1 2018-05-29 Richard Biener <rguenther@suse.de>
3 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
5 2018-05-29 Martin Liska <mliska@suse.cz>
6 David Malcolm <dmalcolm@redhat.com>
8 * vec.c (test_reverse): New.
9 (vec_c_tests): Add new test.
10 * vec.h (vl_ptr>::reverse): New function.
12 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
14 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
16 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
19 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
21 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
23 2018-05-28 Richard Biener <rguenther@suse.de>
25 PR tree-optimization/85933
26 * tree-vect-data-refs.c (vect_record_base_alignments): Only
27 look at stmts marked as vectorizable.
29 2018-05-28 Richard Biener <rguenther@suse.de>
31 PR tree-optimization/85934
32 * tree-vect-generic.c (expand_vector_operations_1): Hoist
33 vector boolean check before scalar optimization.
35 2018-05-28 Jakub Jelinek <jakub@redhat.com>
37 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
40 2018-05-28 Mark Wielaard <mark@klomp.org>
42 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
43 if it is an expression containing a minus sign.
45 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
47 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
49 2018-05-27 Paul Koning <ni1d@arrl.net>
51 * config/pdp11/pdp11.md (truncsihi2): Remove.
53 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
54 Chung-Ju Wu <jasonwucj@gmail.com>
56 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
58 (unaligned_store_dw): Ditto.
59 * config/nds32/nds32-memory-manipulation.c
60 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
61 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
62 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
63 (emit_setmem_word_loop): Rename to ...
64 (emit_setmem_doubleword_loop): ... this.
65 (nds32_gen_dup_4_byte_to_word_value): New function.
66 (nds32_gen_dup_8_byte_to_double_word_value): New function.
67 (nds32_expand_setmem_loop): Refine implementation.
68 (nds32_expand_setmem_loop_v3m): Ditto.
69 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
72 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
74 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
76 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
78 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
79 (nds32_init_machine_status): Initialize machine->attr_naked_p and
80 machine->attr_no_prologue_p.
81 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
82 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
83 (nds32_expand_epilogue): Consider attr_naked_p.
84 (nds32_expand_epilogue_v3pop): Likewise.
85 (nds32_can_use_return_insn): Likewise.
86 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
87 attr_no_prologue_p fields.
88 * config/nds32/nds32.opt (mret-in-naked-func): New option.
90 2018-05-27 Jakub Jelinek <jakub@redhat.com>
93 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
96 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
98 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
100 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
102 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
104 (*<floatsuffix>floatv2div2sf2): Rename to ...
105 (*float<floatunssuffix>v2div2sf2): ... this.
106 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
107 (float<floatunssuffix>v2div2sf2_mask): ... this.
108 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
109 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
110 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
112 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
114 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
116 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
118 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
120 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
122 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
123 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
124 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
125 gen_ufix_truncv8dfv8si2.
126 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
127 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
128 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
129 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
130 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
131 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
132 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
133 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
135 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
139 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
141 2018-05-25 Jim Wilson <jimw@sifive.com>
143 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
144 * config/riscv/riscv.c (struct machine_function): Add
145 interrupt_handler_p and attribute_checked_p fields.
146 (riscv_attribute_table): Add interrupt.
147 (riscv_interrupt_type_p): New.
148 (riscv_save_reg_p): Save extra regs for interrupt handler.
149 (riscv_use_save_libcall): Return false for interrupt handler.
150 (riscv_first_stack_step): Add forward declaration.
151 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
152 for interrupt handler with large frame. Use it for saved reg list.
153 (riscv_expand_prologue): Move flag_stack_usage_info support to
154 eliminate duplication.
155 (riscv_expand_epilogue): Generate mret for interrupt handler.
156 (riscv_epilogue_uses): New.
157 (riscv_can_use_return_insn): Return false for interrupt handler.
158 (riscv_function_ok_for_sibcall): Likewise.
159 (riscv_set_current_function): Add interrupt handler support.
160 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
161 * config/riscv/riscv.md (UNSPECV_MRET): New.
163 (riscv_frflags, riscv_fsflags): Use tab after opcode.
165 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
167 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
169 PR tree-optimization/85712
170 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
171 this candidate has already been replaced in-situ by a copy.
173 2018-05-25 Jason Merrill <jason@redhat.com>
175 PR c++/80485 - inline function non-zero address.
176 * symtab.c (nonzero_address): Check DECL_COMDAT.
178 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
181 * config/alpha/alpha.md (ashlsi3): New insn pattern.
182 (*ashlsi_se): Rename from *ashldi_se. Define as sign
183 extension of SImode operation. Use const123_operand predicate.
185 (*saddl_se_1): Ditto.
187 (*ssubl_se_1): Ditto.
188 * config/alpha/predicates.md (const123_operand): New predicate.
189 * config/alpha/constraints.md (P): Use IN_RANGE.
191 2018-05-25 Richard Biener <rguenther@suse.de>
193 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
195 (ref_maybe_used_by_stmt_p): Likewise.
196 (stmt_may_clobber_ref_p): Likewise.
197 (stmt_may_clobber_ref_p_1): Likewise.
198 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
200 (ref_maybe_used_by_stmt_p): Likewise.
201 (stmt_may_clobber_ref_p): Likewise.
202 (stmt_may_clobber_ref_p_1): Likewise.
203 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
204 the alias oracle to disambiguate DRs with stmts DR analysis
206 (vect_analyze_data_refs): Do not give up on not analyzable
207 DRs for BB vectorization. Remove code truncating the dataref
210 2018-05-25 Jakub Jelinek <jakub@redhat.com>
213 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
214 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
215 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
217 2018-05-25 Richard Biener <rguenther@suse.de>
219 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
220 function, combining stmt data ref gathering and fatal analysis
222 (vect_analyze_data_refs): Remove now redudnant code and simplify.
223 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
224 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
225 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
226 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
228 2018-05-25 Bin Cheng <bin.cheng@arm.com>
230 PR tree-optimization/85720
231 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
232 SCC if all partitions are builtins.
233 (version_loop_by_alias_check): New parameter. Generate cancelable
234 runtime alias check if all partitions are builtins.
235 (distribute_loop): Update call to above function.
237 2018-05-25 Bin Cheng <bin.cheng@arm.com>
239 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
240 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
241 (parm_default_def_partition_arg): Ditto.
242 (set_parm_default_def_partition): Ditto.
243 (get_parm_default_def_partitions): Ditto and make it static.
244 (get_undefined_value_partitions): Ditto and make it static.
245 (remove_ssa_form): Refactor call to init_var_map here.
246 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
247 computation for loop region.
248 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
249 (register_default_def): Delete.
250 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
251 (parm_default_def_partition_arg): Ditto.
252 (set_parm_default_def_partition): Ditto.
253 (get_parm_default_def_partitions): Ditto and make it static.
254 (get_undefined_value_partitions): Ditto and make it static.
255 (coalesce_with_default, coalesce_with_default): Update comment.
256 (create_coalesce_list_for_region): New func factored out from
257 create_outofssa_var_map.
258 (populate_coalesce_list_for_outofssa): New func factored out from
259 create_outofssa_var_map and coalesce_ssa_name.
260 (create_outofssa_var_map): Delete.
261 (coalesce_ssa_name): Refactor to support live range computation.
262 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
263 (get_parm_default_def_partitions): Delete.
264 (get_undefined_value_partitions): Ditto.
265 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
266 computation for loop region.
267 (new_tree_live_info, loe_visit_block): Ditto.
268 (live_worklist, set_var_live_on_entry): Ditto.
269 (calculate_live_on_exit, verify_live_on_entry): Ditto.
270 * tree-ssa-live.h (struct _var_map): New fields.
271 (init_var_map): Change decl.
272 (region_contains_p): New.
274 2018-05-25 Bin Cheng <bin.cheng@arm.com>
276 * tree-ssa-live.h (live_merge_and_clear): Delete.
278 2018-05-25 Richard Biener <rguenther@suse.de>
281 * tree-dump.c (dequeue_and_dump): Remove access to removed
282 operand 2 of a SWITCH_EXPR.
284 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
286 * doc/sourcebuild.texi (vect_double_cond_arith): Include
287 multiplication and division.
288 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
289 (cond_udiv@var{m}, cond_umod@var{m}): Document.
290 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
291 (cond_udiv_optab, cond_umod_optab): New optabs.
292 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
293 (IFN_COND_RDIV): New internal functions.
294 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
295 TRUNC_MOD_EXPR and RDIV_EXPR.
296 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
297 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
299 (SVE_INT_BINARY): Include mult.
300 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
301 (optab, sve_int_op): Handle mult.
302 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
304 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
305 for SVE_INT_BINARY_SD.
307 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
309 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
310 (optab, sve_int_op): Handle div and udiv.
311 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
312 for SVE_INT_BINARY_SD.
313 (*<optab><mode>3): New insn for the same.
315 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
317 * tree-vect-patterns.c: Include predict.h.
318 (vect_recog_divmod_pattern): Restrict check for division support
319 to when optimizing for size.
321 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
323 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
324 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
325 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
326 (gimple_match_op::set_op): Likewise.
327 (gimple_resimplify4): Declare.
328 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
329 (expr::gen_transform): Likewise.
330 (decision_tree::gen): Generate a simplification routine for 4 operands.
331 * gimple-match-head.c (gimple_simplify): Add an overload for
332 4 operands. In the top-level function, handle up to 4 call
333 arguments and call gimple_resimplify4.
334 (gimple_resimplify4): New function.
335 (build_call_internal): Pass a fourth operand.
336 (maybe_push_to_seq): Likewise.
337 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
338 Fold VEC_COND_EXPRs of an operation and a default value into
339 an IFN_COND_* function if possible.
340 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
342 (SVE_COND_FP_BINARY): Include them.
343 (optab, sve_fp_op): Handle them.
344 (SVE_INT_BINARY_REV): New code iterator.
345 (SVE_COND_FP_BINARY_REV): New int iterator.
346 (commutative): New int attribute.
347 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
349 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
351 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
352 (*cond_<optab><mode>): New patterns for reversed operands.
354 2018-05-25 Richard Biener <rguenther@suse.de>
356 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
357 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
358 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
359 (STMT_VINFO_GROUPED_ACCESS): Adjust.
360 * tree-vect-data-refs.c (everywhere): Adjust users.
361 * tree-vect-loop.c (everywhere): Likewise.
362 * tree-vect-slp.c (everywhere): Likewise.
363 * tree-vect-stmts.c (everywhere): Likewise.
364 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
366 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
368 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
370 (gcc_cv_as_section_exclude): ... this.
371 Try Solaris as #exclude syntax.
372 * configure: Regenerate.
373 * config.in: Regenerate.
374 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
376 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
377 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
379 * varasm.c (default_elf_asm_named_section): Don't check if
380 HAVE_GAS_SECTION_EXCLUDE is defined.
382 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
384 * doc/md.texi: Update the documentation of the cond_* optabs
385 to mention the new final operand. Fix GET_MODE_NUNITS call.
386 Describe the scalar case too.
387 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
388 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
390 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
391 (get_conditional_internal_fn): Update comment.
392 * tree-vect-loop.c (vectorizable_reduction): Pass the original
393 accumulator value as a final argument to conditional functions.
394 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
395 a define_expand and add an "else" operand. Assert for now that
396 the else operand is equal to operand 2. Use SVE_INT_BINARY and
397 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
398 (*cond_<optab><mode>): New patterns.
399 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
400 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
401 (UNSPEC_COND_EOR): Delete.
402 (optab): Remove associated mappings.
403 (SVE_INT_BINARY): New code iterator.
404 (sve_int_op): Remove int attribute and add "minus" to the code
406 (SVE_COND_INT_OP): Delete.
407 (SVE_COND_FP_OP): Rename to...
408 (SVE_COND_FP_BINARY): ...this.
410 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
412 * optabs.c (can_reuse_operands_p): New function.
413 (maybe_legitimize_operands): Try to reuse the results for
416 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
418 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
419 Add {q} suffix to insn mnemonic.
421 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
423 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
424 (msp430_warn_func_return): New.
426 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
428 * fold-const.c (tree_nonzero_bits): New function.
429 * fold-const.h (tree_nonzero_bits): Likewise.
430 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
431 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
433 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
437 * varasm.c (assemble_alias): Check ifunc_resolver only on
440 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
443 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
444 when memory input operand is handled.
446 2018-05-24 Luis Machado <luis.machado@linaro.org>
448 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
450 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
452 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
454 * match.pd: Delay FMA folds until after vectorization.
456 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
459 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
460 address check not strict.
462 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
464 * gimple-match.h (gimple_match_op): New class.
465 (mprts_hook): Replace parameters with a gimple_match_op *.
466 (maybe_build_generic_op): Likewise.
467 (gimple_simplified_result_is_gimple_val): Replace parameters with
468 a const gimple_match_op *.
469 (gimple_simplify): Replace code_helper * and tree * parameters with
470 a gimple_match_op * parameter.
471 (gimple_resimplify1): Replace code_helper *, tree and tree *
472 parameters with a gimple_match_op * parameter.
473 (gimple_resimplify2): Likewise.
474 (gimple_resimplify3): Likewise.
475 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
476 parameters with a gimple_match_op * parameter.
477 * gimple-match-head.c (gimple_simplify): Change prototypes of
478 auto-generated functions to take a gimple_match_op * instead of
479 separate code_helper * and tree * parameters. Make the same
480 change in the top-level overload and update calls to the
481 gimple_resimplify routines. Update calls to the auto-generated
482 functions and to maybe_push_res_to_seq in the publicly-facing
483 operation-specific gimple_simplify overloads.
484 (gimple_match_op::MAX_NUM_OPS): Define.
485 (gimple_resimplify1): Replace rcode and ops with a single res_op
486 parameter. Update call to gimple_simplify.
487 (gimple_resimplify2): Likewise.
488 (gimple_resimplify3): Likewise.
489 (mprts_hook): Replace parameters with a gimple_match_op *.
490 (maybe_build_generic_op): Likewise.
