 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12292 CLEANUP_NO_PARTITIONING is not set.
12294 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12296 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12297 for vectors, as a partial revert of r254296.
12298 * rtl.h (const_vec_p): Delete.
12299 (const_vec_duplicate_p): Don't test for vector CONSTs.
12300 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12301 * expmed.c (make_tree): Likewise.
12304 * common.md (E, F): Use CONSTANT_P instead of checking for
12306 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12307 checking for CONST_VECTOR.
12309 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12311 PR middle-end/83575
12312 * predict.c (force_edge_cold): Handle in more sane way edges
12313 with no prediction.
12315 2018-01-09 Carl Love <cel@us.ibm.com>
12317 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12319 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12320 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12321 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12322 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12323 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12324 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12325 * config/rs6000/rs6000-protos.h: Add extern defition for
12326 rs6000_generate_float2_double_code.
12327 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12329 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12330 (float2_v2df): Add define_expand.
12332 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12335 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12336 op_mode in the force_to_mode call.
12338 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12340 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12341 instead of checking each element individually.
12342 (aarch64_evpc_uzp): Likewise.
12343 (aarch64_evpc_zip): Likewise.
12344 (aarch64_evpc_ext): Likewise.
12345 (aarch64_evpc_rev): Likewise.
12346 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12347 instead of checking each element individually. Return true without
12349 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12350 whether all selected elements come from the same input, instead of
12351 checking each element individually. Remove calls to gen_rtx_REG,
12352 start_sequence and end_sequence and instead assert that no rtl is
12355 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12357 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12358 order of HIGH and CONST checks.
12360 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12362 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12363 if the destination isn't an SSA_NAME.
12365 2018-01-09 Richard Biener <rguenther@suse.de>
12367 PR tree-optimization/83668
12368 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12370 (canonicalize_loop_form): ... here, renamed from ...
12371 (canonicalize_loop_closed_ssa_form): ... this and amended to
12372 swap successor edges for loop exit blocks to make us use
12373 the RPO order we need for initial schedule generation.
12375 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12377 PR tree-optimization/64811
12378 * match.pd: When optimizing comparisons with Inf, avoid
12379 introducing or losing exceptions from comparisons with NaN.
12381 2018-01-09 Martin Liska <mliska@suse.cz>
12384 * asan.c (shadow_mem_size): Add gcc_assert.
12386 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12388 Don't save registers in main().
12391 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12392 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12393 * config/avr/avr.c (avr_set_current_function): Don't error if
12394 naked, OS_task or OS_main are specified at the same time.
12395 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12397 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12399 * common/config/avr/avr-common.c (avr_option_optimization_table):
12400 Switch on -mmain-is-OS_task for optimizing compilations.
12402 2018-01-09 Richard Biener <rguenther@suse.de>
12404 PR tree-optimization/83572
12405 * graphite.c: Include cfganal.h.
12406 (graphite_transform_loops): Connect infinite loops to exit
12407 and remove fake edges at the end.
12409 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12411 * ipa-inline.c (edge_badness): Revert accidental checkin.
12413 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12416 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12417 symbols; not inline clones.
12419 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12422 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12423 hard registers. Formatting fixes.
12425 PR preprocessor/83722
12426 * gcc.c (try_generate_repro): Pass
12427 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12428 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12431 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12432 Kito Cheng <kito.cheng@gmail.com>
12434 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12435 (riscv_leaf_function_p): Delete.
12436 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12438 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12440 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12442 (do_ifelse): New function.
12443 (do_isel): New function.
12444 (do_sub3): New function.
12445 (do_add3): New function.
12446 (do_load_mask_compare): New function.
12447 (do_overlap_load_compare): New function.
12448 (expand_compare_loop): New function.
12449 (expand_block_compare): Call expand_compare_loop() when appropriate.
12450 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12451 option description.
12452 (-mblock-compare-inline-loop-limit): New option.
12454 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12457 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12458 Reverse order of second and third operands in first alternative.
12459 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12460 of first and second elements in UNSPEC_VPERMR vector.
12461 (altivec_expand_vec_perm_le): Likewise.
12463 2018-01-08 Jeff Law <law@redhat.com>
12465 PR rtl-optimizatin/81308
12466 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12467 (process_switch): If group_case_labels makes a change, then set
12469 (pass_convert_switch::execute): If a switch is converted, then
12470 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12472 PR rtl-optimization/81308
12473 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12476 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12478 PR target/83663 - Revert r255946
12479 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12480 generation for cases where splatting a value is not useful.
12481 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12482 across a vec_duplicate and a paradoxical subreg forming a vector
12483 mode to a vec_concat.
12485 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12487 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12488 -march=armv8.3-a variants.
12489 * config/arm/t-multilib: Likewise.
12490 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12492 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12494 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12496 (cceq_ior_compare_complement): Give it a name so I can use it, and
12497 change boolean_or_operator predicate to boolean_operator so it can
12498 be used to generate a crand.
12499 (eqne): New code iterator.
12500 (bd/bd_neg): New code_attrs.
12501 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12502 a single define_insn.
12503 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12504 decrement (bdnzt/bdnzf/bdzt/bdzf).
12505 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12506 with the new names of the branch decrement patterns, and added the
12507 names of the branch decrement conditional patterns.
12509 2018-01-08 Richard Biener <rguenther@suse.de>
12511 PR tree-optimization/83563
12512 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12515 2018-01-08 Richard Biener <rguenther@suse.de>
12517 PR middle-end/83713
12518 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12520 2018-01-08 Richard Biener <rguenther@suse.de>
12522 PR tree-optimization/83685
12523 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12524 references to abnormals.
12526 2018-01-08 Richard Biener <rguenther@suse.de>
12529 * dwarf2out.c (output_indirect_strings): Handle empty
12530 skeleton_debug_str_hash.
12531 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12533 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12535 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12536 (emit_store_direct): Likewise.
12537 (arc_trampoline_adjust_address): Likewise.
12538 (arc_asm_trampoline_template): New function.
12539 (arc_initialize_trampoline): Use asm_trampoline_template.
12540 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12541 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12542 * config/arc/arc.md (flush_icache): Delete pattern.
12544 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12546 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12547 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12550 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12553 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12554 by not USED_FOR_TARGET.
12555 (make_pass_resolve_sw_modes): Likewise.
12557 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12559 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12562 2018-01-08 Richard Biener <rguenther@suse.de>
12564 PR middle-end/83580
12565 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12567 2018-01-08 Richard Biener <rguenther@suse.de>
12569 PR middle-end/83517
12570 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12572 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12574 PR middle-end/81897
12575 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12576 basic blocks with a small number of successors.
12577 (convert_control_dep_chain_into_preds): Improve handling of
12579 (dump_predicates): Split apart into...
12580 (dump_pred_chain): ...here...
12581 (dump_pred_info): ...and here.
12582 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12583 (can_chain_union_be_invalidated_p): Improve check for invalidation
12585 (uninit_uses_cannot_happen): Avoid unnecessary if
12586 convert_control_dep_chain_into_preds yielded nothing.
12588 2018-01-06 Martin Sebor <msebor@redhat.com>
12590 PR tree-optimization/83640
12591 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12592 subtracting negative offset from size.
12593 (builtin_access::overlap): Adjust offset bounds of the access to fall
12594 within the size of the object if possible.
12596 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12598 PR rtl-optimization/83699
12599 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12600 extract_bit_field_as_subreg to cases in which the extracted
12601 value is also a vector.
12603 * lra-constraints.c (process_alt_operands): Test for the equivalence
12604 substitutions when detecting a possible reload cycle.
12606 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12609 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12610 by default if flag_selective_schedling{,2}. Formatting fixes.
12612 PR rtl-optimization/83682
12613 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12614 if it has non-VECTOR_MODE element mode.
12615 (vec_duplicate_p): Likewise.