491 (build_call_internal): Replace type, nargs and ops with
493 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
494 with a single gimple_match_op *. Update calls to mprts_hook,
495 build_call_internal and gimple_simplified_result_is_gimple_val.
496 Factor out code that is common to the tree_code and combined_fn cases.
497 * genmatch.c (expr::gen_transform): Replace tem_code and
498 tem_ops with a gimple_match_op called tem_op. Update calls
499 to the gimple_resimplify functions and maybe_push_res_to_seq.
500 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
501 res_ops. Update call to the gimple_resimplify functions.
502 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
503 (decision_tree::gen): Make the functions take a gimple_match_op *
504 called res_op instead of separate res_code and res_ops parameters.
505 Update call accordingly.
506 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
507 and ops with a single res_op parameter. Update calls to
508 maybe_build_generic_op and maybe_push_res_to_seq.
509 (fold_stmt_1): Update calls to gimple_simplify and
510 replace_stmt_with_simplification.
511 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
512 and gimple_simplified_result_is_gimple_val.
513 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
515 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
516 with a gimple_match_op *.
517 (vn_nary_build_or_lookup): Likewise. Update call to
518 vn_nary_build_or_lookup_1.
519 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
520 gimple_match_op *. Update calls to the gimple_resimplify routines
521 and to gimple_simplified_result_is_gimple_val.
522 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
523 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
524 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
525 (visit_nary_op): Likewise.
526 (visit_reference_op_load): Likewise.
528 2018-05-23 Luis Machado <luis.machado@linaro.org>
530 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
531 modifier for printing the step amount.
533 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
536 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
539 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
541 * doc/sourcebuild.texi (Endianness): New subsubsection.
543 2018-05-23 Luis Machado <luis.machado@linaro.org>
545 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
546 <prefetch_dynamic_strides>: New const bool field.
547 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
548 prefetch_dynamic_strides.
549 (exynosm1_prefetch_tune): Likewise.
550 (thunderxt88_prefetch_tune): Likewise.
551 (thunderx_prefetch_tune): Likewise.
552 (thunderx2t99_prefetch_tune): Likewise.
553 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
555 (aarch64_override_options_internal): Update to set
556 PARAM_PREFETCH_DYNAMIC_STRIDES.
557 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
558 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
559 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
560 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
561 prefetch-dynamic-strides setting.
563 2018-05-23 Luis Machado <luis.machado@linaro.org>
565 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
566 <minimum_stride>: New const int field.
567 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
568 minimum_stride field defaulting to -1.
569 (exynosm1_prefetch_tune): Likewise.
570 (thunderxt88_prefetch_tune): Likewise.
571 (thunderx_prefetch_tune): Likewise.
572 (thunderx2t99_prefetch_tune): Likewise.
573 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
574 <default_opt_level>: Set to 3.
575 (aarch64_override_options_internal): Update to set
576 PARAM_PREFETCH_MINIMUM_STRIDE.
577 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
578 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
579 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
580 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
581 stride is constant and is below the minimum stride threshold.
583 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
585 * config/arm/arm-cpus.in (mode26): Delete.
586 (armv4): Delete mode26 reference.
587 * config/arm/arm.c (arm_configure_build_target): Delete use of
590 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
592 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
594 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
595 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
597 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
598 (floatunsdidf2): Ditto.
600 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
602 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
603 (fixuns_trunc<mode>si2_avx512f): Ditto.
604 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
605 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
606 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
608 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
610 PR rtl-optimization/79985
611 * df-scan.c (df_insn_refs_collect): Remove special case for
612 global registers and asm statements.
614 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
616 * extend.texi (Global Register Variables): Rewrite the bullet list.
617 Note that the register is available for allocation. Note that access
618 via inline asm must use constraints. Add note about async-signal
619 handlers. Remove paragraph about automagic register selection.
621 2018-05-23 Richard Biener <rguenther@suse.de>
623 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
624 of fixed offset from memset VN.
626 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
628 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
630 (alloc_cand_and_find_basis): Initialize first_interp field.
631 (slsr_process_mul): Modify first_interp field.
632 (slsr_process_add): Likewise.
633 (slsr_process_cast): Modify first_interp field for each new
635 (slsr_process_copy): Likewise.
636 (dump_candidate): Dump first_interp field.
637 (replace_mult_candidate): Process all interpretations, not just
639 (replace_rhs_if_not_dup): Likewise.
640 (replace_one_candidate): Likewise.
642 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
644 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
646 (aarch64_needs_frame_chain): New function.
647 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
649 2018-05-23 Sudakshina Das <sudi.das@arm.com>
652 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
653 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
654 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
655 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
656 as true for strict-align.
657 (aarch64_can_inline_p): Perform checks even when callee has no
658 attributes to check for strict alignment.
659 * doc/extend.texi (AArch64 Function Attributes): Document
661 * doc/invoke.texi: (AArch64 Options): Likewise.
663 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
665 PR tree-optimization/85853
666 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
667 the handling of the root of the node to...
668 (vect_slp_analyze_node_operations_1): ...this new function,
669 and run the whole thing with the child nodes' def types
670 set according to their SLP node's def type.
672 2018-05-23 Richard Biener <rguenther@suse.de>
675 * tree-data-ref.c (create_runtime_alias_checks): Defer
676 and ignore overflow warnings.
678 2018-05-23 Yury Gribov <tetra2005@gmail.com>
680 PR tree-optimization/85822
681 * tree-vrp.c (is_masked_range_test): Fix handling of negative
684 2018-05-23 Richard Biener <rguenther@suse.de>
686 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
687 memset constants via native_interpret_expr.
689 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
692 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
694 (cgraph_node::create_alias): Likewise.
695 (cgraph_node::get_availability): Check ifunc_resolver instead
696 of looking up ifunc attribute.
697 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
698 * varasm.c (do_assemble_alias): Likewise.
699 (assemble_alias): Likewise.
700 (default_binds_local_p_3): Likewise.
701 * cgraph.h (cgraph_node): Add ifunc_resolver.
702 (cgraph_node::only_called_directly_or_aliased_p): Return false
704 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
706 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
707 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
708 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
709 instead of looking up ifunc attribute.
711 2018-05-22 Luis Machado <luis.machado@linaro.org>
713 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
715 2018-05-22 Martin Sebor <msebor@redhat.com>
718 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
719 only when expasion succeeds.
720 (expand_builtin_strcmp): Same.
721 (expand_builtin_strncmp): Same.
723 2018-05-22 Martin Sebor <msebor@redhat.com>
725 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
727 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
728 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
730 * config/aarch64/aarch64-ldpstp.md: Replace uses of
731 aarch64_mem_pair_operand with memory_operand and delete operand swapping
733 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
734 Add check for legitimate_address.
735 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
736 (aarch64_swap_ldrstr_operands): New.
737 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
740 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
741 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
743 * config/aarch64/aarch64.md: New patterns to generate stp
745 (store_pair_sw, store_pair_dw): New patterns to generate stp for
746 single words and double words.
747 (load_pair_sw, load_pair_dw): Likewise.
748 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
750 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
752 * config/aarch64/aarch64-ldpstp.md: Modify peephole
753 for different mode ldpstp and add peephole for merged zero stores.
755 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
757 (aarch64_gen_store_pair): Rename calls to match new patterns.
758 (aarch64_gen_load_pair): Rename calls to match new patterns.
759 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
760 (load_pair<DREG:mode><DREG2:mode>): ... This.
761 (store_pair<mode>): Rename to...
762 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
763 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
765 (V_INT_EQUIV): Handle SImode.
766 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
769 2018-05-22 Martin Sebor <msebor@redhat.com>
772 * calls.c (maybe_warn_nonstring_arg): Use string length to set
773 or ajust the presumed bound on an operation to avoid unnecessary
776 2018-05-22 Martin Sebor <msebor@redhat.com>
778 PR tree-optimization/85826
779 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
780 assuming that a DECL necesarily has a constant size.
782 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
785 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
787 2018-05-22 Richard Biener <rguenther@suse.de>
789 PR tree-optimization/85834
790 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
791 non-constant and non-zero memset arguments.
793 2018-05-22 Martin Liska <mliska@suse.cz>
796 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
798 2018-05-22 Richard Biener <rguenther@suse.de>
800 PR tree-optimization/85863
801 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
802 comparisons when vectype is specified.
803 (vectorizable_condition): Do not specify vectype for
804 vect_is_simple_cond when SLP vectorizing.
806 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
809 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
810 define __ibm128 as long double.
811 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
812 as a distinct type when IEEE 128-bit support is enabled.
813 (init_float128_ieee): Fix up conversions between IFmode and IEEE
814 128-bit types to use the correct functions.
815 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
816 convert between 128-bit floating point types that have different
817 modes but the same representation, instead of using gen_lowpart to
819 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
821 (IFKF_reg): New attributes to give the register constraints for
823 (extend<mode>tf2_internal): New insns to mark an explicit
824 conversion between 128-bit floating point types that have a
825 different mode but share the same representation.
827 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
829 PR tree-optimization/85814
830 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
831 a null return from get_strinfo when unsharing the next
832 strinfo in the chain.
834 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
837 * varasm.c (weak_finish): Clean up weak_decls.
839 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
841 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
842 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
843 UNSPEC_UADALP values.
844 * config/aarch64/iterators.md (ABAL): New int iterator.
847 (sur): Add mappings for the above.
848 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
850 (aarch64_<sur>abal<mode>_4): Likewise.
851 (aarch64_<sur>adalp<mode>_3): Likewise.
852 (<sur>sadv16qi): New define_expand.
854 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
856 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
857 (*movdf_internal): Ditto.
858 (*rcpsf2_sse): Ditto.
859 (*rsqrtsf2_sse): Ditto.
860 (*sqrt<mode>2_sse): Ditto.
862 2018-05-21 Tamar Christina <tamar.christina@arm.com>
864 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
866 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
867 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
868 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
869 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
871 * config/aarch64/arm_neon.h: Likewise.
872 * config/aarch64/iterators.md (VQ_I): New.
874 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
876 * config.gcc: Add arc/t-multilib-linux to tmake_file for
878 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
881 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
883 * config/nds32/constraints.md (S): New constraint.
884 * config/nds32/nds32.md (call_internal): Use constraint S.
885 (call_value_internal): Likewise.
886 (sibcall_internal): Likewise.
887 (sibcall_value_internal): Likewise.
889 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
890 Chung-Ju Wu <jasonwucj@gmail.com>
892 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
895 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
896 Chung-Ju Wu <jasonwucj@gmail.com>
898 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
899 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
900 (nds32_rtx_costs_impl): Simplify.
901 (nds32_address_cost_impl): Simplify.
902 (nds32_init_rtx_costs): New function.
903 (nds32_rtx_costs_speed_prefer): Likewise.
904 (nds32_rtx_costs_size_prefer): Likewise.
905 (nds32_address_cost_speed_prefer): Likewise.
906 (nds32_address_cost_speed_fwprop): Likewise.
907 (nds32_address_cost_size_prefer): Likewise.
908 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
909 * config/nds32/nds32.c (nds32_option_override): Use
910 nds32_init_rtx_costs function.
912 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
914 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
915 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
916 (TARGET_PIPELINE_N8): Likewise.
917 (TARGET_PIPELINE_N10): Likewise.
918 (TARGET_PIPELINE_N13): Likewise.
919 (TARGET_PIPELINE_GRAYWOLF): Likewise.
921 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
923 * config/nds32/nds32-fpu.md: Update copyright year.
925 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
927 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
929 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
931 * config/nds32/nds32.c
932 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
933 * config/nds32/nds32.opt (minline-asm-r15): New option.
935 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
937 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
939 * config/nds32/nds32.md (abssi2): New pattern.
941 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
943 * config/i386/i386.md (rex64namesuffix): New mode attribute.
944 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
945 Merge insn pattern from sse_cvtsi2ss<round_name> and
946 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
947 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
948 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
949 using SWI48 mode iterator.
950 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
951 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
952 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
953 pattern from sse_cvttss2si<round_saeonly_name>
954 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
955 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
956 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
957 using SWI48 mode iterator.
958 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
959 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
960 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
961 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
962 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
963 using SWI48 mode iterator.
964 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
965 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
966 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
967 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
968 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
970 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
971 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
972 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
973 pattern from sse_cvttsd2si<round_saeonly_name>
974 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
976 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
978 * config/nds32/nds32-md-auxiliary.c
979 (nds32_valid_smw_lwm_base_p): Refine.
980 (nds32_output_smw_single_word): Refine.
981 (nds32_output_smw_double_word): New.
982 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
984 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
986 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
987 (nds32_output_stack_pop): Refine.
988 (nds32_expand_unaligned_load): Refine.
989 (nds32_expand_unaligned_store): Refine.
991 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
992 Chung-Ju Wu <jasonwucj@gmail.com>
994 * config/nds32/constants.md: Add TP_REGNUM constant.
995 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
996 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
998 * config/nds32/nds32-doubleword.md: Consider flag_pic.
999 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1000 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1001 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1002 and PIC code generation.
1003 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1005 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1007 * config/nds32/nds32.md: Support TLS and PIC.
1008 * config/nds32/nds32.c: Support TLS and PIC.
1009 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1010 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1013 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1015 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1016 mode with E_ prefix.
1018 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1019 Chung-Ju Wu <jasonwucj@gmail.com>
1021 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1022 * config/nds32/nds32-md-auxiliary.c
1023 (symbolic_reference_mentioned_p): New.
1024 (nds32_legitimize_ict_address): New.
1025 (nds32_expand_ict_move): New.
1026 (nds32_indirect_call_referenced_p): New.
1027 (nds32_symbol_binds_local_p): Delete.
1028 (nds32_long_call_p): Modify.
1029 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1030 * config/nds32/nds32-protos.h
1031 (symbolic_reference_mentioned_p): Declare.
1032 (nds32_legitimize_ict_address): Declare.
1033 (nds32_expand_ict_move): Declare.
1034 (nds32_indirect_call_referenced_p): Declare.
1035 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1036 (nds32_relax_group): Use nds32_ict_const_p as condition.
1037 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1038 (nds32_asm_file_start): Output ict_model directive in asm code.
1039 (nds32_legitimate_address_p): Consider indirect call.
1040 (nds32_print_operand): Consider indirect call.
1041 (nds32_print_operand_address): Consider indirect call.
1042 (nds32_insert_attributes): Handle "indirect_call" attribute.
1043 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1044 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1045 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1046 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1047 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1048 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1049 (TARGET_ICT_MODEL_SMALL): Define.
1050 (TARGET_ICT_MODEL_LARGE): Define.
1051 * config/nds32/nds32.md (movsi): Consider ict model.
1052 (call, call_value): Consider ict model.
1053 (sibcall, sibcall_value): Consider ict model.
1054 * config/nds32/nds32.opt (mict-model): New option.
1055 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1058 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1059 Monk Chiang <sh.chiang04@gmail.com>
1060 Jim Wilson <jimw@sifive.com>
1062 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1063 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1064 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1065 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1066 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1067 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1068 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1069 compute save_libcall_adjustment properly.
1070 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1071 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1072 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1073 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1074 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1075 (ABI_SPEC): Handle mabi=ilp32e.
1076 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1077 (RVE): Add RVE mask.
1078 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1079 <-march>: Add rv32e as an example.
1081 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1084 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1085 (intra_create_variable_infos): Handle C++ constructors.
1087 2018-05-18 Martin Liska <mliska@suse.cz>
1089 * passes.def: Remove a redundant pass.
1091 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1094 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1096 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1098 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1100 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1102 (armv2, armv2a, armv3, armv3m): Delete architectures.
1103 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1104 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1105 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1107 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1108 (*mulsidi3adddi): Likewise.
1109 (mulsidi3): Likewise.
1110 (*mulsidi3_nov6): Likewise.
1111 (umulsidi3): Likewise.
1112 (umulsidi3_nov6): Likewise.
1113 (umaddsidi4): Likewise.
1114 (*umulsidi3adddi): Likewise.
1115 (smulsi3_highpart): Likewise.
1116 (*smulsi3_highpart_nov6): Likewise.
1117 (umulsi3_highpart): Likewise.
1118 (*umulsi3_highpart_nov6): Likewise.
1119 * config/arm/arm.h (arm_arch3m): Delete.
1120 * config/arm/arm.c (arm_arch3m): Delete.
1121 (arm_option_override_internal): Update armv3-related comment.
1122 (arm_configure_build_target): Delete use of isa_bit_mode32.
1123 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1124 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1125 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1127 (mulusa3): Likewise.
1128 * config/arm/arm-protos.h (arm_arch3m): Delete.
1129 * config/arm/arm-tables.opt: Regenerate.
1130 * config/arm/arm-tune.md: Likewise.
1131 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1132 deleted architectures.
1134 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1136 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1137 (armv5t, armv5te): New features.
1138 (ARMv5, ARMv5e): Delete fgroups.
1139 (ARMv5t, ARMv5te): Adjust for above changes.
1141 (armv5, armv5e): Delete arches.
1142 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1144 (*call_reg_arm): Likewise.
1145 (*call_value_reg_armv5): Likewise.
1146 (*call_value_reg_arm): Likewise.
1147 (*call_symbol): Likewise.
1148 (*call_value_symbol): Likewise.
1149 (*sibcall_insn): Likewise.
1150 (*sibcall_value_insn): Likewise.
1152 (prefetch): Likewise.
1153 (define_split and define_peephole2 dependent on arm_arch5):
1155 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1157 (TARGET_ARM_QBIT): Likewise.
1158 (TARGET_DSP_MULTIPLY): Likewise.
1159 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1160 (arm_arch5, arm_arch5e): Delete.
1161 (arm_arch5t, arm_arch5te): Declare.
1162 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1163 (arm_arch5t): Declare.
1164 (arm_option_reconfigure_globals): Update for the above.
1165 (arm_options_perform_arch_sanity_checks): Update comment, replace
1166 use of arm_arch5 with arm_arch5t.
1167 (use_return_insn): Likewise.
1168 (arm_emit_call_insn): Likewise.
1169 (output_return_instruction): Likewise.
1170 (arm_final_prescan_insn): Likewise.
1171 (arm_coproc_builtin_available): Likewise.
1172 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1173 arm_arch5e with arm_arch5t and arm_arch5te.
1174 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1175 (arm_arch5t, arm_arch5te): Declare.
1176 * config/arm/arm-tables.opt: Regenerate.
1177 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1178 * config/arm/t-multilib: Likewise.
1179 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1180 instead of arm_arch5.
1181 (*call_reg_thumb1): Likewise.
1182 (*call_value_reg_thumb1_v5): Likewise.
1183 (*call_value_reg_thumb1): Likewise.
1184 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1186 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1188 2018-05-18 Martin Liska <mliska@suse.cz>
1190 PR gcov-profile/84846
1191 * doc/gcov.texi: Document -t option of gcov tool.
1193 2018-05-18 Martin Liska <mliska@suse.cz>
1195 PR gcov-profile/84846
1196 * gcov.c (print_usage): Add new -t option.
1197 (process_args): Handle the option.
1198 (generate_results): Use stdout as output when requested by
1201 2018-05-18 Martin Liska <mliska@suse.cz>
1203 PR gcov-profile/84846
1204 * coverage.c (coverage_init): Write PWD to .gcno file.
1205 * doc/gcov.texi: Document how working directory is printed.
1206 * gcov-dump.c (dump_gcov_file): Print PWD.
1207 * gcov.c (output_intermediate_file): Likewise.
1208 (read_graph_file): Read PWD string.
1209 (output_lines): Print PWD.
1211 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1214 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1215 for retval and return false if all args to phi are zero.
1217 2018-05-18 Richard Biener <rguenther@suse.de>
1219 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1221 (evrp_dom_walker::before_dom_children): Call it.
1223 2018-05-18 Richard Biener <rguenther@suse.de>
1225 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1226 results when processing array refs with variable index.
1228 2018-05-18 Toon Moene <toon@moene.org>
1230 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1231 directly after that of -floop-interchange. Indicate that both
1232 options are enabled by default when specifying -O3.
1234 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1236 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1237 iterator. Delete separate integer-mode vec_set<mode> expander.
1238 (aarch64_simd_vec_setv2di): Delete.
1239 (vec_setv2di): Delete.
1240 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1241 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1242 the "w, r" alternative.
1244 2018-05-18 Martin Liska <mliska@suse.cz>
1246 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1247 * tree-pass.h (make_pass_lower_switch_O0): New function.
1248 * tree-switch-conversion.c (node_has_low_bound): Remove.
1249 (node_has_high_bound): Likewise.
1250 (node_is_bounded): Likewise.
1251 (class pass_lower_switch): Make it a template type and create
1253 (pass_lower_switch::execute): Add template argument.
1254 (make_pass_lower_switch): New function.
1255 (make_pass_lower_switch_O0): New function.
1256 (do_jump_if_equal): Remove.
1257 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1258 all the hard work to tree optimization passes.
1260 2018-05-18 Martin Liska <mliska@suse.cz>
1262 * dbgcnt.c (limit_low): Renamed from limit.
1263 (limit_high): New variable.
1264 (dbg_cnt_is_enabled): Check for upper limit.
1265 (dbg_cnt): Adjust dumping.
1266 (dbg_cnt_set_limit_by_index): Add new argument for high
1268 (dbg_cnt_set_limit_by_name): Likewise.
1269 (dbg_cnt_process_single_pair): Parse new format.
1270 (dbg_cnt_process_opt): Use strtok.
1271 (dbg_cnt_list_all_counters): Remove 'value' and add
1273 * doc/invoke.texi: Document changes.
1275 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1277 * doc/sourcebuild.texi (scalar_all_fma): Document.
1278 * tree.def (FMA_EXPR): Delete.
1279 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1280 * internal-fn.c (ternary_direct): New macro.
1281 (expand_ternary_optab_fn): Likewise.
1282 (direct_ternary_optab_supported_p): Likewise.
1283 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1284 * builtins.c (fold_builtin_fma): Delete.
1285 (fold_builtin_3): Don't call it.
1286 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1287 * expr.c (expand_expr_real_2): Likewise.
1288 * fold-const.c (operand_equal_p): Likewise.
1289 (fold_ternary_loc): Likewise.
1290 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1291 * gimple.c (DEFTREECODE): Likewise.
1292 * gimplify.c (gimplify_expr): Likewise.
1293 * optabs-tree.c (optab_for_tree_code): Likewise.
1294 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1295 * tree-eh.c (operation_could_trap_p): Likewise.
1296 (stmt_could_throw_1_p): Likewise.
1297 * tree-inline.c (estimate_operator_cost): Likewise.
1298 * tree-pretty-print.c (dump_generic_node): Likewise.
1299 (op_code_prio): Likewise.
1300 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1301 * tree-ssa-operands.c (get_expr_operands): Likewise.
1302 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1303 * fold-const-call.h (fold_fma): Delete.
1304 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1305 CFN_FNMA and CFN_FNMS.
1307 * genmatch.c (combined_fn): New enum.
1308 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1309 (commutative_op): New function.
1310 (commutate): Use it. Handle more than 2 operands.
1311 (dt_operand::gen_gimple_expr): Use commutative_op.
1312 (parser::parse_expr): Allow :c to be used with non-binary
1313 operators if the commutative operand is known.
1314 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1315 CFN_FMS, CFN_FNMA and CFN_FNMS.
1316 (backprop::process_assign_use): Remove FMA_EXPR handling.
1317 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1318 (gen_hsa_fma): New function.
1319 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1320 IFN_FNMA and IFN_FNMS.
1321 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1322 * gimple-fold.h (follow_all_ssa_edges): Declare.
1323 * gimple-fold.c (follow_all_ssa_edges): New function.
1324 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1325 gimple_build interface and use follow_all_ssa_edges to fold the result.
1326 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1327 instead of checking for optabs directly.
1328 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1329 rather than FMA_EXPRs.
1330 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1331 call to IFN_FMA instead of an FMA_EXPR.
1333 2018-05-17 Jim Wilson <jimw@sifive.com>
1335 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1336 sign extended promoted subreg, and the range does not have the sign bit
1337 set, then do a sign extend.
1339 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1340 test, check for sign extended subreg and/or constant operands, and
1341 do a sign extend in that case.
1343 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1345 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1346 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1348 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1349 Change logics_shift_reg to logics_shift_imm.
1350 (thunderx2t99_fp_loadpair_basic): Delete.
1351 (thunderx2t99_fp_storepair_basic): Delete.
1352 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1353 (thunderx2t99_asimd_polynomial): Delete.
1354 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1355 and neon_fp_mul_d_scalar_q.
1356 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1357 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1358 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1359 (thunderx2t99_asimd_lut): Add missing tbl types.
1360 (thunderx2t99_asimd_ext): Delete.
1361 (thunderx2t99_asimd_load1_1_mult): Delete.
1362 (thunderx2t99_asimd_load1_2_mult): Delete.
1363 (thunderx2t99_asimd_load1_ldp): New.
1364 (thunderx2t99_asimd_load1): New.
1365 (thunderx2t99_asimd_load2): Add missing *load2* types.
1366 (thunderx2t99_asimd_load3): New.
1367 (thunderx2t99_asimd_load4): New.
1368 (thunderx2t99_asimd_store1_1_mult): Delete.
1369 (thunderx2t99_asimd_store1_2_mult): Delete.
1370 (thunderx2t99_asimd_store2_mult): Delete.
1371 (thunderx2t99_asimd_store2_onelane): Delete.
1372 (thunderx2t99_asimd_store_stp): New.
1373 (thunderx2t99_asimd_store1): New.
1374 (thunderx2t99_asimd_store2): New.
1375 (thunderx2t99_asimd_store3): New.
1376 (thunderx2t99_asimd_store4): New.
1378 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1380 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1381 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1383 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1384 Segher Boessenkool <segher@kernel.crashing.org>
1387 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1390 2018-05-17 Richard Biener <rguenther@suse.de>
1392 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1393 for pruning loop and prune defs feeding only already visited PHIs.
1395 2018-05-17 Richard Biener <rguenther@suse.de>
1397 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1399 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1400 Richard Biener <rguenther@suse.de>
1402 PR tree-optimization/85793
1403 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1404 for VMAT_ELEMENTWISE.
1406 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1408 * internal-fn.h (lookup_internal_fn): Declare
1409 * internal-fn.c (lookup_internal_fn): New function.
1410 * gimple.c (gimple_build_call_from_tree): Handle calls to
1412 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1413 internal function names.
1414 * tree-pretty-print.c (dump_generic_node): Likewise.
1415 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1417 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1419 * gimple-fold.h (gimple_build): Make the function forms take
1420 combined_fn rather than built_in_function.
1421 (gimple_simplify): Likewise.
1422 * gimple-match-head.c (gimple_simplify): Likewise.
1423 * gimple-fold.c (gimple_build): Likewise.
1424 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1425 rather than gimple_build_call_internal.
1426 (get_initial_defs_for_reduction): Likewise.
1427 (vect_create_epilog_for_reduction): Likewise.
1428 (vectorizable_live_operation): Likewise.
1430 2018-05-17 Martin Liska <mliska@suse.cz>
1432 * gimple-ssa-sprintf.c (format_directive): Do not use
1433 space in between 'G_' and '('.
1435 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1438 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1439 even if the mask is not all ones.
1442 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1444 (ix86_gimple_fold_builtin): Likewise.
1447 * config/i386/i386.c: Include tree-vector-builder.h.
1448 (ix86_vector_shift_count): New function.
1449 (ix86_fold_builtin): Fold shift builtins by scalar count.
1450 (ix86_gimple_fold_builtin): Likewise.
1452 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1453 _mm512_setzero): New intrinsics.
1455 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1456 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1458 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1459 code generation for cases where splatting a value is not useful.
1460 * simplify-rtx.c (simplify_ternary_operation): Simplify
1461 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1462 mode to a vec_concat.
1464 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1466 * config.gcc: Support "goldmont-plus".
1467 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1469 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1470 PROCESSOR_GOLDMONT_PLUS.
1471 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1472 (processor_target_table): Add "goldmont-plus".