12617 PR middle-end/83694
12618 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12619 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12621 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12624 * config/i386/i386-builtin.def
12625 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12626 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12627 Require also OPTION_MASK_ISA_AVX512F in addition to
12628 OPTION_MASK_ISA_GFNI.
12629 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12630 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12631 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12632 to OPTION_MASK_ISA_GFNI.
12633 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12634 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12635 OPTION_MASK_ISA_AVX512BW.
12636 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12637 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12638 addition to OPTION_MASK_ISA_GFNI.
12639 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12640 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12641 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12642 to OPTION_MASK_ISA_GFNI.
12643 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12644 a requirement for all ISAs rather than any of them with a few
12646 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12648 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12649 bitmasks to be enabled with 3 exceptions, instead of requiring any
12650 enabled ISA with lots of exceptions.
12651 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12652 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12653 Change avx512bw in isa attribute to avx512f.
12654 * config/i386/sgxintrin.h: Add license boilerplate.
12655 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12656 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12657 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12658 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12660 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12661 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12662 temporarily sse2 rather than sse if not enabled already.
12665 * config/i386/sse.md (VI248_VLBW): Rename to ...
12666 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12667 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12668 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12669 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12670 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12671 mode iterator instead of VI248_VLBW.
12673 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12675 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12676 (record_modified): Skip clobbers; add debug output.
12677 (param_change_prob): Use sreal frequencies.
12679 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12681 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12682 punt for user-aligned variables.
12684 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12686 * tree-chrec.c (chrec_contains_symbols): Return true for
12689 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12692 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12693 of (x|y) == x for BICS pattern.
12695 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12697 PR tree-optimization/83605
12698 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12699 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12702 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12704 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12705 * config/epiphany/rtems.h: New file.
12707 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12708 Uros Bizjak <ubizjak@gmail.com>
12711 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12712 QIreg_operand instead of register_operand predicate.
12713 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12714 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12715 comments instead of -fmitigate[-_]rop.
12717 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12720 * cgraphunit.c (symbol_table::compile): Switch to text_section
12721 before calling assembly_start debug hook.
12722 * run-rtl-passes.c (run_rtl_passes): Likewise.
12725 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12727 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12728 range_int_cst_p rather than !symbolic_range_p before calling
12729 extract_range_from_multiplicative_op_1.
12731 2018-01-04 Jeff Law <law@redhat.com>
12733 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12734 redundant test in assertion.
12736 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12738 * doc/rtl.texi: Document machine_mode wrapper classes.
12740 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12742 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12743 using tree_to_uhwi.
12745 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12747 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12748 the VEC_PERM_EXPR fold to fail.
12750 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12753 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12754 to switched_sections.
12756 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12759 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12760 test for d.testing.
12762 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12765 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12766 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12768 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12771 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12772 is BLKmode and bitpos not zero or mode change is needed.
12774 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12777 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12780 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12783 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12784 instead of MULT rtx. Update all corresponding splitters.
12785 (*saddl_se): Ditto.
12786 (*ssub<modesuffix>): Ditto.
12787 (*ssubl_se): Ditto.
12788 (*cmp_sadd_di): Update split patterns.
12789 (*cmp_sadd_si): Ditto.
12790 (*cmp_sadd_sidi): Ditto.
12791 (*cmp_ssub_di): Ditto.
12792 (*cmp_ssub_si): Ditto.
12793 (*cmp_ssub_sidi): Ditto.
12794 * config/alpha/predicates.md (const23_operand): New predicate.
12795 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12796 Look for ASHIFT, not MULT inner operand.
12797 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12799 2018-01-04 Martin Liska <mliska@suse.cz>
12801 PR gcov-profile/83669
12802 * gcov.c (output_intermediate_file): Add version to intermediate
12804 * doc/gcov.texi: Document new field 'version' in intermediate
12805 file format. Fix location of '-k' option of gcov command.
12807 2018-01-04 Martin Liska <mliska@suse.cz>
12810 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12812 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12814 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12816 2018-01-03 Martin Sebor <msebor@redhat.com>
12818 PR tree-optimization/83655
12819 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12820 checking calls with invalid arguments.
12822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12824 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12825 (vectorizable_mask_load_store): Delete.
12826 (vectorizable_call): Return false for masked loads and stores.
12827 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12828 instead of gimple_assign_rhs1.
12829 (vectorizable_load): Handle IFN_MASK_LOAD.
12830 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12834 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12836 (vectorizable_mask_load_store): ...here.
12837 (vectorizable_load): ...and here.
12839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12841 * tree-vect-stmts.c (vect_build_all_ones_mask)
12842 (vect_build_zero_merge_argument): New functions, split out from...
12843 (vectorizable_load): ...here.
12845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12847 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12849 (vectorizable_mask_load_store): ...here.
12850 (vectorizable_store): ...and here.
12852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12854 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12856 (vectorizable_mask_load_store): ...here.
12858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12860 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12861 (vect_model_store_cost): Take a vec_load_store_type instead of a
12863 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12864 (vect_model_store_cost): Take a vec_load_store_type instead of a
12866 (vectorizable_mask_load_store): Update accordingly.
12867 (vectorizable_store): Likewise.
12868 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12872 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12873 IFN_MASK_LOAD calls here rather than...
12874 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12877 Alan Hayward <alan.hayward@arm.com>
12878 David Sherwood <david.sherwood@arm.com>
12880 * expmed.c (extract_bit_field_1): For vector extracts,
12881 fall back to extract_bit_field_as_subreg if vec_extract
12884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12885 Alan Hayward <alan.hayward@arm.com>
12886 David Sherwood <david.sherwood@arm.com>
12888 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12889 they are variable or constant sized.
12890 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12891 slots for constant-sized data.
12893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12894 Alan Hayward <alan.hayward@arm.com>
12895 David Sherwood <david.sherwood@arm.com>
12897 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12898 handling COND_EXPRs with boolean comparisons, try to find a better
12899 basis for the mask type than the boolean itself.
12901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12903 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12904 is calculated and how it can be overridden.
12905 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12906 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12908 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12912 Alan Hayward <alan.hayward@arm.com>
12913 David Sherwood <david.sherwood@arm.com>
12915 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12916 Remove the mode argument.
12917 (aarch64_simd_valid_immediate): Remove the mode and inverse
12919 * config/aarch64/iterators.md (bitsize): New iterator.
12920 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12921 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12922 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12923 aarch64_simd_valid_immediate.
12924 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12925 (aarch64_reg_or_bic_imm): Likewise.
12926 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12927 with an insn_type enum and msl with a modifier_type enum.
12928 Replace element_width with a scalar_mode. Change the shift
12929 to unsigned int. Add constructors for scalar_float_mode and
12930 scalar_int_mode elements.
12931 (aarch64_vect_float_const_representable_p): Delete.
12932 (aarch64_can_const_movi_rtx_p)
12933 (aarch64_simd_scalar_immediate_valid_for_move)
12934 (aarch64_simd_make_constant): Update call to
12935 aarch64_simd_valid_immediate.
12936 (aarch64_advsimd_valid_immediate_hs): New function.
12937 (aarch64_advsimd_valid_immediate): Likewise.
12938 (aarch64_simd_valid_immediate): Remove mode and inverse
12939 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12940 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12941 and aarch64_float_const_representable_p on the result.
12942 (aarch64_output_simd_mov_immediate): Remove mode argument.
12943 Update call to aarch64_simd_valid_immediate and use of
12944 simd_immediate_info.
12945 (aarch64_output_scalar_simd_mov_immediate): Update call
12948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12949 Alan Hayward <alan.hayward@arm.com>
12950 David Sherwood <david.sherwood@arm.com>
12952 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12953 (mode_nunits): Likewise CONST_MODE_NUNITS.
12954 * machmode.def (ADJUST_NUNITS): Document.
12955 * genmodes.c (mode_data::need_nunits_adj): New field.
12956 (blank_mode): Update accordingly.