1473 (PTA_GOLDMONT_PLUS): Define.
1474 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1475 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1476 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1477 (fold_builtin_cpu): Add "goldmont-plus".
1478 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1479 (ix86_option_override_internal): Add "goldmont-plus".
1480 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1481 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1482 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1483 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1485 2018-05-17 Richard Biener <rguenther@suse.de>
1487 PR tree-optimization/85757
1488 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1489 remove defs that only feed that PHI from further processing.
1491 2018-05-16 Jim Wilson <jimw@sifive.com>
1493 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1495 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1497 2018-05-16 Mark Wielaard <mark@klomp.org>
1499 * dwarf2out.c (count_index_strings): New function.
1500 (output_indirect_strings): Call count_index_strings and generate
1501 header for dwarf_version >= 5.
1503 2018-05-16 Mark Wielaard <mark@klomp.org>
1505 * dwarf2out.c (dwarf_FORM): New function.
1506 (set_indirect_string): Use dwarf_FORM.
1507 (reset_indirect_string): Likewise.
1508 (size_of_die): Likewise.
1509 (value_format): Likewise.
1510 (output_die): Likewise.
1511 (add_skeleton_AT_string): Likewise.
1512 (output_macinfo_op): Likewise.
1513 (index_string): Likewise.
1514 (output_index_string_offset): Likewise.
1515 (output_index_string): Likewise.
1516 (count_index_strings): Likewise.
1518 2018-05-16 Carl Love <cel@us.ibm.com>
1520 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1521 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1523 2018-05-16 Martin Jambor <mjambor@suse.cz>
1525 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1526 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1528 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1530 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1531 (fnma<mode>4): Likewise.
1532 (fms<mode>4): Likewise.
1533 (fnms<mode>4): Likewise.
1534 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1535 (aarch64_fnma<mode>4): Likewise.
1536 (aarch64_fms<mode>4): Likewise.
1537 (aarch64_fnms<mode>4): Likewise.
1538 (aarch64_fnmadd<mode>4): Likewise.
1540 2018-05-16 Jason Merrill <jason@redhat.com>
1542 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1544 2018-05-16 Richard Biener <rguenther@suse.de>
1546 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1547 (dump_stmt_cost): Declare.
1548 (add_stmt_cost): Dump cost we add.
1549 (add_stmt_costs): New function.
1550 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1552 (vect_analyze_stmt): Adjust prototype.
1553 (vectorizable_condition): Likewise.
1554 (vectorizable_live_operation): Likewise.
1555 (vectorizable_reduction): Likewise.
1556 (vectorizable_induction): Likewise.
1557 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1558 cost vector to pass to vectorizable_ and record afterwards.
1559 (vect_model_reduction_cost): Take cost vector argument and adjust.
1560 (vect_model_induction_cost): Likewise.
1561 (vectorizable_reduction): Likewise.
1562 (vectorizable_induction): Likewise.
1563 (vectorizable_live_operation): Likewise.
1564 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1565 SLP_TREE_NUMBER_OF_VEC_STMTS.
1566 (vect_analyze_slp_cost_1): Remove.
1567 (vect_analyze_slp_cost): Likewise.
1568 (vect_slp_analyze_node_operations): Take visited args and
1569 a target cost vector. Avoid processing already visited stmt sets.
1570 (vect_slp_analyze_operations): Use a local cost vector to gather
1571 costs and register those of non-discarded instances.
1572 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1573 (vect_schedule_slp_instance): Remove copying of
1574 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1576 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1577 adding cost. Record cost entry location.
1578 (vect_prologue_cost_for_slp_op): Function to compute cost of
1579 a constant or invariant generated for SLP vect in the prologue,
1580 split out from vect_analyze_slp_cost_1.
1581 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1582 (vect_model_promotion_demotion_cost): Likewise.
1583 (vect_model_store_cost): Likewise, make static.
1584 (vect_model_load_cost): Likewise.
1585 (vectorizable_bswap): Add cost vector arg and adjust.
1586 (vectorizable_call): Likewise.
1587 (vectorizable_simd_clone_call): Likewise.
1588 (vectorizable_conversion): Likewise.
1589 (vectorizable_assignment): Likewise.
1590 (vectorizable_shift): Likewise.
1591 (vectorizable_operation): Likewise.
1592 (vectorizable_store): Likewise.
1593 (vectorizable_load): Likewise.
1594 (vectorizable_condition): Likewise.
1595 (vectorizable_comparison): Likewise.
1596 (can_vectorize_live_stmts): Likewise.
1597 (vect_analyze_stmt): Likewise.
1598 (vect_transform_stmt): Adjust calls to vectorizable_*.
1599 * tree-vectorizer.c: Include gimple-pretty-print.h.
1600 (dump_stmt_cost): New function.
1602 2018-05-16 Richard Biener <rguenther@suse.de>
1604 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1605 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1606 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1607 (check_name): New callback.
1608 (dse_classify_store): Track cycles via a visited bitmap of PHI
1609 defs and simplify handling of in-loop and across loop dead stores
1610 and properly fail for loop-variant refs. Handle byte-tracking with
1611 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1614 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1616 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1617 (vect_get_mask_type_for_stmt): Likewise.
1618 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1620 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1621 to determine the statement's vector type and the vector type that
1622 should be used for calculating nunits. Deal with cases in which
1623 the type has to be deferred.
1624 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1625 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1626 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1627 (vect_determine_vf_for_stmt): New functions, split out from...
1628 (vect_determine_vectorization_factor): ...here.
1629 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1630 (vect_get_mask_type_for_stmt): New functions, split out from
1631 vect_determine_vectorization_factor.
1633 2018-05-16 Richard Biener <rguenther@suse.de>
1635 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1636 verify the [VEC_]COND_EXPR embedded comparison.
1638 2018-05-15 Martin Sebor <msebor@redhat.com>
1640 PR tree-optimization/85753
1641 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1642 RECORD_TYPE in addition to ARRAY_TYPE.
1644 2018-05-15 Martin Sebor <msebor@redhat.com>
1647 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1649 2018-05-15 Richard Biener <rguenther@suse.de>
1651 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1652 add by_clobber_p one. Change algorithm to collect all defs
1653 representing uses we need to walk and try reducing them to
1654 a single one before failing.
1655 (dse_dom_walker::dse_optimize_stmt): Adjust.
1657 2018-05-13 Mark Wielaard <mark@klomp.org>
1659 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1660 (size_of_loc_descr): Likewise.
1661 (output_loc_operands): Likewise.
1662 (output_loc_operands_raw): Likewise.
1663 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1664 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1665 (hash_loc_operands): Likewise.
1666 (compare_loc_operands): Likewise.
1668 2018-05-14 Mark Wielaard <mark@klomp.org>
1670 * dwarf2out.c (count_index_addrs): New function.
1671 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1673 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1675 PR tree-optimization/83648
1676 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1677 return value as malloc candidate.
1679 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1682 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1683 as true in call to suggest_attribute.
1685 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1687 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1688 -mreadonly-in-sdata.
1690 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1692 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1694 (aarch64_crypto_aesd_fused): Likewise.
1696 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1698 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1699 (movsi_aarch64): Likewise.
1700 (load_pairsi): Likewise.
1701 (load_pairdi): Likewise.
1702 (store_pairsi): Likewise.
1703 (store_pairdi): Likewise.
1704 (load_pairsf): Likewise.
1705 (load_pairdf): Likewise.
1706 (store_pairsf): Likewise.
1707 (store_pairdf): Likewise.
1708 (zero_extend): Likewise.
1709 (trunc): Swap alternatives.
1710 (fcvt_target): Add '?' to prefer w over r.
1712 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1715 * config/i386/i386.md: Disallow non-commutative arithmetics in
1716 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1717 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1718 in the peephole2 before it.
1720 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1722 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1723 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1724 (ix86_handle_option): Handle -mcldemote.
1725 * config.gcc: New header.
1726 * config/i386/cldemoteintrin.h: New file.
1727 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1728 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1730 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1731 OPTION_MASK_ISA_CLDEMOTE.
1732 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1733 (ix86_valid_target_attribute_inner_p): Ditto.
1734 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1735 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1736 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1737 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1738 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1740 * config/i386/i386.opt: Add -mcldemote.
1741 * config/i386/x86intrin.h: New header.
1742 * doc/invoke.texi: Add -mcldemote.
1744 2018-05-14 Richard Biener <rguenther@suse.de>
1746 * doc/match-and-simplify.texi: Adjust :s documentation.
1748 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1750 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1751 intended memcpy size.
1752 (REORDER_45): Likewise.
1754 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1756 * sort.cc: New file.
1757 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1758 * vec.c (qsort_chk): Use gcc_qsort.
1759 * Makefile.in (OBJS-libcommon): Add sort.o.
1760 (build/sort.o): New target. Use it...
1761 (BUILD_RTL): ... here, and...
1762 (build/gencfn-macros): ... here, and...
1763 (build/genmatch): ... here.
1765 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1766 Chung-Ju Wu <jasonwucj@gmail.com>
1768 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1769 * config/nds32/nds32-graywolf.md: New file.
1770 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1771 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1773 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1774 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1775 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1776 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1777 * config/nds32/pipelines.md: Include n15 settings.
1779 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1780 Chung-Ju Wu <jasonwucj@gmail.com>
1782 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1783 * config/nds32/nds32-n13.md: New file.
1784 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1785 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1787 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1788 * config/nds32/nds32.md (pipeline_model): Add n13.
1789 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1790 * config/nds32/pipelines.md: Include n13 settings.
1792 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1793 Chung-Ju Wu <jasonwucj@gmail.com>
1795 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1796 * config/nds32/nds32-n10.md: New file.
1797 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1798 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1800 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1801 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1802 * config/nds32/nds32.md (pipeline_model): Add n10.
1803 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1804 * config/nds32/pipelines.md: Include n10 settings.
1806 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1807 Kito Cheng <kito.cheng@gmail.com>
1808 Chung-Ju Wu <jasonwucj@gmail.com>
1810 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1811 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1812 Add enum values for DSP extension instructions.
1813 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1815 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1816 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1818 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1819 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1820 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1821 * config/nds32/nds32-intrinsic.md: Likewise.
1822 * config/nds32/nds32_intrinsic.h: Likewise.
1823 * config/nds32/nds32-md-auxiliary.c: Likewise.
1824 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1825 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1826 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1827 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1828 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1829 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1830 TYPE_DMAC in switch statement.
1831 * config/nds32/nds32.c: New checking and implementation for DSP
1832 extension instructions.
1833 * config/nds32/nds32.h: Likewise.
1834 * config/nds32/nds32.md: Likewise.
1835 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1836 * config/nds32/predicates.md: Implement new predicates for DSP
1839 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1841 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1842 Reformat alternatives and attributes so it is easier to identify
1843 which constraints/attributes go with which instruction.
1844 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1845 (mov<mode>_softfloat32, FMOVE64): Likewise.
1846 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1847 (mov<mode>_softfloat64, FMOVE64): Likewise.
1849 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1851 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1853 (Basic PowerPC Built-in Functions): The new name of the
1854 subsection previously known as "PowerPC Built-in Functions".
1855 (Basic PowerPC Built-in Functions Available on all Configurations):
1857 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1858 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1859 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1860 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1862 2018-05-11 Martin Jambor <mjambor@suse.cz>
1865 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1868 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1871 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1873 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1875 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1876 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1877 (ix86_handle_option): Handle -mwaitpkg.
1878 * config.gcc: New header.
1879 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1880 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1881 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1883 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1884 OPTION_MASK_ISA_WAITPKG.
1885 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1886 (ix86_option_override_internal): Add PTA_WAITPKG.
1887 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1888 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1889 IX86_BUILTIN_TPAUSE.
1890 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1891 __builtin_ia32_umwait and __builtin_ia32_tpause.
1892 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1893 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1894 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1895 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1896 UNSPECV_TPAUSE): New.
1897 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1898 * config/i386/i386.opt: Add -mwaitpkg.
1899 * config/i386/waitpkgintrin.h: New file.
1900 * config/i386/x86intrin.h: New header.
1901 * doc/invoke.texi: Add -mwaitpkg.
1903 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1906 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1908 (cortex-m0): Use armv6s-m isa.
1909 (cortex-m0plus): Likewise.
1910 (cortex-m1): Likewise.
1911 (cortex-m0.small-multiply): Likewise.
1912 (cortex-m0plus.small-multiply): Likewise.
1913 (cortex-m1.small-multiply): Likewise.
1915 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1916 Jakub Jelinek <jakub@redhat.com>
1918 PR tree-optimization/85692
1919 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1920 source permute as well.
1922 2018-05-11 Martin Liska <mliska@suse.cz>
1925 * doc/extend.texi: Document LLVM style format for no_sanitize
1928 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1930 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1931 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1932 (mode_supports_vsx_dform_quad): Likewise.
1933 (mode_supports_vmx_dform): Move these functions to be next to the
1934 other mode_supports functions.
1935 (mode_supports_dq_form): Likewise.
1936 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1937 mode_supports_dq_form.
1938 (reg_offset_addressing_ok_p): Likewise.
1939 (offsettable_ok_by_alignment): Likewise.
1940 (rs6000_legitimate_offset_address_p): Likewise.
1941 (legitimate_lo_sum_address_p): Likewise.
1942 (rs6000_legitimize_address): Likewise.
1943 (rs6000_legitimize_reload_address): Likewise.
1944 (rs6000_secondary_reload_inner): Likewise.
1945 (rs6000_preferred_reload_class): Likewise.
1946 (rs6000_output_move_128bit): Likewise.
1948 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1950 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1951 Generate SImode target register for null target.
1952 <case IX86_BUILTIN_XGETBV>: Ditto.
1953 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1954 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1956 2018-05-10 Carl Love <cel@us.ibm.com>
1958 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1959 dcbtt and dcbtstt if operands[2] is 0.
1961 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1964 * config/i386/sse.md (usadv64qi): New expander.
1966 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1968 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1969 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1970 -maltivec=be support.
1971 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1972 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1973 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1974 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1975 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1976 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1977 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1978 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1979 altivec_vsumsws): Adjust.
1980 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1981 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1982 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1984 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1985 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1986 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1987 (altivec_lve<VI_char>x): Delete expand.
1988 (*altivec_lve<VI_char>x_internal): Rename to...
1989 (altivec_lve<VI_char>x): ... this.
1990 (altivec_lvxl_<mode>): Delete expand.
1991 (*altivec_lvxl_<mode>_internal): Rename to ...
1992 (altivec_lvxl_<mode>): ... this.
1993 (altivec_stvxl_<mode>): Delete expand.
1994 (*altivec_stvxl_<mode>_internal): Rename to ...
1995 (altivec_stvxl_<mode>): ... this.
1996 (altivec_stve<VI_char>x): Delete expand.
1997 (*altivec_stve<VI_char>x_internal): Rename to ...
1998 (altivec_stve<VI_char>x): ... this.
1999 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2000 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2001 reduc_plus_scal_<mode>): Adjust.
2002 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2004 (rs6000_cpu_cpp_builtins): Adjust.
2005 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2006 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2007 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2008 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2009 -maltivec=be support.
2010 (rs6000_split_vec_extract_var): Adjust.
2011 (rs6000_split_v4si_init): Adjust.
2012 (swap_selector_for_mode): Delete.
2013 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2014 altivec_expand_stvex_be): Delete.
2015 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2016 -maltivec=be support.
2017 (rs6000_gimple_fold_builtin): Ditto.
2018 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2020 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2021 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2022 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2023 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2024 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2025 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2026 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2027 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2028 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2029 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2030 anonymous split): Adjust.
2031 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2032 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2034 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2036 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2037 when --with-gxx-include-dir is also specified.
2038 * configure: Regenerate.
2040 2018-05-09 Jim Wilson <jimw@sifive.com>
2043 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2044 * config/riscv/t-withmultilib: New.
2045 * config/riscv/withmultilib.h: New.
2046 * doc/install.texi: Document RISC-V --with-multilib-list support.
2048 2018-05-09 Richard Biener <rguenther@suse.de>
2050 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2052 (vect_bb_vectorization_profitable_p): Adjust. Compute
2053 actual scalar cost using the cost vector and the add_stmt_cost
2056 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2058 PR rtl-optimization/85645
2059 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2060 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2062 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2064 PR rtl-optimization/85645
2065 * shrink-wrap.c (spread_components): Return a boolean saying if
2066 anything was changed.
2067 (try_shrink_wrapping_separate): Iterate spread_components until
2068 nothing changes anymore.
2070 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2072 PR rtl-optimization/85645
2073 * regrename.c (build_def_use): Also kill the chains that include the
2074 destination of a REG_CFA_REGISTER note.
2076 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2078 PR rtl-optimization/85645
2079 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2080 insn that has a REG_CFA_REGISTER note.
2082 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2084 * cfgexpand.c (expand_clobber): New function.
2085 (expand_gimple_stmt_1): Use it.
2086 * tree-vect-stmts.c (vect_clobber_variable): New function,
2088 (vectorizable_simd_clone_call): ...here.
2089 (vectorizable_store): Emit a clobber either side of an
2090 IFN_STORE_LANES sequence.
2091 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2093 2018-05-09 Tom de Vries <tom@codesourcery.com>
2096 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2097 (define_insn "trap_if_false"): Add exit after trap.
2099 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2101 PR rtl-optimization/85638
2102 * bb-reorder.c: Include common/common-target.h.
2103 (create_forwarder_block): New function extracted from...
2104 (fix_up_crossing_landing_pad): ...here. Rename into...
2105 (dw2_fix_up_crossing_landing_pad): ...this.
2106 (sjlj_fix_up_crossing_landing_pad): New function.
2107 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2108 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2109 from both partitions and exit the loop after one iteration.
2111 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2114 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2116 (Basic PowerPC Built-in Functions): The new name of the
2117 subsection previously known as "PowerPC Built-in Functions".
2118 (Basic PowerPC Built-in Functions Available on all Configurations):
2120 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2122 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2124 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2126 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2129 2018-05-08 Jim Wilson <jimw@sifive.com>
2131 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2132 (LD_EMUL_SUFFIX): New.
2133 (LINK_SPEC): Use it.
2135 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2137 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2139 (Basic PowerPC Built-in Functions): The new name of the
2140 subsection previously known as "PowerPC Built-in Functions".
2141 (Basic PowerPC Built-in Functions Available on all Configurations):
2143 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2145 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2147 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2149 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2152 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2155 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2156 after cmpelim optimization.
2158 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2160 * config.gcc: Support "goldmont".
2161 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2162 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2164 * config/i386/i386.c (m_GOLDMONT): Define.
2165 (processor_target_table): Add "goldmont".
2166 (PTA_GOLDMONT): Define.
2167 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2168 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2169 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2170 (fold_builtin_cpu): Add "goldmont".
2171 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2172 (ix86_option_override_internal): Add "goldmont".
2173 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2174 (processor_type): Add PROCESSOR_GOLDMONT.
2175 * config/i386/i386.md: Add CPU "glm".
2176 * config/i386/glm.md: New file.
2177 * config/i386/x86-tune.def: Add m_GOLDMONT.
2178 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2180 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2183 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2185 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2186 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2187 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2190 * config/i386/i386.c (ix86_fold_builtin): Handle
2191 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2194 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2195 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2197 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2200 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2201 (check_arch): Likewise.
2202 (check_fpu): Return the result rather than printing it.
2203 (end arch): Fix operator precedence.
2204 (end cpu): Likewise.
2205 (END): Print the result from check_fpu.
2207 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2208 Alan Hayward <alan.hayward@arm.com>
2209 David Sherwood <david.sherwood@arm.com>
2211 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2212 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2213 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2214 (*fcmuo<mode>_and): New patterns.
2216 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2218 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2219 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2220 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2221 (cmp_op, sve_imm_con): New code attributes.
2222 (SVE_COND_INT_CMP, imm_con): Delete.
2223 (cmp_op): Remove above unspecs from int attribute.
2224 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2226 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2227 comparison-specific unspecs.
2228 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2229 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2230 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2231 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2232 (*vec_fcm<cmp_op><mode>): Rename to...
2233 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2234 (*vec_fcmuo<mode>): Rename to...
2235 (*fcmuo<mode>): ...this and adjust likewise.
2236 (*pred_fcm<cmp_op><mode>): New pattern.
2237 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2238 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2240 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2242 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2243 (aarch64_emit_sve_predicated_cond): New function.
2244 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2245 (aarch64_emit_unspec_cond_or): Replace with...
2246 (aarch64_emit_sve_or_conds): ...this new function. Use
2247 aarch64_emit_sve_ptrue_op for the individual comparisons and
2248 aarch64_emit_binop to OR them together.
2249 (aarch64_emit_inverted_unspec_cond): Replace with...
2250 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2251 aarch64_emit_sve_ptrue_op for the comparison and
2252 aarch64_emit_unop to invert the result.
2253 (aarch64_expand_sve_vec_cmp_float): Update after the above
2254 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2256 2018-05-07 Nathan Sidwell <nathan@acm.org>
2258 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2259 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2260 (Backwards Compatibility): Likewise.
2262 2018-05-07 Luis Machado <luis.machado@linaro.org>
2266 2018-05-07 Luis Machado <luis.machado@linaro.org>
2268 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2269 <prefetch_dynamic_strides>: New const bool field.
2270 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2271 prefetch_dynamic_strides.
2272 (exynosm1_prefetch_tune): Likewise.
2273 (thunderxt88_prefetch_tune): Likewise.
2274 (thunderx_prefetch_tune): Likewise.
2275 (thunderx2t99_prefetch_tune): Likewise.
2276 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2277 (aarch64_override_options_internal): Update to set
2278 PARAM_PREFETCH_DYNAMIC_STRIDES.
2279 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2280 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2281 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2282 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2283 prefetch-dynamic-strides setting.
2285 2018-05-07 Luis Machado <luis.machado@linaro.org>
2287 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2288 <minimum_stride>: New const int field.
2289 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2290 minimum_stride field.
2291 (exynosm1_prefetch_tune): Likewise.
2292 (thunderxt88_prefetch_tune): Likewise.
2293 (thunderx_prefetch_tune): Likewise.
2294 (thunderx2t99_prefetch_tune): Likewise.
2295 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2296 (aarch64_override_options_internal): Update to set
2297 PARAM_PREFETCH_MINIMUM_STRIDE.
2298 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2299 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2300 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2301 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2302 stride is constant and is below the minimum stride threshold.
2304 2018-05-07 Luis Machado <luis.machado@linaro.org>
2306 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2309 2018-05-07 Luis Machado <luis.machado@linaro.org>
2311 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2312 <prefetch_dynamic_strides>: New const bool field.
2313 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2314 prefetch_dynamic_strides.
2315 (exynosm1_prefetch_tune): Likewise.
2316 (thunderxt88_prefetch_tune): Likewise.
2317 (thunderx_prefetch_tune): Likewise.
2318 (thunderx2t99_prefetch_tune): Likewise.
2319 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2320 (aarch64_override_options_internal): Update to set
2321 PARAM_PREFETCH_DYNAMIC_STRIDES.
2322 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2323 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2324 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2325 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2326 prefetch-dynamic-strides setting.
2328 2018-05-07 Luis Machado <luis.machado@linaro.org>
2330 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2331 <minimum_stride>: New const int field.
2332 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2333 minimum_stride field.
2334 (exynosm1_prefetch_tune): Likewise.
2335 (thunderxt88_prefetch_tune): Likewise.
2336 (thunderx_prefetch_tune): Likewise.
2337 (thunderx2t99_prefetch_tune): Likewise.
2338 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2339 (aarch64_override_options_internal): Update to set
2340 PARAM_PREFETCH_MINIMUM_STRIDE.
2341 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2342 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2343 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2344 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2345 stride is constant and is below the minimum stride threshold.
2347 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2350 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2351 the type is addressable. Don't force op into register if it has
2354 2018-05-05 Roland McGrath <mcgrathr@google.com>
2357 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2358 any section for which we don't know a specific type it should have,
2359 regardless of name. Previously this was done only for the exact
2360 names ".init_array", ".fini_array", and ".preinit_array".
2361 (default_elf_asm_named_section): Add comment about
2362 relationship with default_section_type_flags and SECTION_NOTYPE.
2363 (get_section): Don't consider it a type conflict if one side has
2364 SECTION_NOTYPE and the other doesn't, as long as neither has the
2365 SECTION_BSS et al used in the default_section_type_flags logic.
2367 2018-05-05 Tom de Vries <tom@codesourcery.com>
2370 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2371 (workaround_barsyncs): New function.
2372 (nvptx_reorg): Use workaround_barsyncs.
2373 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2374 (define_expand "nvptx_membar_cta"): New define_expand.
2375 (define_insn "*nvptx_membar_cta"): New insn.
2377 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2379 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2380 To improve optimization opportunities.
2381 * builtin-types.def: The new needed builtin types for the above.
2383 2018-05-04 Richard Biener <rguenther@suse.de>
2385 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2386 * gimple-ssa-store-merging.c
2387 (imm_store_chain_info::output_merged_store): Remove redundant create,
2388 release split_store vector contents on failure.
2389 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2390 scalar stmt vector on cache hit.
2392 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2394 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2396 * config.gcc (powerpc-xilinx-eabi*): Remove.
2397 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2399 (fusion_addis_mem_combo_load): Ditto.
2400 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2402 (rs6000_cpu_cpp_builtins): Ditto.
2403 * config/rs6000/rs6000-linux.c
2404 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2405 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2406 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2408 (rs6000_setup_reg_addr_masks): Ditto.
2409 (rs6000_init_hard_regno_mode_ok): Ditto.
2410 (rs6000_option_override_internal): Ditto.
2411 (legitimate_lo_sum_address_p): Ditto.
2412 (rs6000_legitimize_address): Ditto.
2413 (rs6000_legitimize_reload_address): Ditto.
2414 (rs6000_legitimate_address_p): Ditto.
2415 (abi_v4_pass_in_fpr): Ditto.
2416 (setup_incoming_varargs): Ditto.
2417 (rs6000_gimplify_va_arg): Ditto.
2418 (rs6000_split_multireg_move): Ditto.
2419 (rs6000_savres_strategy): Ditto.
2420 (rs6000_emit_prologue_components): Ditto.
2421 (rs6000_emit_epilogue_components): Ditto.
2422 (rs6000_emit_prologue): Ditto.
2423 (rs6000_emit_epilogue): Ditto.
2424 (rs6000_elf_file_end): Ditto.
2425 (rs6000_function_value): Ditto.
2426 (rs6000_libcall_value): Ditto.
2427 * config/rs6000/rs6000.h: Ditto.
2428 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2429 (TARGET_MINMAX): ... this. New.
2430 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2431 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2432 (*movsi_internal1_single): Delete.
2433 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2434 mfpu=, mxilinx-fpu): Delete.
2435 * config/rs6000/singlefp.h: Delete.
2436 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2437 * config/rs6000/t-rs6000: Ditto.
2438 * config/rs6000/t-xilinx: Delete.
2439 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2440 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2441 (VStype_simple): Delete.
2442 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2443 * config/rs6000/xfpu.h: Delete.
2444 * config/rs6000/xfpu.md: Delete.
2445 * config/rs6000/xilinx.h: Delete.
2446 * config/rs6000/xilinx.opt: Delete.
2447 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2448 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2450 2018-05-04 Tom de Vries <tom@codesourcery.com>
2453 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2456 2018-05-04 Richard Biener <rguenther@suse.de>
2459 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2460 (expand_complex_div_wide): Likewise.
2461 (expand_complex_operations_1): Likewise.
2462 (expand_complex_libcall): Preserve EH info of the original stmt.
2463 (tree_lower_complex): Handle removed blocks.
2464 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2465 on complex multiplication and division libcall builtins.
2467 2018-05-04 Richard Biener <rguenther@suse.de>
2470 * fold-const.c (negate_expr_p): Restrict negation of operand
2471 zero of a division to when we know that can happen without
2473 (fold_negate_expr_1): Likewise.