12957 (adj_nunits): New variable.
12958 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12960 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12961 listed in adj_nunits.
12962 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12963 listed in adj_nunits. Don't emit case statements for such modes.
12964 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12965 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12966 nothing if adj_nunits is nonnull.
12967 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12968 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12969 (emit_mode_fbit): Update use of print_maybe_const_decl.
12970 (emit_move_size): Likewise. Treat the array as non-const
12972 (emit_mode_adjustments): Handle adj_nunits.
12974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12976 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12977 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12978 (VECTOR_MODES): Use it.
12979 (make_vector_modes): Take the prefix as an argument.
12981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12982 Alan Hayward <alan.hayward@arm.com>
12983 David Sherwood <david.sherwood@arm.com>
12985 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12986 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12987 for MODE_VECTOR_BOOL.
12988 * machmode.def (VECTOR_BOOL_MODE): Document.
12989 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12990 (make_vector_bool_mode): New function.
12991 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12993 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12994 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12996 * stor-layout.c (int_mode_for_mode): Likewise.
12997 * tree.c (build_vector_type_for_mode): Likewise.
12998 * varasm.c (output_constant_pool_2): Likewise.
12999 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13000 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13001 for MODE_VECTOR_BOOL.
13002 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13003 of mode class checks.
13004 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13005 instead of a list of mode class checks.
13006 (expand_vector_scalar_condition): Likewise.
13007 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13010 Alan Hayward <alan.hayward@arm.com>
13011 David Sherwood <david.sherwood@arm.com>
13013 * machmode.h (mode_size): Change from unsigned short to
13015 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13016 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13017 or if measurement_type is not polynomial.
13018 (fixed_size_mode::includes_p): Check for constant-sized modes.
13019 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13020 return a poly_uint16 rather than an unsigned short.
13021 (emit_mode_size): Change the type of mode_size from unsigned short
13022 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13023 (emit_mode_adjustments): Cope with polynomial vector sizes.
13024 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13026 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13028 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13029 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13030 * caller-save.c (setup_save_areas): Likewise.
13031 (replace_reg_with_saved_mem): Likewise.
13032 * calls.c (emit_library_call_value_1): Likewise.
13033 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13034 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13035 (gen_lowpart_for_combine): Likewise.
13036 * convert.c (convert_to_integer_1): Likewise.
13037 * cse.c (equiv_constant, cse_insn): Likewise.
13038 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13039 (cselib_subst_to_values): Likewise.
13040 * dce.c (word_dce_process_block): Likewise.
13041 * df-problems.c (df_word_lr_mark_ref): Likewise.
13042 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13043 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13044 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13045 (rtl_for_decl_location): Likewise.
13046 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13047 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13048 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13049 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13050 (expand_expr_real_1): Likewise.
13051 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13052 (pad_below): Likewise.
13053 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13054 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13055 * ira.c (get_subreg_tracking_sizes): Likewise.
13056 * ira-build.c (ira_create_allocno_objects): Likewise.
13057 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13058 (ira_sort_regnos_for_alter_reg): Likewise.
13059 * ira-costs.c (record_operand_costs): Likewise.
13060 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13061 (resolve_simple_move): Likewise.
13062 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13063 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13064 (lra_constraints): Likewise.
13065 (CONST_POOL_OK_P): Reject variable-sized modes.
13066 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13067 (add_pseudo_to_slot, lra_spill): Likewise.
13068 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13069 * optabs-query.c (get_best_extraction_insn): Likewise.
13070 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13071 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13072 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13073 * recog.c (offsettable_address_addr_space_p): Likewise.
13074 * regcprop.c (maybe_mode_change): Likewise.
13075 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13076 * regrename.c (build_def_use): Likewise.
13077 * regstat.c (dump_reg_info): Likewise.
13078 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13079 (find_reloads, find_reloads_subreg_address): Likewise.
13080 * reload1.c (eliminate_regs_1): Likewise.
13081 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13082 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13083 (simplify_binary_operation_1, simplify_subreg): Likewise.
13084 * targhooks.c (default_function_arg_padding): Likewise.
13085 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13086 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13087 (verify_gimple_assign_ternary): Likewise.
13088 * tree-inline.c (estimate_move_cost): Likewise.
13089 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13090 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13091 (get_address_cost_ainc): Likewise.
13092 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13093 (vect_supportable_dr_alignment): Likewise.
13094 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13095 (vectorizable_reduction): Likewise.
13096 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13097 (vectorizable_operation, vectorizable_load): Likewise.
13098 * tree.c (build_same_sized_truth_vector_type): Likewise.
13099 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13100 * var-tracking.c (emit_note_insn_var_location): Likewise.
13101 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13102 (ADDR_VEC_ALIGN): Likewise.
13104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13105 Alan Hayward <alan.hayward@arm.com>
13106 David Sherwood <david.sherwood@arm.com>
13108 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13110 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13111 or if measurement_type is polynomial.
13112 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13113 * combine.c (make_extraction): Likewise.
13114 * dse.c (find_shift_sequence): Likewise.
13115 * dwarf2out.c (mem_loc_descriptor): Likewise.
13116 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13117 (extract_bit_field, extract_low_bits): Likewise.
13118 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13119 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13120 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13121 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13122 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13123 * reload.c (find_reloads): Likewise.
13124 * reload1.c (alter_reg): Likewise.
13125 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13126 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13127 * tree-if-conv.c (predicate_mem_writes): Likewise.
13128 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13129 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13130 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13131 * valtrack.c (dead_debug_insert_temp): Likewise.
13132 * varasm.c (mergeable_constant_section): Likewise.
13133 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13136 Alan Hayward <alan.hayward@arm.com>
13137 David Sherwood <david.sherwood@arm.com>
13139 * expr.c (expand_assignment): Cope with polynomial mode sizes
13140 when assigning to a CONCAT.
13142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13143 Alan Hayward <alan.hayward@arm.com>
13144 David Sherwood <david.sherwood@arm.com>
13146 * machmode.h (mode_precision): Change from unsigned short to
13148 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13150 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13151 or if measurement_type is not polynomial.
13152 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13153 in which the mode is already known to be a scalar_int_mode.
13154 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13155 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13157 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13158 for GET_MODE_PRECISION.
13159 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13160 for GET_MODE_PRECISION.
13161 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13163 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13164 (expand_field_assignment, make_extraction): Likewise.
13165 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13166 (get_last_value): Likewise.
13167 * convert.c (convert_to_integer_1): Likewise.
13168 * cse.c (cse_insn): Likewise.
13169 * expr.c (expand_expr_real_1): Likewise.
13170 * lra-constraints.c (simplify_operand_subreg): Likewise.
13171 * optabs-query.c (can_atomic_load_p): Likewise.
13172 * optabs.c (expand_atomic_load): Likewise.
13173 (expand_atomic_store): Likewise.
13174 * ree.c (combine_reaching_defs): Likewise.
13175 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13176 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13177 * tree.h (type_has_mode_precision_p): Likewise.
13178 * ubsan.c (instrument_si_overflow): Likewise.
13180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13181 Alan Hayward <alan.hayward@arm.com>
13182 David Sherwood <david.sherwood@arm.com>
13184 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13185 polynomial numbers of units.
13186 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13187 (valid_vector_subparts_p): New function.
13188 (build_vector_type): Remove temporary shim and take the number
13189 of units as a poly_uint64 rather than an int.
13190 (build_opaque_vector_type): Take the number of units as a
13191 poly_uint64 rather than an int.
13192 * tree.c (build_vector_from_ctor): Handle polynomial
13193 TYPE_VECTOR_SUBPARTS.
13194 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13195 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13196 (build_vector_from_val): If the number of units is variable,
13197 use build_vec_duplicate_cst for constant operands and
13198 VEC_DUPLICATE_EXPR otherwise.
13199 (make_vector_type): Remove temporary is_constant ().