2475 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2478 * real.h (real_nextafter): Declare.
2479 * real.c (real_nextafter): New function.
2480 * fold-const-call.c (fold_const_nextafter): New function.
2481 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2482 CASE_CFN_NEXTTOWARD.
2483 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2484 even when arg1_mode is different from arg0_mode.
2486 2018-05-03 Nathan Sidwell <nathan@acm.org>
2488 * doc/extend.texi (Deprecated Features): Remove
2490 (Backwards Compatibility): Likewise.
2491 * doc/invoke.texi (C++ Language Options): Likewise.
2492 (C++ Dialect Options): Likewise.
2494 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2497 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2498 _mm512_mask_mullox_epi64): New intrinsics.
2500 2018-05-03 Tom de Vries <tom@codesourcery.com>
2503 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2504 dump files): Add offload-tree.
2506 2018-05-03 Richard Biener <rguenther@suse.de>
2508 PR tree-optimization/85615
2509 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2510 to loops not nested in BBs loop father to avoid creating multi-entry
2513 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2515 PR tree-optimization/70291
2516 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2517 arguments. Change return type to tree. Emit libcall as a new
2518 statement rather than replacing existing one when inplace_p is true.
2519 (expand_complex_multiplication_components): New function.
2520 (expand_complex_multiplication): Expand floating-point complex
2521 multiplication using the above.
2522 (expand_complex_division): Rename inner_type parameter to type.
2523 Update expand_complex_libcall call-site.
2524 (expand_complex_operations_1): Update expand_complex_multiplication
2525 and expand_complex_division call-sites.
2527 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2530 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2531 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2532 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2533 the highest significant bit of the shift count mask is clear. In
2534 check whether and[sq]i3 is needed verify that all significant bits
2535 of the shift count other than the highest are set.
2537 2018-05-02 Tom de Vries <tom@codesourcery.com>
2540 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2541 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2542 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2543 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2544 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2545 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2546 __builtin_goacc_parlevel_size.
2548 2018-05-02 Richard Biener <rguenther@suse.de>
2550 PR tree-optimization/85597
2551 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2552 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2555 2018-05-02 Tom de Vries <tom@codesourcery.com>
2558 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2559 dump files): Add ltrans-tree.
2561 2018-05-02 Tom de Vries <tom@codesourcery.com>
2564 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2565 dump files): Add wpa-ipa.
2567 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2569 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2570 powerpc*-*-linux*paired* target.
2571 * config/rs6000/750cl.h: Delete.
2572 * config/rs6000/paired.h: Delete.
2573 * config/rs6000/paired.md: Delete.
2574 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2576 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2577 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2578 comment. Remove paired float support.
2579 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2580 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2582 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2583 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2585 * config/rs6000/rs6000.c: Remove paired float support.
2586 (paired_expand_vector_init, paired_expand_vector_move,
2587 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2588 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2589 paired_expand_builtin, paired_expand_predicate_builtin,
2590 paired_init_builtins): Delete.
2591 * config/rs6000/rs6000.h: Remove paired float support.
2592 * config/rs6000/rs6000.md: Remove paired float support.
2593 (move_from_CR_ov_bit): Delete.
2594 * config/rs6000/rs6000.opt (mpaired): Delete.
2595 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2596 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2598 2018-05-02 Richard Biener <rguenther@suse.de>
2601 * gimplify.c (gimplify_save_expr): When in SSA form allow
2602 SAVE_EXPRs to compute to SSA vars.
2604 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2607 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2608 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2609 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2610 clobber operands[2], instead use a new pseudo. Formatting fixes.
2612 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2614 PR tree-optimization/85586
2615 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2616 exit early for statements in the same group if the accesses are
2619 2018-05-02 Tom de Vries <tom@codesourcery.com>
2622 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2625 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2627 PR tree-optimization/85143
2628 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2630 2018-05-01 Tom de Vries <tom@codesourcery.com>
2633 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2634 not found" error message.
2636 2018-05-01 Tom de Vries <tom@codesourcery.com>
2639 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2640 * vec.c (test_ordered_remove_if): New function.
2641 (vec_c_tests): Call test_ordered_remove_if.
2642 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2643 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2644 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2645 VEC_ORDERED_REMOVE_IF.
2647 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2649 PR tree-optimization/82665
2650 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2651 pointer subtraction where arguments come from a memchr call.
2653 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2655 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2656 --push-state --as-needed and --pop-state instead of --as-needed and
2657 --no-as-needed if ld supports it.
2658 * configure: Regenerated.
2661 * doc/install.texi2html: Replace _002d with - and _002a with * in
2662 generated html files using sed.
2664 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2667 * gcc-rich-location.c (blank_line_before_p): New function.
2668 (use_new_line): New function.
2669 (gcc_rich_location::add_fixit_insert_formatted): New function.
2670 * gcc-rich-location.h
2671 (gcc_rich_location::add_fixit_insert_formatted): New function.
2673 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2675 * selftest.c (assert_streq): Rename "expected" and "actual" to
2676 "val1" and "val2". Extend NULL-handling to cover both inputs
2677 symmetrically, while still requiring both to be non-NULL for a pass.
2678 * selftest.h (assert_streq): Rename "expected" and "actual" to
2680 (ASSERT_EQ): Likewise.
2681 (ASSERT_EQ_AT): Likewise.
2682 (ASSERT_KNOWN_EQ): Likewise.
2683 (ASSERT_KNOWN_EQ_AT): Likewise.
2684 (ASSERT_NE): Likewise.
2685 (ASSERT_MAYBE_NE): Likewise.
2686 (ASSERT_MAYBE_NE_AT): Likewise.
2687 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2688 the assertion to pass.
2689 (ASSERT_STREQ_AT): Likewise.
2691 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2693 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2694 interaction with -pie.
2696 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2698 * selftest.h: Fix alphabetization of per-source-file selftest
2701 2018-04-30 Jason Merrill <jason@redhat.com>
2703 PR c++/61982 - dead stores to destroyed objects.
2704 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2707 2018-04-30 Jason Merrill <jason@redhat.com>
2709 * tree.c (build_clobber): New.
2710 * tree.h: Declare it.
2711 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2713 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2715 * diagnostic-show-locus.c (layout::layout): Update for
2716 location_get_source_line returning a char_span.
2717 (struct char_span): Move to input.h.
2718 (struct correction): Update for fields in char_span becoming
2720 (struct source_line): Update for location_get_source_line
2721 returning a char_span.
2722 (layout::print_line): Likewise.
2723 * edit-context.c (edited_file::print_content): Likewise.
2724 (edited_file::print_diff_hunk): Likewise.
2725 (edited_file::print_run_of_changed_lines): Likewise.
2726 (edited_file::get_num_lines): Likewise.
2727 (edited_line::edited_line): Likewise.
2728 * final.c (asm_show_source): Likewise.
2729 * input.c (location_get_source_line): Convert return type
2730 from const char * to char_span, losing the final "line_len"
2732 (dump_location_info): Update for the above.
2733 (get_substring_ranges_for_loc): Likewise. Use a char_span
2734 when handling the literal within the line.
2735 (test_reading_source_line): Update for location_get_source_line
2736 returning a char_span.
2737 * input.h (class char_span): Move here from
2738 diagnostic-show-locus.c, converting from a struct to a class.
2739 Make data members private.
2740 (char_span::operator bool): New.
2741 (char_span::length): New.
2742 (char_span::get_buffer): New.
2743 (char_span::operator[]): New.
2744 (char_span::subspan): Make const.
2745 (char_span::xstrdup): New.
2746 (location_get_source_line): Convert return type from const char *
2747 to char_span, losing the final "line_size" param.
2749 2018-04-30 Jan Hubicka <jh@suse.cz>
2751 * lto-wrapper.c (ltrans_priorities): New static var.
2752 (cmp_priority): New.
2753 (run_gcc): Read priorities and if doing parallel build order
2754 the Makefile by them.
2756 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2758 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2760 2018-04-30 Richard Biener <rguenther@suse.de>
2762 * tree-cfg.c (verify_address): Remove base argument, add
2763 flag whether to check TREE_ADDRESSABLE and do that.
2764 (verify_expr): Remove.
2765 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2766 (verify_gimple_assign_single): Likewise.
2767 (verify_gimple_switch): Likewise.
2768 (verify_expr_location_1): Dereference tp once. Add (disabled)
2769 piece from verify_expr.
2770 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2772 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2774 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2776 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2778 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2779 (small_data_pattern): Likewise.
2780 (arc_rewrite_small_data): Likewise.
2781 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2782 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2783 (get_symbol_alignment): New function.
2784 (legitimate_small_data_address_p): Likewise.
2785 (legitimate_scaled_address): Update, call
2786 legitimate_small_data_address_p.
2787 (output_sdata): New static variable.
2788 (arc_print_operand): Update how we handle small data operands.
2789 (arc_print_operand_address): Likewise.
2790 (arc_legitimate_address_p): Update, use
2791 legitimate_small_data_address_p.
2792 (arc_rewrite_small_data_p): Remove.
2793 (arc_rewrite_small_data_1): Likewise.
2794 (arc_rewrite_small_data): Likewise.
2795 (small_data_pattern): Likewise.
2796 (compact_sda_memory_operand): Update to use
2797 legitimate_small_data_address_p and get_symbol_alignment.
2798 (prepare_move_operands): Don't rewite sdata pattern.
2799 (prepare_extend_operands): Remove.
2800 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2802 (zero_extendqisi2): Likewise.
2803 (zero_extendhisi2): Likewise.
2804 (extendqihi2): Likewise.
2805 (extendqisi2): Likewise.
2806 (extendhisi2): Likewise.
2810 * config/arc/constraints.md (Usd): Change it to memory constraint.
2812 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2814 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2815 as source of std instructions.
2816 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2817 allow 6-bit constants as source for store instructions.
2818 (movdi_insn): Update instruction pattern to allow 6-bit constants
2819 as source for store instructions.
2821 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2823 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2825 2018-04-30 Nathan Sidwell <nathan@acm.org>
2826 Sandra Loosemore <sandra@codesourcery.com>
2828 * dumpfile.c (dump_open): Allow '-' for stdout.
2829 * doc/invoke.texi (Developer Options): Document dump filename
2830 determination early. Document stdin/stdout selection.
2832 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2834 Microblaze Target: PIC data text relative
2836 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2837 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2839 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2840 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2841 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2842 New addressing mode for data-text relative position indepenedent code.
2843 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2844 'ADDRESS_SYMBOLIC_TXT_REL'.
2845 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2846 (microblaze_legitimate_pic_operand): Exclude function calls from
2847 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2848 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2850 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2851 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2852 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2853 for 'address + offset'.
2854 (microblaze_expand_prologue): Add new function prologue call for
2856 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2857 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2858 table in case of TARGET_PIC_DATA_TEXT_REL.
2859 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2860 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2861 Add new macros 'UNSPEC_TEXT',
2862 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2863 + exclude function calls from 'UNSPEC_PLT' in case of data text
2865 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2866 new target hook for generating address diff vector tables in case of
2868 * doc/tm.texi : Regenerate.
2869 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2870 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2871 of addr diff vector generation.
2872 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2873 target hook definition.
2874 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2875 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2876 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2877 option for data text relative.
2879 2018-04-30 Richard Biener <rguenther@suse.de>
2881 * tree-chrec.h (evolution_function_is_constant_p): Remove
2883 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2885 2018-04-30 Richard Biener <rguenther@suse.de>
2888 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2890 2018-04-30 Richard Biener <rguenther@suse.de>
2892 PR tree-optimization/28364
2893 PR tree-optimization/85275
2894 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2895 copying first exit test.
2897 2018-04-28 Mark Wielaard <mark@klomp.org>
2899 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2901 (dwarf_AT): Handle DW_AT_addr_base.
2902 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2904 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2907 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2908 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2909 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2910 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2912 2018-04-28 Richard Biener <rguenther@suse.de>
2914 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2915 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2916 to reflect use. Only add interesting stmts.
2918 2018-04-27 Martin Jambor <mjambor@suse.cz>
2921 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2922 the jump function allows for passing through aggregate values.
2924 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2926 * input.h (in_system_header_at): Convert from macro to inline
2928 (from_macro_expansion_at): Likewise.
2929 (from_macro_definition_at): Likewise.
2931 2018-04-27 Jeff Law <law@redhat.com>
2933 * config.gcc: Mark tile* targets as deprecated/obsolete.
2935 2018-04-27 Richard Biener <rguenther@suse.de>
2937 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2940 2018-04-27 Richard Biener <rguenther@suse.de>
2942 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2944 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2946 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2947 with Yd constraint. Set "preferred_for_speed" attribute from
2948 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2950 (*movdi_internal): Ditto.
2951 (movti_interunit splitters): Remove
2952 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2953 (movdi_interunit splitters): Ditto.
2954 * config/i386/constraints.md (Ye): Remove.
2955 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2957 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2960 * config/aarch64/constraints.md (Usg): Limit to 31.
2963 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2965 PR tree-optimization/85529
2966 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2967 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2968 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2969 zero extension or masking of the MSB bit.
2970 (optimize_range_tests): Add FIRST_BB argument, pass it through
2971 to optimize_range_tests_var_bound.
2972 (maybe_optimize_range_tests, reassociate_bb): Adjust
2973 optimize_range_tests callers.
2975 2018-04-26 Richard Biener <rguenther@suse.de>
2976 Jakub Jelinek <jakub@redhat.com>
2978 * cgraph.h (symbol_table): Just declare debug method here.
2979 * symtab.c (symbol_table::debug): Define.
2981 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2983 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2985 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2987 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2988 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2989 (*movdi_internal): Substitute Yi and Yj constraint with x
2990 and Ym and Yn constraint with y constraint. Update "isa"
2991 attribute and set "preferred_for_speed" attribute from
2992 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2993 (*movsi_internal): Ditto.
2994 (*movdf_internal): Ditto.
2995 (*movsf_internal): Ditto.
2996 (*zero_extendsidi2): Ditto.
2997 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2998 (sse2_loadld): Ditto.
2999 (*vec_extract<ssevecmodelower>_0): Ditto.
3000 (*vec_extractv4si_0_zext_sse4): Ditto.
3001 (vec_concatv2di): Ditto.
3002 (*vec_dup<mode>): Ditto.
3003 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3004 * config/i386/constraints.md (Yi): Remove.
3009 2018-04-26 Nathan Sidwell <nathan@acm.org>
3011 * dumpfile.c (dump_open): New.
3012 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3013 (dump_finish): Detect stdio/stderr by value not name.
3015 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3017 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3019 2018-04-26 Tom de Vries <tom@codesourcery.com>
3022 * config/nvptx/nvptx.c (verify_neutering_jumps)
3023 (verify_neutering_labels): New function
3024 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3026 2018-04-26 Tom de Vries <tom@codesourcery.com>
3029 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3030 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3033 2018-04-26 Richard Biener <rguenther@suse.de>
3034 Tom de Vries <tom@codesourcery.com>
3037 * lto-streamer-out.c (output_function): Fixup loops if required to match
3038 discovery done in the reader.
3040 2018-04-26 Richard Biener <rguenther@suse.de>
3042 PR tree-optimization/85116
3043 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3044 have a loop exit from the single latch predecessor. Remove
3045 case of header with just condition.
3046 (ch_base::copy_headers): Exclude infinite loops from any
3048 (pass_ch::execute): Record exits.
3050 2018-04-26 Richard Biener <rguenther@suse.de>
3052 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3053 prologue cost vector and pass it to vect_get_load_cost.
3054 (vect_get_peeling_costs_all_drs): Likewise.
3055 (vect_peeling_hash_get_lowest_cost): Likewise.
3056 (vect_enhance_data_refs_alignment): Likewise.
3058 2018-04-26 Richard Biener <rguenther@suse.de>
3061 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3062 checking of integer<->pointer conversions.
3063 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3064 sign-/zero-extending pointer types.
3065 (expand_omp_for_static_chunk): Likewise.
3067 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3068 Jean Lee <xiaoyur347@gmail.com>
3070 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3071 (TARGET_ASAN_SHADOW_OFFSET): Define.
3072 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3073 true for -fsanitize=address.
3075 2018-04-25 Mark Wielaard <mark@klomp.org>
3077 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3080 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3082 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3083 than "alu", remove explicit "memory" and "imm_disp" attributes.
3084 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3087 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3088 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3089 gen_lowpart_no_emit.
3091 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3094 * config/i386/i386.c (ix86_expand_builtin): Change memory
3095 operand to XI, extend p0 to Pmode.
3096 * config/i386/i386.md: Change unspec volatile and operand
3097 1 mode to XI, change operand 0 mode to P.
3099 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3101 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3102 GET_MODE_MASK before any checking.
3103 (nds32_can_use_bset_p): Likewise.
3104 (nds32_can_use_btgl_p): Likewise.
3106 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3108 * config/nds32/nds32-doubleword.md: New define_split pattern for
3109 illegal register number.
3111 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3113 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3115 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3117 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3119 2018-04-25 Richard Biener <rguenther@suse.de>
3121 * lto-streamer.h (LTO_major_version): Bump to 8.
3123 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3125 * BASE-VER: Set to 9.0.0.
3127 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3129 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3130 in __abskf2 and __powikf2.
3132 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3135 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3136 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3137 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3138 Use the above on operand 2. Reindent.
3139 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3141 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3144 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3146 * config/i386/i386.opt (mcet): Removed.
3147 * doc/install.texi: Remove -mcet documentation.
3148 * doc/invoke.texi: Likewise.
3150 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3153 * doc/install.texi: Remove -mcet from bootstrap-cet.
3155 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3158 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3159 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3163 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3164 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3165 containing a CONST_VECTOR.
3167 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3169 * doc/install.texi: Update newlib dependency for nvptx.
3171 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3174 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3175 instead of INTVAL when shifting x left.
3177 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3179 PR tree-optimization/85478
3180 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3181 vect_grouped_store_supported for single element vectors.
3183 2018-04-24 Richard Biener <rguenther@suse.de>
3186 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3187 load cost increase to the case of non-constant step.
3189 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3192 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3193 destination if any_malformed_asm.
3195 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3198 * expr.c (store_field): In the bitfield case, if the value comes from
3199 a function call and is returned in registers by means of a PARALLEL,
3200 do not change the mode of the temporary unless BLKmode and VOIDmode.
3202 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3204 PR rtl-optimization/85423
3205 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3206 dependencies to debug insns when the previous insn is non-debug.
3208 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3210 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3211 enums into a single definition.
3212 (fls): Fix predicates and printing.
3215 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3217 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3218 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3219 and short u6 immediate.
3220 (check_if_valid_sleep_operand): Remove.
3221 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3223 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3225 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3226 flag_always_save_lp condition.
3227 * config/nds32/nds32.opt (malways-save-lp): New option.
3229 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3231 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3232 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3233 * config/nds32/nds32.h
3234 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3235 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3237 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3239 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3240 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3242 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3243 Chung-Ju Wu <jasonwucj@gmail.com>
3245 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3247 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3248 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3250 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3252 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3254 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3256 * config/nds32/nds32-protos.h (nds32_data_alignment,
3257 nds32_local_alignment): Declare.
3258 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3259 nds32_local_alignment): New functions.
3260 (TARGET_CONSTANT_ALIGNMENT): Define.
3261 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3263 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3265 * config/nds32/nds32.c
3266 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3267 (TARGET_MODES_TIEABLE_P): Likewise.
3269 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3271 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3274 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3275 Chung-Ju Wu <jasonwucj@gmail.com>
3277 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3278 for unaligned access.
3279 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3281 * config/nds32/nds32-intrinsic.md: Likewise.
3282 * config/nds32/nds32_intrinsic.h: Likewise.
3283 * config/nds32/nds32.h (nds32_builtins): Likewise.
3284 * config/nds32/nds32.opt (munaligned-access): New option.
3285 * config/nds32/nds32.c (nds32_asm_file_start): Display
3286 flag_unaligned_access status.
3288 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3290 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3291 -mno-relax is present.
3292 * config/riscv/linux.h (LINK_SPEC): Ditto.
3294 2018-04-20 Martin Sebor <msebor@redhat.com>
3297 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3299 (gimple_fold_builtin_stxcpy_chk): Same.
3300 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3302 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3305 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3306 __powikf2 when long double is IEEE 128-bit.
3308 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3310 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3311 step to make sure stack always aligned.
3313 2018-04-20 Carl Love <cel@us.ibm.com>
3316 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3317 size check for arg0.
3319 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3320 Tom de Vries <tom@codesourcery.com>
3323 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3324 Emit insns for calls too.
3325 (nvptx_find_par): Always look for worker-level predecessor insn.
3326 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3328 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3329 (nvptx_process_pars): Propagate frames for calls.
3331 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3334 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3336 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3337 (ix86_handle_option): Don't handle OPT_mibt.
3338 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3340 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3342 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3343 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3344 (ix86_target_macros): Define __CET__ with flag_cf_protection
3345 for -fcf-protection.
3346 * config/i386/i386.c (isa2_opts): Remove -mibt.
3347 * config/i386/i386.h (TARGET_IBT): Removed.
3348 (TARGET_IBT_P): Likewise.
3349 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3350 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3351 * config/i386/i386.opt (mcet): Update help message.
3354 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3355 -mcet as an alias for -mshstk.
3357 2018-04-20 Richard Biener <rguenther@suse.de>
3360 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3361 complexity by forcing a single use of the multiply operand.
3363 2018-04-20 Martin Jambor <mjambor@suse.cz>
3366 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3367 recursion dependency to only apply to non-clones.
3369 2018-04-20 Martin Jambor <mjambor@suse.cz>
3372 * ipa-cp.c (create_specialized_node): Check that clones of
3373 self-recursive edges exist during IPA-CP.
3375 2018-04-19 Toon Moene <toon@moene.org>
3377 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3380 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3382 PR tree-optimization/85467
3383 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3384 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3385 VECTOR_CST element to type.
3387 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3390 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3391 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3392 (builtin_longjmp): Likewise.
3393 (save_stack_nonlocal): New pattern.
3394 (restore_stack_nonlocal): Likewise.
3396 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3399 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3400 Replace ASM_OUTPUT_LABEL with fprintf.
3402 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3405 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3406 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3407 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3408 define __IBT__ and __SHSTK__ for -fcf-protection.
3409 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3411 (ix86_trampoline_init): Likewise.
3412 (x86_output_mi_thunk): Likewise.
3413 (ix86_notrack_prefixed_insn_p): Likewise.
3414 (ix86_option_override_internal): Don't disallow -fcf-protection.
3415 * config/i386/i386.md (rdssp<mode>): Also enable for
3417 (incssp<mode>): Likewise.
3418 (nop_endbr): Likewise.
3419 * config/i386/i386.opt (mcet): Change help message to built-in
3423 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3424 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3425 enable CET built-in functions.
3427 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3429 * common/config/i386/i386-common.c
3430 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3431 OPTION_MASK_ISA_MOVDIRI_UNSET,
3432 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3433 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3434 * config.gcc (movdirintrin.h): New header.
3435 * config/i386/cpuid.h (bit_MOVDIRI,
3436 bit_MOVDIR64B): New bits.
3437 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3439 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3440 (VOID, PVOID, PCVOID)): New function types.
3441 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3442 __builtin_ia32_directstoreu_u64,
3443 __builtin_ia32_movdir64b): New builtins.
3444 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3445 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3447 (ix86_valid_target_attribute_inner_p): Ditto.
3448 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3449 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3450 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3451 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3452 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3453 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3454 (movdiri<mode>, movdir64b_<mode>): New.
3455 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3456 * config/i386/immintrin.h: Include movdirintrin.h.
3457 * config/i386/movdirintrin.h: New file.
3458 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3460 2018-04-19 Richard Biener <rguenther@suse.de>
3463 * cfg.c (clear_bb_flags): When loop state says we have
3464 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3466 2018-04-19 Richard Biener <rguenther@suse.de>
3468 PR tree-optimization/84737
3469 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3470 copying restrict info.
3471 (vect_setup_realignment): Use it.
3472 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3473 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3474 the first DR to all generated stores.
3475 (vectorizable_load): Likewise for loads.
3477 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3479 PR tree-optimization/85446
3480 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3481 the integral and pointer types to have the same precision.
3483 * doc/install.texi: Document --disable-cet being the default and
3486 2018-04-18 Martin Liska <mliska@suse.cz>
3488 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3491 2018-04-18 Martin Liska <mliska@suse.cz>
3494 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3497 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3498 arguments if they are comparable.
3500 2018-04-18 Martin Liska <mliska@suse.cz>
3503 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3506 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3509 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3512 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3513 ENDBR after calling __morestack.
3515 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3518 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3519 by using gcc_base_ver to generate a gcc_driver_version, and use
3520 it when generating GCC_DRIVER_NAME.
3521 * configure: Regenerate.
3523 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3526 * config.gcc: Obsolete powerpc*-*-*spe*.
3528 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3531 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3532 (stabstr_D): Change type of unum from unsigned int to
3533 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3536 2018-04-17 Jim Wilson <jimw@sifive.com>
3539 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3540 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3541 Set arg_pointer_offset after using pretend_args_size.
3543 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3545 PR rtl-optimization/85431
3546 * dse.c (record_store): Ignore zero width stores.
3549 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3550 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3551 __builtin_stack_restore rather than after it.
3552 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3553 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3554 argument instead of virtual_dynamic_stack_rtx.
3556 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3558 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3560 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3561 Add note to error message to explain internal mapping of overloaded
3562 built-in function name to non-overloaded built-in function name.
3563 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3566 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3569 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3570 where the inputs overlap with the output.
3572 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3574 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3575 (=v, v) alternative and explicit "memory" attribute.
3576 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3577 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3579 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3580 "sselog1" type instead of "sselog".
3581 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3582 "sselog". Remove explicit "memory" attribute.
3583 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3584 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3586 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3587 "sselog1" type instead of "sselog". Remove explicit "memory"
3589 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3590 use "sselog1" type instead of "sselog". Remove explicit "memory"
3592 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3593 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3595 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3596 "sselog1" type instead of "sselog". Remove explicit "memory"
3598 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3599 use "sselog1" type instead of "sselog". Remove explicit "memory"
3603 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3606 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3609 2018-04-17 Martin Jambor <mjambor@suse.cz>
3612 * ipa-cp.c (create_specialized_node): Call
3613 expand_all_artificial_thunks if necessary.
3615 2018-04-17 Martin Liska <mliska@suse.cz>
3618 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3619 in message, remote space in between '_G' and '('.
3621 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3624 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3625 avx512f_vmcmp<mode>3<round_saeonly_name>,
3626 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3627 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3628 avx512f_rndscale<mode><round_saeonly_name>,
3629 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3630 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3631 Use %<iptr>2 instead of %2 for -masm=intel.
3632 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3633 avx512f_vcvttss2usi<round_saeonly_name>,
3634 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3636 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3637 avx512f_vcvttsd2usi<round_saeonly_name>,
3638 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3639 Use %q1 instead of %1 for -masm=intel.
3640 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3641 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3642 of %3 for -masm=intel.
3643 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3645 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3647 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3649 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3650 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3652 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3654 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3655 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3656 %g1 and one with %0 and %1.
3657 (avx512er_vmrcp28<mode><round_saeonly_name>,
3658 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3660 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3661 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3662 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3663 of %0 and %{%4%} for -masm=intel.
3664 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3665 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3666 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3667 order of %0 and %{%5%}%{z%} for -masm=intel.
3669 2018-04-17 Jan Hubicka <jh@suse.cz>
3672 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3674 2018-04-17 Martin Liska <mliska@suse.cz>
3677 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3678 for target_clone error message. Make default implementation
3679 clone to be a local declaration.
3680 (separate_attrs): Add new argument and check for an empty
3682 (expand_target_clones): Handle it.
3683 (ipa_target_clone): Make redirection just for target_clones
3686 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3687 Tom de Vries <tom@codesourcery.com>
3690 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3691 tiled basic blocks without omp continue statements.