13200 (build_vector_type, build_opaque_vector_type): Take the number of
13201 units as a poly_uint64 rather than an int.
13202 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13204 * cfgexpand.c (expand_debug_expr): Likewise.
13205 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13206 (store_constructor, expand_expr_real_1): Likewise.
13207 (const_scalar_mask_from_tree): Likewise.
13208 * fold-const-call.c (fold_const_reduction): Likewise.
13209 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13210 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13211 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13212 (fold_relational_const): Likewise.
13213 (native_interpret_vector): Likewise. Change the size from an
13214 int to an unsigned int.
13215 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13216 TYPE_VECTOR_SUBPARTS.
13217 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13218 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13219 duplicating a non-constant operand into a variable-length vector.
13220 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13221 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13222 * ipa-icf.c (sem_variable::equals): Likewise.
13223 * match.pd: Likewise.
13224 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13225 * print-tree.c (print_node): Likewise.
13226 * stor-layout.c (layout_type): Likewise.
13227 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13228 * tree-cfg.c (verify_gimple_comparison): Likewise.
13229 (verify_gimple_assign_binary): Likewise.
13230 (verify_gimple_assign_ternary): Likewise.
13231 (verify_gimple_assign_single): Likewise.
13232 * tree-pretty-print.c (dump_generic_node): Likewise.
13233 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13234 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13235 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13236 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13237 (vect_shift_permute_load_chain): Likewise.
13238 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13239 (expand_vector_condition, optimize_vector_constructor): Likewise.
13240 (lower_vec_perm, get_compute_type): Likewise.
13241 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13242 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13243 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13244 (vect_recog_mask_conversion_pattern): Likewise.
13245 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13246 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13247 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13248 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13249 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13250 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13251 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13252 (supportable_widening_operation): Likewise.
13253 (supportable_narrowing_operation): Likewise.
13254 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13256 * varasm.c (output_constant): Likewise.
13258 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13259 Alan Hayward <alan.hayward@arm.com>
13260 David Sherwood <david.sherwood@arm.com>
13262 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13263 so that both the length == 3 and length != 3 cases set up their
13264 own permute vectors. Add comments explaining why we know the
13265 number of elements is constant.
13266 (vect_permute_load_chain): Likewise.
13268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13269 Alan Hayward <alan.hayward@arm.com>
13270 David Sherwood <david.sherwood@arm.com>
13272 * machmode.h (mode_nunits): Change from unsigned char to
13274 (ONLY_FIXED_SIZE_MODES): New macro.
13275 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13276 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13277 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13279 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13280 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13281 or if measurement_type is not polynomial.
13282 * genmodes.c (ZERO_COEFFS): New macro.
13283 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13285 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13286 Use ZERO_COEFFS when emitting initializers.
13287 * data-streamer.h (bp_pack_poly_value): New function.
13288 (bp_unpack_poly_value): Likewise.
13289 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13290 for GET_MODE_NUNITS.
13291 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13292 for GET_MODE_NUNITS.
13293 * tree.c (make_vector_type): Remove temporary shim and make
13294 the real function take the number of units as a poly_uint64
13295 rather than an int.
13296 (build_vector_type_for_mode): Handle polynomial nunits.
13297 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13298 * emit-rtl.c (const_vec_series_p_1): Likewise.
13299 (gen_rtx_CONST_VECTOR): Likewise.
13300 * fold-const.c (test_vec_duplicate_folding): Likewise.
13301 * genrecog.c (validate_pattern): Likewise.
13302 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13303 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13304 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13305 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13306 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13307 * rtlanal.c (subreg_get_info): Likewise.
13308 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13309 (vect_grouped_load_supported): Likewise.
13310 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13311 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13312 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13313 (simplify_const_unary_operation, simplify_binary_operation_1)
13314 (simplify_const_binary_operation, simplify_ternary_operation)
13315 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13316 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13317 instead of CONST_VECTOR_NUNITS.
13318 * varasm.c (output_constant_pool_2): Likewise.
13319 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13320 explicit-encoded elements in the XVEC for variable-length vectors.
13322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13324 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13327 Alan Hayward <alan.hayward@arm.com>
13328 David Sherwood <david.sherwood@arm.com>
13330 * coretypes.h (fixed_size_mode): Declare.
13331 (fixed_size_mode_pod): New typedef.
13332 * builtins.h (target_builtins::x_apply_args_mode)
13333 (target_builtins::x_apply_result_mode): Change type to
13334 fixed_size_mode_pod.
13335 * builtins.c (apply_args_size, apply_result_size, result_vector)
13336 (expand_builtin_apply_args_1, expand_builtin_apply)
13337 (expand_builtin_return): Update accordingly.
13339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13341 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13342 * cselib.c (cselib_hash_rtx): Likewise.
13343 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13344 CONST_VECTOR encoding.
13346 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13347 Jeff Law <law@redhat.com>
13350 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13351 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13352 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13353 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13356 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13357 explicitly probe *sp in a noreturn function if there were any callee
13358 register saves or frame pointer is needed.
13360 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13363 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13364 BLKmode for ternary, binary or unary expressions.
13367 * var-tracking.c (delete_vta_debug_insn): New inline function.
13368 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13369 insns from get_insns () to NULL instead of each bb separately.
13370 Use delete_vta_debug_insn. No longer static.
13371 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13372 delete_vta_debug_insns callers.
13373 * rtl.h (delete_vta_debug_insns): Declare.
13374 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13375 instead of variable_tracking_main.
13377 2018-01-03 Martin Sebor <msebor@redhat.com>
13379 PR tree-optimization/83603
13380 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13381 arguments past the endof the argument list in functions declared
13382 without a prototype.
13383 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13384 Avoid checking when arguments are null.
13386 2018-01-03 Martin Sebor <msebor@redhat.com>
13389 * doc/extend.texi (attribute const): Fix a typo.
13390 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13391 issuing -Wsuggest-attribute for void functions.
13393 2018-01-03 Martin Sebor <msebor@redhat.com>
13395 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13396 offset_int::from instead of wide_int::to_shwi.
13397 (maybe_diag_overlap): Remove assertion.
13398 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13399 * gimple-ssa-sprintf.c (format_directive): Same.
13400 (parse_directive): Same.
13401 (sprintf_dom_walker::compute_format_length): Same.
13402 (try_substitute_return_value): Same.
13404 2018-01-03 Jeff Law <law@redhat.com>
13406 PR middle-end/83654
13407 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13408 non-constant residual for zero at runtime and avoid probing in
13409 that case. Reorganize code for trailing problem to mirror handling
13412 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13414 PR tree-optimization/83501
13415 * tree-ssa-strlen.c (get_string_cst): New.
13416 (handle_char_store): Call get_string_cst.
13418 2018-01-03 Martin Liska <mliska@suse.cz>
13420 PR tree-optimization/83593
13421 * tree-ssa-strlen.c: Include tree-cfg.h.
13422 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13423 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13424 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13426 (strlen_dom_walker::before_dom_children): Call
13427 gimple_purge_dead_eh_edges. Dump tranformation with details
13429 (strlen_dom_walker::before_dom_children): Update call by adding
13430 new argument cleanup_eh.
13431 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13433 2018-01-03 Martin Liska <mliska@suse.cz>
13436 * cif-code.def (VARIADIC_THUNK): New enum value.
13437 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13440 2018-01-03 Jan Beulich <jbeulich@suse.com>
13442 * sse.md (mov<mode>_internal): Tighten condition for when to use
13443 vmovdqu<ssescalarsize> for TI and OI modes.
13445 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13447 Update copyright years.
13449 2018-01-03 Martin Liska <mliska@suse.cz>
13452 * ipa-visibility.c (function_and_variable_visibility): Skip
13453 functions with noipa attribure.
13455 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13457 * gcc.c (process_command): Update copyright notice dates.
13458 * gcov-dump.c (print_version): Ditto.
13459 * gcov.c (print_version): Ditto.
13460 * gcov-tool.c (print_version): Ditto.
13461 * gengtype.c (create_file): Ditto.
13462 * doc/cpp.texi: Bump @copying's copyright year.
13463 * doc/cppinternals.texi: Ditto.
13464 * doc/gcc.texi: Ditto.
13465 * doc/gccint.texi: Ditto.
13466 * doc/gcov.texi: Ditto.
13467 * doc/install.texi: Ditto.
13468 * doc/invoke.texi: Ditto.
13470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13472 * vector-builder.h (vector_builder::m_full_nelts): Change from
13473 unsigned int to poly_uint64.
13474 (vector_builder::full_nelts): Update prototype accordingly.
13475 (vector_builder::new_vector): Likewise.
13476 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13477 (vector_builder::operator ==): Likewise.
13478 (vector_builder::finalize): Likewise.
13479 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13480 Take the number of elements as a poly_uint64 rather than an
13482 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13483 from unsigned int to poly_uint64.
13484 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13485 (vec_perm_indices::new_vector): Likewise.
13486 (vec_perm_indices::length): Likewise.
13487 (vec_perm_indices::nelts_per_input): Likewise.
13488 (vec_perm_indices::input_nelts): Likewise.
13489 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13490 number of elements per input as a poly_uint64 rather than an
13491 unsigned int. Use the original encoding for variable-length
13492 vectors, rather than clamping each individual element.
13493 For the second and subsequent elements in each pattern,
13494 clamp the step and base before clamping their sum.
13495 (vec_perm_indices::series_p): Handle polynomial element counts.
13496 (vec_perm_indices::all_in_range_p): Likewise.
13497 (vec_perm_indices_to_tree): Likewise.
13498 (vec_perm_indices_to_rtx): Likewise.
13499 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13500 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13501 (tree_vector_builder::new_binary_operation): Handle polynomial
13502 element counts. Return false if we need to know the number
13503 of elements at compile time.
13504 * fold-const.c (fold_vec_perm): Punt if the number of elements
13505 isn't known at compile time.
13507 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13509 * vec-perm-indices.h (vec_perm_builder): Change element type
13510 from HOST_WIDE_INT to poly_int64.
13511 (vec_perm_indices::element_type): Update accordingly.
13512 (vec_perm_indices::clamp): Handle polynomial element_types.
13513 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13514 (vec_perm_indices::all_in_range_p): Likewise.
13515 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13517 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13518 polynomial vec_perm_indices element types.
13519 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13520 * fold-const.c (fold_vec_perm): Likewise.
13521 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13522 * tree-vect-generic.c (lower_vec_perm): Likewise.
13523 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13524 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13525 element type to HOST_WIDE_INT.
13527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13528 Alan Hayward <alan.hayward@arm.com>
13529 David Sherwood <david.sherwood@arm.com>
13531 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13532 rather than an int. Use plus_constant.
13533 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13534 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13537 Alan Hayward <alan.hayward@arm.com>
13538 David Sherwood <david.sherwood@arm.com>
13540 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13541 a HOST_WIDE_INT to a poly_int64.
13543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13544 Alan Hayward <alan.hayward@arm.com>
13545 David Sherwood <david.sherwood@arm.com>
13547 * calls.c (load_register_parameters): Cope with polynomial
13548 mode sizes. Require a constant size for BLKmode parameters
13549 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13550 forces a parameter to be padded at the lsb end in order to
13551 fill a complete number of words, require the parameter size
13552 to be ordered wrt UNITS_PER_WORD.
13554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13555 Alan Hayward <alan.hayward@arm.com>
13556 David Sherwood <david.sherwood@arm.com>
13558 * reload1.c (spill_stack_slot_width): Change element type
13559 from unsigned int to poly_uint64_pod.
13560 (alter_reg): Treat mode sizes as polynomial.
13562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13563 Alan Hayward <alan.hayward@arm.com>
13564 David Sherwood <david.sherwood@arm.com>
13566 * reload.c (complex_word_subreg_p): New function.
13567 (reload_inner_reg_of_subreg, push_reload): Use it.
13569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13570 Alan Hayward <alan.hayward@arm.com>
13571 David Sherwood <david.sherwood@arm.com>
13573 * lra-constraints.c (process_alt_operands): Reject matched
13574 operands whose sizes aren't ordered.
13575 (match_reload): Refer to this check here.
13577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13578 Alan Hayward <alan.hayward@arm.com>
13579 David Sherwood <david.sherwood@arm.com>
13581 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13582 that the mode size is in the set {1, 2, 4, 8, 16}.
13584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13585 Alan Hayward <alan.hayward@arm.com>
13586 David Sherwood <david.sherwood@arm.com>
13588 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13589 Use plus_constant instead of gen_rtx_PLUS.
13591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13592 Alan Hayward <alan.hayward@arm.com>
13593 David Sherwood <david.sherwood@arm.com>
13595 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13596 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13597 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13598 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13599 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13600 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13601 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13602 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13603 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13604 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13606 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13607 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13608 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13609 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13610 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13611 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13612 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13613 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13614 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13615 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13617 * expr.c (emit_move_resolve_push): Treat the input and result
13618 of PUSH_ROUNDING as a poly_int64.
13619 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13620 (emit_push_insn): Likewise.
13621 * lra-eliminations.c (mark_not_eliminable): Likewise.
13622 * recog.c (push_operand): Likewise.
13623 * reload1.c (elimination_effects): Likewise.
13624 * rtlanal.c (nonzero_bits1): Likewise.
13625 * calls.c (store_one_arg): Likewise. Require the padding to be
13626 known at compile time.
13628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13629 Alan Hayward <alan.hayward@arm.com>
13630 David Sherwood <david.sherwood@arm.com>
13632 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13633 Use plus_constant instead of gen_rtx_PLUS.
13635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13636 Alan Hayward <alan.hayward@arm.com>
13637 David Sherwood <david.sherwood@arm.com>
13639 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13640 rather than an int.
13642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13643 Alan Hayward <alan.hayward@arm.com>
13644 David Sherwood <david.sherwood@arm.com>
13646 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13647 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13648 via stack temporaries. Treat the mode size as polynomial too.
13650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13651 Alan Hayward <alan.hayward@arm.com>
13652 David Sherwood <david.sherwood@arm.com>
13654 * expr.c (expand_expr_real_2): When handling conversions involving
13655 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13656 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13657 as a poly_uint64 too.
13659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13660 Alan Hayward <alan.hayward@arm.com>
13661 David Sherwood <david.sherwood@arm.com>
13663 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13666 Alan Hayward <alan.hayward@arm.com>
13667 David Sherwood <david.sherwood@arm.com>
13669 * combine.c (can_change_dest_mode): Handle polynomial
13670 REGMODE_NATURAL_SIZE.
13671 * expmed.c (store_bit_field_1): Likewise.
13672 * expr.c (store_constructor): Likewise.
13673 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13674 and polynomial REGMODE_NATURAL_SIZE.
13675 (gen_lowpart_common): Likewise.
13676 * reginfo.c (record_subregs_of_mode): Likewise.
13677 * rtlanal.c (read_modify_subreg_p): Likewise.
13679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13680 Alan Hayward <alan.hayward@arm.com>
13681 David Sherwood <david.sherwood@arm.com>
13683 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13684 numbers of elements.
13686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13687 Alan Hayward <alan.hayward@arm.com>
13688 David Sherwood <david.sherwood@arm.com>
13690 * match.pd: Cope with polynomial numbers of vector elements.
13692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13693 Alan Hayward <alan.hayward@arm.com>
13694 David Sherwood <david.sherwood@arm.com>
13696 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13697 in a POINTER_PLUS_EXPR.
13699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13700 Alan Hayward <alan.hayward@arm.com>
13701 David Sherwood <david.sherwood@arm.com>
13703 * omp-simd-clone.c (simd_clone_subparts): New function.
13704 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13705 (ipa_simd_modify_function_body): Likewise.
13707 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13708 Alan Hayward <alan.hayward@arm.com>
13709 David Sherwood <david.sherwood@arm.com>
13711 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13712 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13713 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13714 (expand_vector_condition, vector_element): Likewise.
13715 (subparts_gt): New function.
13716 (get_compute_type): Use subparts_gt.
13717 (count_type_subparts): Delete.
13718 (expand_vector_operations_1): Use subparts_gt instead of
13719 count_type_subparts.
13721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13722 Alan Hayward <alan.hayward@arm.com>
13723 David Sherwood <david.sherwood@arm.com>
13725 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13726 (vect_compile_time_alias): ...this new function. Do the calculation
13727 on poly_ints rather than trees.
13728 (vect_prune_runtime_alias_test_list): Update call accordingly.
13730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13731 Alan Hayward <alan.hayward@arm.com>
13732 David Sherwood <david.sherwood@arm.com>
13734 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13736 (vect_schedule_slp_instance): Likewise.
13738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13739 Alan Hayward <alan.hayward@arm.com>
13740 David Sherwood <david.sherwood@arm.com>
13742 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13743 constant and extern definitions for variable-length vectors.
13744 (vect_get_constant_vectors): Note that the number of units
13745 is known to be constant.
13747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13748 Alan Hayward <alan.hayward@arm.com>
13749 David Sherwood <david.sherwood@arm.com>
13751 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13752 of units as polynomial. Choose between WIDE and NARROW based
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 * tree-vect-stmts.c (simd_clone_subparts): New function.
13760 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13767 vectors as polynomial. Use build_index_vector for
13768 IFN_GOMP_SIMD_LANE.
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-stmts.c (get_load_store_type): Treat the number of
13775 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13776 for variable-length vectors.
13777 (vectorizable_mask_load_store): Treat the number of units as
13778 polynomial, asserting that it is constant if the condition has
13779 already been enforced.
13780 (vectorizable_store, vectorizable_load): Likewise.
13782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13783 Alan Hayward <alan.hayward@arm.com>
13784 David Sherwood <david.sherwood@arm.com>
13786 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13787 of units as polynomial. Punt if we can't tell at compile time
13788 which vector contains the final result.
13790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13791 Alan Hayward <alan.hayward@arm.com>
13792 David Sherwood <david.sherwood@arm.com>
13794 * tree-vect-loop.c (vectorizable_induction): Treat the number
13795 of units as polynomial. Punt on SLP inductions. Use an integer
13796 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13797 cast of such a series for variable-length floating-point
13800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13801 Alan Hayward <alan.hayward@arm.com>
13802 David Sherwood <david.sherwood@arm.com>
13804 * tree.h (build_index_vector): Declare.
13805 * tree.c (build_index_vector): New function.
13806 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13807 of units as polynomial, forcibly converting it to a constant if
13808 vectorizable_reduction has already enforced the condition.
13809 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13810 to create a {1,2,3,...} vector.
13811 (vectorizable_reduction): Treat the number of units as polynomial.
13812 Choose vectype_in based on the largest scalar element size rather
13813 than the smallest number of units. Enforce the restrictions
13816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13817 Alan Hayward <alan.hayward@arm.com>
13818 David Sherwood <david.sherwood@arm.com>
13820 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13821 number of units as polynomial.
13823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13824 Alan Hayward <alan.hayward@arm.com>
13825 David Sherwood <david.sherwood@arm.com>
13827 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13828 * target.def (autovectorize_vector_sizes): Return the vector sizes
13829 by pointer, using vector_sizes rather than a bitmask.
13830 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13831 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13832 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13834 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13835 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13836 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13837 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13838 * omp-general.c (omp_max_vf): Likewise.
13839 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13840 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13841 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13842 * tree-vect-slp.c (vect_slp_bb): Likewise.
13843 * doc/tm.texi: Regenerate.
13844 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13846 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13847 the vector size as a poly_uint64 rather than an unsigned int.
13848 (current_vector_size): Change from an unsigned int to a poly_uint64.
13849 (get_vectype_for_scalar_type): Update accordingly.
13850 * tree.h (build_truth_vector_type): Take the size and number of
13851 units as a poly_uint64 rather than an unsigned int.
13852 (build_vector_type): Add a temporary overload that takes
13853 the number of units as a poly_uint64 rather than an unsigned int.
13854 * tree.c (make_vector_type): Likewise.
13855 (build_truth_vector_type): Take the number of units as a poly_uint64
13856 rather than an unsigned int.
13858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13859 Alan Hayward <alan.hayward@arm.com>
13860 David Sherwood <david.sherwood@arm.com>
13862 * target.def (get_mask_mode): Take the number of units and length
13863 as poly_uint64s rather than unsigned ints.
13864 * targhooks.h (default_get_mask_mode): Update accordingly.
13865 * targhooks.c (default_get_mask_mode): Likewise.
13866 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13867 * doc/tm.texi: Regenerate.
13869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13870 Alan Hayward <alan.hayward@arm.com>
13871 David Sherwood <david.sherwood@arm.com>
13873 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13874 * omp-general.c (omp_max_vf): Likewise.
13875 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13876 (expand_omp_simd): Handle polynomial safelen.
13877 * omp-low.c (omplow_simd_context): Add a default constructor.
13878 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13879 (lower_rec_simd_input_clauses): Update accordingly.
13880 (lower_rec_input_clauses): Likewise.
13882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13883 Alan Hayward <alan.hayward@arm.com>
13884 David Sherwood <david.sherwood@arm.com>
13886 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13887 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13888 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13889 (vect_analyze_slp_cost): Likewise.
13890 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13891 (vect_model_load_cost): Likewise.
13893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13894 Alan Hayward <alan.hayward@arm.com>
13895 David Sherwood <david.sherwood@arm.com>
13897 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13898 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13899 from an unsigned int * to a poly_uint64_pod *.
13900 (calculate_unrolling_factor): New function.
13901 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13904 Alan Hayward <alan.hayward@arm.com>
13905 David Sherwood <david.sherwood@arm.com>
13907 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13908 from an unsigned int to a poly_uint64.
13909 (_loop_vec_info::slp_unrolling_factor): Likewise.
13910 (_loop_vec_info::vectorization_factor): Change from an int
13912 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13913 (vect_get_num_vectors): New function.
13914 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13915 (vect_get_num_copies): Use vect_get_num_vectors.
13916 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13917 to an unsigned int *.
13918 (vect_analyze_data_refs): Change min_vf from an int * to a
13920 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13921 than an unsigned HOST_WIDE_INT.
13922 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13923 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13924 to an unsigned int *.
13925 (vect_analyze_data_ref_dependences): Likewise.
13926 (vect_compute_data_ref_alignment): Handle polynomial vf.
13927 (vect_enhance_data_refs_alignment): Likewise.
13928 (vect_prune_runtime_alias_test_list): Likewise.
13929 (vect_shift_permute_load_chain): Likewise.
13930 (vect_supportable_dr_alignment): Likewise.
13931 (dependence_distance_ge_vf): Take the vectorization factor as a
13932 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13933 (vect_analyze_data_refs): Change min_vf from an int * to a
13935 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13936 vfm1 as a poly_uint64 rather than an int. Make the same change
13937 for the returned bound_scalar.
13938 (vect_gen_vector_loop_niters): Handle polynomial vf.
13939 (vect_do_peeling): Likewise. Update call to
13940 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13941 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13943 * tree-vect-loop.c (vect_determine_vectorization_factor)
13944 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13945 (vect_get_known_peeling_cost): Likewise.
13946 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13947 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13948 (vect_transform_loop): Likewise. Use the lowest possible VF when
13949 updating the upper bounds of the loop.
13950 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13951 rather than an int.
13952 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13953 polynomial unroll factors.
13954 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13955 (vect_make_slp_decision): Likewise.
13956 (vect_supported_load_permutation_p): Likewise, and polynomial
13958 (vect_analyze_slp_cost): Handle polynomial vf.
13959 (vect_slp_analyze_node_operations): Likewise.
13960 (vect_slp_analyze_bb_1): Likewise.
13961 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13962 than an unsigned HOST_WIDE_INT.
13963 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13964 (vectorizable_load): Handle polynomial vf.
13965 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13967 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13970 Alan Hayward <alan.hayward@arm.com>
13971 David Sherwood <david.sherwood@arm.com>
13973 * match.pd: Handle bit operations involving three constants
13974 and try to fold one pair.
13976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13978 * tree-vect-loop-manip.c: Include gimple-fold.h.
13979 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13980 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13981 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13982 Add a path that uses a step of VF instead of 1, but disable it
13984 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13985 and niters_no_overflow parameters. Update calls to
13986 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13987 Create a new SSA name if the latter choses to use a ste other
13988 than zero, and return it via niters_vector_mult_vf_var.
13989 * tree-vect-loop.c (vect_transform_loop): Update calls to
13990 vect_do_peeling, vect_gen_vector_loop_niters and
13991 slpeel_make_loop_iterate_ntimes.
13992 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13993 (vect_gen_vector_loop_niters): Update declarations after above changes.
13995 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13997 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13998 128-bit round to integer instructions.
13999 (ceil<mode>2): Likewise.
14000 (btrunc<mode>2): Likewise.
14001 (round<mode>2): Likewise.
14003 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14005 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14006 unaligned VSX load/store on P8/P9.
14007 (expand_block_clear): Allow the use of unaligned VSX
14008 load/store on P8/P9.
14010 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14012 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14014 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14015 swap associated with both a load and a store.
14017 2018-01-02 Andrew Waterman <andrew@sifive.com>
14019 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14020 * config/riscv/riscv.md (clear_cache): Use it.
14022 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14024 * web.c: Remove out-of-date comment.
14026 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14028 * expr.c (fixup_args_size_notes): Check that any existing
14029 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14030 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14031 (emit_single_push_insn): ...here.
14033 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14035 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14036 (const_vector_encoded_nelts): New function.
14037 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14038 (const_vector_int_elt, const_vector_elt): Declare.
14039 * emit-rtl.c (const_vector_int_elt_1): New function.
14040 (const_vector_elt): Likewise.
14041 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14042 of CONST_VECTOR_ELT.
14044 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14046 * expr.c: Include rtx-vector-builder.h.
14047 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14048 directly on the tree encoding.
14049 (const_vector_from_tree): Likewise.
14050 * optabs.c: Include rtx-vector-builder.h.
14051 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14052 sequence of "u" values.
14053 * vec-perm-indices.c: Include rtx-vector-builder.h.
14054 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14055 directly on the vec_perm_indices encoding.
14057 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14059 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14060 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14061 * rtx-vector-builder.h: New file.
14062 * rtx-vector-builder.c: Likewise.
14063 * rtl.h (rtx_def::u2): Add a const_vector field.
14064 (CONST_VECTOR_NPATTERNS): New macro.
14065 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14066 (CONST_VECTOR_DUPLICATE_P): Likewise.
14067 (CONST_VECTOR_STEPPED_P): Likewise.
14068 (CONST_VECTOR_ENCODED_ELT): Likewise.
14069 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14070 (unwrap_const_vec_duplicate): Likewise.
14071 (const_vec_series_p): Check for a non-duplicated vector encoding.
14072 Say that the function only returns true for integer vectors.
14073 * emit-rtl.c: Include rtx-vector-builder.h.
14074 (gen_const_vec_duplicate_1): Delete.
14075 (gen_const_vector): Call gen_const_vec_duplicate instead of
14076 gen_const_vec_duplicate_1.
14077 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14078 (gen_const_vec_duplicate): Use rtx_vector_builder.
14079 (gen_const_vec_series): Likewise.
14080 (gen_rtx_CONST_VECTOR): Likewise.
14081 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14082 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14083 Build a new vector rather than modifying a CONST_VECTOR in-place.
14084 (handle_special_swappables): Update call accordingly.
14085 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14086 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14087 Build a new vector rather than modifying a CONST_VECTOR in-place.
14088 (handle_special_swappables): Update call accordingly.
14090 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14092 * simplify-rtx.c (simplify_const_binary_operation): Use
14093 CONST_VECTOR_ELT instead of XVECEXP.
14095 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14097 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14098 the selector elements to be different from the data elements
14099 if the selector is a VECTOR_CST.
14100 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14101 ssizetype for the selector.
14103 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14105 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14106 before testing each element individually.
14107 * tree-vect-generic.c (lower_vec_perm): Likewise.
14109 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14111 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14112 * selftest-run-tests.c (selftest::run_tests): Call it.
14113 * vector-builder.h (vector_builder::operator ==): New function.
14114 (vector_builder::operator !=): Likewise.
14115 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14116 (vec_perm_indices::all_from_input_p): New function.
14117 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14118 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14119 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14120 instead of reading the VECTOR_CST directly. Detect whether both
14121 vector inputs are the same before constructing the vec_perm_indices,
14122 and update the number of inputs argument accordingly. Use the
14123 utility functions added above. Only construct sel2 if we need to.
14125 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14127 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14128 the broadcast of the low byte.
14129 (expand_mult_highpart): Use an explicit encoding for the permutes.
14130 * optabs-query.c (can_mult_highpart_p): Likewise.
14131 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14132 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14133 (vectorizable_bswap): Likewise.
14134 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14135 explicit encoding for the power-of-2 permutes.
14136 (vect_permute_store_chain): Likewise.
14137 (vect_grouped_load_supported): Likewise.
14138 (vect_permute_load_chain): Likewise.
14140 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14142 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14143 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14144 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14145 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14146 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14147 (vect_gen_perm_mask_any): Likewise.
14149 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14151 * int-vector-builder.h: New file.
14152 * vec-perm-indices.h: Include int-vector-builder.h.
14153 (vec_perm_indices): Redefine as an int_vector_builder.
14154 (auto_vec_perm_indices): Delete.
14155 (vec_perm_builder): Redefine as a stand-alone class.
14156 (vec_perm_indices::vec_perm_indices): New function.
14157 (vec_perm_indices::clamp): Likewise.
14158 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14159 (vec_perm_indices::new_vector): New function.
14160 (vec_perm_indices::new_expanded_vector): Update for new
14161 vec_perm_indices class.
14162 (vec_perm_indices::rotate_inputs): New function.
14163 (vec_perm_indices::all_in_range_p): Operate directly on the
14164 encoded form, without computing elided elements.
14165 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14166 encoding. Update for new vec_perm_indices class.
14167 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14168 the given vec_perm_builder.
14169 (expand_vec_perm_var): Update vec_perm_builder constructor.
14170 (expand_mult_highpart): Use vec_perm_builder instead of
14171 auto_vec_perm_indices.
14172 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14173 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14174 or double series encoding as appropriate.
14175 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14176 vec_perm_indices instead of auto_vec_perm_indices.
14177 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14178 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14179 (vect_permute_store_chain): Likewise.
14180 (vect_grouped_load_supported): Likewise.
14181 (vect_permute_load_chain): Likewise.
14182 (vect_shift_permute_load_chain): Likewise.
14183 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14184 (vect_transform_slp_perm_load): Likewise.
14185 (vect_schedule_slp_instance): Likewise.
14186 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14187 (vectorizable_mask_load_store): Likewise.
14188 (vectorizable_bswap): Likewise.
14189 (vectorizable_store): Likewise.
14190 (vectorizable_load): Likewise.
14191 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14192 vec_perm_indices instead of auto_vec_perm_indices. Use
14193 tree_to_vec_perm_builder to read the vector from a tree.
14194 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14195 vec_perm_builder instead of a vec_perm_indices.
14196 (have_whole_vector_shift): Use vec_perm_builder and
14197 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14198 truncation to calc_vec_perm_mask_for_shift.
14199 (vect_create_epilog_for_reduction): Likewise.
14200 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14201 from auto_vec_perm_indices to vec_perm_indices.
14202 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14203 instead of changing individual elements.
14204 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14205 the vector in d.perm.
14206 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14207 from auto_vec_perm_indices to vec_perm_indices.
14208 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14209 instead of changing individual elements.
14210 (arm_vectorize_vec_perm_const): Use new_vector to install
14211 the vector in d.perm.
14212 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14213 Update vec_perm_builder constructor.
14214 (rs6000_expand_interleave): Likewise.
14215 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14216 (rs6000_expand_interleave): Likewise.
14218 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14220 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14221 to qimode could truncate the indices.
14222 * optabs.c (expand_vec_perm_var): Likewise.
14224 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14226 * Makefile.in (OBJS): Add vec-perm-indices.o.
14227 * vec-perm-indices.h: New file.
14228 * vec-perm-indices.c: Likewise.
14229 * target.h (vec_perm_indices): Replace with a forward class
14231 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14232 * optabs.h: Include vec-perm-indices.h.
14233 (expand_vec_perm): Delete.
14234 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14235 (expand_vec_perm_const): Declare.
14236 * target.def (vec_perm_const_ok): Replace with...
14237 (vec_perm_const): ...this new hook.
14238 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14239 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14240 * doc/tm.texi: Regenerate.
14241 * optabs.def (vec_perm_const): Delete.
14242 * doc/md.texi (vec_perm_const): Likewise.
14243 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14244 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14245 expand_vec_perm for constant permutation vectors. Assert that
14246 the mode of variable permutation vectors is the integer equivalent
14247 of the mode that is being permuted.
14248 * optabs-query.h (selector_fits_mode_p): Declare.
14249 * optabs-query.c: Include vec-perm-indices.h.
14250 (selector_fits_mode_p): New function.
14251 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14252 is defined, instead of checking whether the vec_perm_const_optab
14253 exists. Use targetm.vectorize.vec_perm_const instead of
14254 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14255 fit in the vector mode before using a variable permute.
14256 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14257 vec_perm_indices instead of an rtx.
14258 (expand_vec_perm): Replace with...
14259 (expand_vec_perm_const): ...this new function. Take the selector
14260 as a vec_perm_indices rather than an rtx. Also take the mode of
14261 the selector. Update call to shift_amt_for_vec_perm_mask.
14262 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14263 Use vec_perm_indices::new_expanded_vector to expand the original
14264 selector into bytes. Check whether the indices fit in the vector
14265 mode before using a variable permute.
14266 (expand_vec_perm_var): Make global.
14267 (expand_mult_highpart): Use expand_vec_perm_const.
14268 * fold-const.c: Includes vec-perm-indices.h.
14269 * tree-ssa-forwprop.c: Likewise.
14270 * tree-vect-data-refs.c: Likewise.
14271 * tree-vect-generic.c: Likewise.
14272 * tree-vect-loop.c: Likewise.
14273 * tree-vect-slp.c: Likewise.
14274 * tree-vect-stmts.c: Likewise.
14275 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14277 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14278 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14279 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14280 (aarch64_vectorize_vec_perm_const): ...this new function.
14281 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14282 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14283 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14284 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14285 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14286 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14287 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14289 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14290 check for NEON modes.
14291 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14292 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14293 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14294 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14296 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14297 the old VEC_PERM_CONST conditions.
14298 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14299 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14300 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14301 (ia64_vectorize_vec_perm_const_ok): Merge into...
14302 (ia64_vectorize_vec_perm_const): ...this new function.
14303 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14304 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14305 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14306 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14307 * config/mips/mips.c (mips_expand_vec_perm_const)
14308 (mips_vectorize_vec_perm_const_ok): Merge into...
14309 (mips_vectorize_vec_perm_const): ...this new function.
14310 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14311 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14312 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14313 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14314 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14315 (rs6000_expand_vec_perm_const): Delete.
14316 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14318 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14319 (altivec_expand_vec_perm_const_le): Take each operand individually.
14320 Operate on constant selectors rather than rtxes.
14321 (altivec_expand_vec_perm_const): Likewise. Update call to
14322 altivec_expand_vec_perm_const_le.
14323 (rs6000_expand_vec_perm_const): Delete.
14324 (rs6000_vectorize_vec_perm_const_ok): Delete.
14325 (rs6000_vectorize_vec_perm_const): New function.
14326 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14327 an element count and rtx array.
14328 (rs6000_expand_extract_even): Update call accordingly.
14329 (rs6000_expand_interleave): Likewise.
14330 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14331 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14332 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14333 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14334 (rs6000_expand_vec_perm_const): Delete.
14335 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14336 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14337 (altivec_expand_vec_perm_const_le): Take each operand individually.
14338 Operate on constant selectors rather than rtxes.
14339 (altivec_expand_vec_perm_const): Likewise. Update call to
14340 altivec_expand_vec_perm_const_le.
14341 (rs6000_expand_vec_perm_const): Delete.
14342 (rs6000_vectorize_vec_perm_const_ok): Delete.
14343 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14344 reference to the SPE evmerge intructions.
14345 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14346 an element count and rtx array.
14347 (rs6000_expand_extract_even): Update call accordingly.
14348 (rs6000_expand_interleave): Likewise.
14349 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14350 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14352 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14354 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14356 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14357 vector mode and that that mode matches the mode of the data
14359 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14360 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14361 directly using expand_vec_perm_1 when forcing selectors into
14363 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14365 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14367 * optabs-query.h (can_vec_perm_p): Delete.
14368 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14369 * optabs-query.c (can_vec_perm_p): Split into...
14370 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14371 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14372 particular selector is valid.
14373 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14374 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14375 (vect_grouped_load_supported): Likewise.
14376 (vect_shift_permute_load_chain): Likewise.
14377 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14378 (vect_transform_slp_perm_load): Likewise.
14379 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14380 (vectorizable_bswap): Likewise.
14381 (vect_gen_perm_mask_checked): Likewise.
14382 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14383 implementations of variable permutation vectors into account
14384 when deciding which selector to use.
14385 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14386 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14387 with a false third argument.
14388 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14389 to test whether the constant selector is valid and can_vec_perm_var_p
14390 to test whether a variable selector is valid.
14392 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14394 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14395 * optabs-query.c (can_vec_perm_p): Likewise.
14396 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14397 instead of vec_perm_indices.
14398 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14399 (vect_gen_perm_mask_checked): Likewise,
14400 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14401 (vect_gen_perm_mask_checked): Likewise,
14403 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14405 * optabs-query.h (qimode_for_vec_perm): Declare.
14406 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14407 (qimode_for_vec_perm): ...this new function.
14408 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14410 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14412 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14413 does not have a conditional at the top.
14415 2018-01-02 Richard Biener <rguenther@suse.de>
14417 * ipa-inline.c (big_speedup_p): Fix expression.
14419 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14422 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14425 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14429 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14430 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14431 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14432 cond_taken_branch_cost 3->4.
14434 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14436 PR tree-optimization/83581
14437 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14438 TODO_cleanup_cfg if any changes have been made.
14440 PR middle-end/83608
14441 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14442 convert_modes if target mode has the right side, but different mode
14445 PR middle-end/83609
14446 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14447 last argument when extracting from CONCAT. If either from_real or
14448 from_imag is NULL, use expansion through memory. If result is not
14449 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14450 the parts directly to inner mode, if even that fails, use expansion
14453 PR middle-end/83623
14454 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14455 check for bswap in mode rather than HImode and use that in expand_unop
14458 Copyright (C) 2018 Free Software Foundation, Inc.
14460 Copying and distribution of this file, with or without modification,
14461 are permitted in any medium without royalty provided the copyright
14462 notice and this notice are preserved.