3693 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3696 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3697 vec_extract expression as having side effects to make sure it gets
3700 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3703 * config/i386/i386.c (get_builtin_code_for_version): Check
3706 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3709 * gcc/config.gcc: Support "skylake".
3710 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3712 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3713 (processor_target_table): Add "skylake".
3714 (ix86_option_override_internal): Add "skylake".
3715 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3716 PROCESSOR_CANNONLAKE.
3717 (get_builtin_code_for_version): Fix priority for
3718 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3719 PROCESSOR_SKYLAKE-AVX512.
3720 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3721 (processor_type): Add PROCESSOR_SKYLAKE.
3723 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3724 Jason Merrill <jason@redhat.com>
3727 * convert.c (convert_to_integer_1): Use direct recursion for
3728 enumeral types and types with a precision less than the number
3729 of bits in their mode.
3731 2018-04-16 Julia Koval <julia.koval@intel.com>
3734 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3735 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3737 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3740 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3741 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3742 and -mno-direct-move.
3744 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3747 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3748 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3749 Ensure negative shifts result in {0}.
3751 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3753 PR rtl-optimization/79916
3754 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3755 regs (if any) to define how to gnerate SD moves when LRA is in
3758 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3760 PR rtl-optimization/85393
3761 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3762 * except.c (expand_dw2_landing_pad_for_region): Make static.
3763 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3764 a label and unconditional jump to old_bb, rather than
3765 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3768 PR rtl-optimization/85376
3769 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3770 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3771 instead of a specific value.
3773 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3774 Bin Cheng <bin.cheng@arm.com>
3776 PR tree-optimization/82965
3777 PR tree-optimization/83991
3778 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3779 by_profile_only parameter.
3780 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3781 information if the loop was predicted to iterate too many times.
3782 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3784 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3787 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3790 2018-04-13 Martin Liska <mliska@suse.cz>
3791 Jakub Jelinek <jakub@redhat.com>
3794 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3795 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3796 * builtins.c (expand_builtin_memory_copy_args): Use
3797 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3798 handle dest_addr == pc_rtx.
3800 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3803 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3804 asked to not generate direct moves.
3805 (fix_trunc<mode>si2_stfiwx): Similar.
3806 (fix_trunc<mode>si2_internal): Similar.
3808 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3811 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3812 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3813 lookup if dest in some wider mode is known to be const0_rtx and
3814 if so, record permanent equivalence for it to be ZERO_EXTEND of
3815 the narrower mode destination.
3817 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3819 * lto-streamer-out.c (output_function): Revert 259346.
3820 * omp-expand.c (expand_oacc_for): Likewise.
3822 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3824 PR rtl-optimization/85354
3825 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3826 * sel-sched.c (sel_global_init): ... here.
3828 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3831 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3832 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3833 mode for PE-COFF targets.
3834 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3835 (i386_pe_asm_lto_end): Likewise.
3836 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3837 (TARGET_ASM_LTO_END): Likewise.
3838 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3839 (i386_pe_asm_lto_start): New function.
3840 (i386_pe_asm_lto_end): Likewise.
3842 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3843 Richard Biener <rguenther@suse.de>
3846 * lto-streamer-out.c (output_function): Fix CFG loop state before
3848 * omp-expand.c (expand_oacc_for): Handle calls to internal
3849 functions like regular functions.
3851 2018-04-12 Richard Biener <rguenther@suse.de>
3854 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3855 for the early LTO debug to properly generate references to it
3856 during DIE emission. Do not re-use that for the skeleton for
3858 (dwarf2out_early_finish): Likewise.
3860 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3863 * config/i386/sse.md
3864 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3865 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3866 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3867 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3868 and output is a reg, avoid creating invalid lowpart subreg, but
3869 instead split into a 512-bit move. Don't split if not AVX512VL,
3870 input is xmm16+ reg and output is a mem.
3871 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3872 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3873 xmm16+ reg and output is a mem.
3875 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3877 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3878 also for flag_dwarf2_cfi_asm.
3880 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3882 PR rtl-optimization/85342
3883 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3884 a bool scalar var inside of the loop instead. Don't try to update
3885 recog_data.operand after failed apply_change_group.
3887 2018-04-12 Tom de Vries <tom@codesourcery.com>
3890 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3891 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3892 array with flexible array member as array without given dimension.
3893 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3894 argument for undefined param to true.
3896 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3899 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3900 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3901 from PowerPC section.
3902 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3903 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3904 help text that is too long.
3905 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3906 help text that is too long.
3907 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3908 help text that is too long.
3910 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3912 * config/alpha/alpha.md (stack_probe_internal): Rename
3913 from "probe_stack". Update all callers.
3915 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3917 PR rtl-optimization/84566
3918 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3919 sched_macro_fuse_insns.
3921 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3924 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3925 (compute_block_dependences): ... from here.
3927 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3929 PR tree-optimization/85331
3930 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3931 from int to HOST_WIDE_INT.
3933 2018-04-11 Martin Jambor <mjambor@suse.cz>
3936 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3937 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3938 not the same as the source val.
3939 (cgraph_edge_brings_value_p): New parameter.
3940 (gather_edges_for_value): Pass destination value to
3941 cgraph_edge_brings_value_p.
3942 (perhaps_add_new_callers): Likewise.
3943 (get_info_about_necessary_edges): Likewise and exclude values brought
3944 only by self-recursive edges.
3945 (create_specialized_node): Redirect only clones of self-calling edges.
3946 (+self_recursive_pass_through_p): New function.
3947 (find_more_scalar_values_for_callers_subset): Use it.
3948 (find_aggregate_values_for_callers_subset): Likewise.
3949 (known_aggs_to_agg_replacement_list): Removed.
3950 (decide_whether_version_node): Re-calculate known constants for all
3951 remaining context clones.
3953 2018-04-11 Richard Biener <rguenther@suse.de>
3956 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3957 from early DWARF output.
3958 (dwarf2out_early_finish): Output line info unconditionally into
3959 early DWARF and add reference to it.
3961 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3964 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3965 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3966 other than V2DFmode using iptr mode attribute.
3967 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3969 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3971 PR rtl-optimization/84659
3972 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3974 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3977 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3979 (output_loc_list): Pass address of a dummy size variable even in the
3980 locview handling loop.
3981 (index_location_lists): Add comment on why skip_loc_list_entry can't
3984 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3987 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3990 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3993 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3994 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3995 and -mstring-compare-inline-limit.
3997 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4000 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4001 for stack clash protection in a register whenever we need it to be in
4004 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4006 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4007 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4009 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4012 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4014 (mlong-double-): Ditto.
4015 * config/rs6000/sysv4.opt (msdata=): Ditto.
4016 (mtls-size=): Ditto.
4018 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4020 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4021 erroneous entries for
4022 "vector int vec_ldl (int, long int *)", and
4023 "vector unsigned int vec_ldl (int, unsigned long int *)".
4024 Add comments and entries for
4025 "vector bool char vec_ldl (int, bool char *)",
4026 "vector bool short vec_ldl (int, bool short *)",
4027 "vector bool int vec_ldl (int, bool int *)",
4028 "vector bool long long vec_ldl (int, bool long long *)",
4029 "vector pixel vec_ldl (int, pixel *)",
4030 "vector long long vec_ldl (int, long long *)",
4031 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4032 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4033 type tree bool_long_long_type_node and correct definition of
4034 bool_V2DI_type_node to make reference to this new type tree.
4035 (rs6000_mangle_type): Replace erroneous reference to
4036 bool_long_type_node with bool_long_long_type_node.
4037 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4038 comments to emphasize sign distinctions for char and int types and
4039 replace RS6000_BTI_bool_long constant with
4040 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4041 use of RS6000_BTI_pixel.
4042 (bool_long_type_node): Remove this macro definition.
4043 (bool_long_long_type_node): New macro definition
4045 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4047 PR rtl-optimization/85300
4048 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4049 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4050 simplify_unary_operation fails.
4052 2018-04-10 Martin Liska <mliska@suse.cz>
4054 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4055 cgraph_edge and ipa_ref.
4057 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4061 * config/i386/sse.md
4062 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4063 computation of the VEC_MERGE selector from mask.
4064 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4065 Fix decoding of the VEC_MERGE selector into mask.
4067 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4069 PR tree-optimization/85286
4070 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4072 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4074 * final.c (final_1): Set insn_last_address as well as
4075 insn_current_address.
4077 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4080 * explow.c (emit_stack_probe): Call validize_mem on memory location
4081 before passing it to gen_probe_stack. Create address operand and
4082 legitimize it for the probe_stack_address case.
4084 2018-04-09 Jan Hubicka <jh@suse.cz>
4087 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4088 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4089 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4091 (free_lang_data): Rebuild type inheritance graph.
4093 2018-04-09 Martin Sebor <msebor@redhat.com>
4095 * invoke.texi (-finline-small-functions): Mention other optimization
4097 (-findirect-inlining, -fpartial-inlining): Same.
4098 (-finline-functions-called-once): Same.
4099 (-freorder-blocks-and-partition): Same.
4101 2018-04-09 Jan Hubicka <jh@suse.cz>
4104 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4105 jumps; choose last target that matches the criteria (i.e.
4106 no partition changes for non-crossing jumps).
4107 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4108 support for redirecting crossing jumps to non-crossing.
4110 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4112 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4113 also for naked functions.
4115 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4117 * config/arc/arc.md (add_shift): New pattern.
4118 (add_shift2): Likewise.
4119 (sub_shift): Likewise.
4120 (sub_shift_cmp0_noout): Likewise.
4121 (compare_si_ashiftsi): Likewise.
4122 (xbfu_cmp0_noout): New combine pattern.
4123 (xbfu_cmp0"): Likewise.
4124 (movsi_set_cc_insn): Place the predicable variant first.
4125 (commutative_binary_cmp0_noout): Remove clobber.
4126 (commutative_binary_cmp0): New pattern.
4127 (noncommutative_binary_cmp0): Likewise.
4128 (noncommutative_binary_cmp0_noout): Likewise.
4129 (noncommutative_binary_comparison_result_used): Removed.
4130 (rsub_cmp0): New pattern.
4131 (rsub_cmp0_noout): Likewise.
4132 (extzvsi): Changed, keep only meaningful variants.
4133 (SQH, SEZ): New iterators.
4134 (SQH_postfix): New mode attribute.
4135 (SEZ_prefix): New code attribute.
4136 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4137 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4138 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4140 (noncommutative_operator): Check the availability of barrel
4143 2018-04-09 Richard Biener <rguenther@suse.de>
4145 PR tree-optimization/85284
4146 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4147 Only use the niter constraining form of simple_iv when the exit
4150 2018-04-09 Tom de Vries <tom@codesourcery.com>
4153 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4154 (define_expand "*memory_barrier"): New define_expand.
4155 (define_insn "memory_barrier"): New insn.
4157 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4159 PR rtl-optimization/80463
4160 PR rtl-optimization/83972
4161 PR rtl-optimization/83480
4163 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4164 correct producer for the insn.
4165 (tidy_control_flow): Fixup seqnos in case of debug insns.
4167 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4169 PR rtl-optimization/83913
4171 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4172 different sched-times when merging exprs.
4174 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4176 PR rtl-optimization/83962
4178 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4179 tidy_fallthru_edge and tidy_control_flow.
4181 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4183 PR rtl-optimization/83530
4185 * sel-sched.c (force_next_insn): New global variable.
4186 (remove_insn_for_debug): When force_next_insn is true, also leave only
4187 next insn in the ready list.
4188 (sel_sched_region): When the region wasn't scheduled, make another pass
4189 over it with force_next_insn set to 1.
4191 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4193 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4195 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4196 for interrupt control.
4197 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4198 functions for interrupt control.
4199 * config/nds32/nds32-intrinsic.md: Likewise.
4200 * config/nds32/nds32_intrinsic.h: Likewise.
4201 * config/nds32/nds32.h (nds32_builtins): Likewise.
4203 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4205 * config/nds32/nds32.c (nds32_init_machine_status,
4206 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4207 strict_aligned_p field.
4208 (nds32_expand_to_rtl_hook): New function.
4209 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4210 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4212 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4213 Chung-Ju Wu <jasonwucj@gmail.com>
4215 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4216 * config/nds32/nds32-n7.md: New file.
4217 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4218 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4220 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4221 * config/nds32/nds32.md (pipeline_model): Add n7.
4222 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4223 * config/nds32/pipelines.md: Include n7 settings.
4225 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4226 Chung-Ju Wu <jasonwucj@gmail.com>
4228 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4229 * config/nds32/nds32-e8.md: New file.
4230 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4231 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4233 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4234 * config/nds32/nds32.md (pipeline_model): Add e8.
4235 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4236 * config/nds32/pipelines.md: Include e8 settings.
4238 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4239 Chung-Ju Wu <jasonwucj@gmail.com>
4241 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4242 * config/nds32/nds32-n8.md: New file.
4243 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4244 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4246 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4247 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4248 * config/nds32/nds32.md (pipeline_model): Add n8.
4249 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4250 * config/nds32/pipelines.md: Include n8 settings.
4252 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4253 Chung-Ju Wu <jasonwucj@gmail.com>
4255 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4256 * config/nds32/nds32-n9-2r1w.md: New file.
4257 * config/nds32/nds32-n9-3r2w.md: New file.
4258 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4259 nds32_register_ports): New or modify for cpu n9.
4260 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4262 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4263 * config/nds32/nds32-utils.c: New file.
4264 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4265 TARGET_MUL_SLOW): Define.
4266 * config/nds32/nds32.md (pipeline_model): New attribute.
4267 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4268 New options that support cpu n9.
4269 * config/nds32/pipelines.md: Include n9 settings.
4270 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4272 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4274 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4275 information if necessary.
4276 (output_cond_branch_compare_zero): Likewise.
4277 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4278 (nds32_target_alignment): Refine for alignment.
4279 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4280 (FUNCTION_BOUNDARY): Modify.
4281 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4283 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4285 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4287 * config/nds32/constants.md (unspec_volatile_element): Add values for
4288 TLB operation and data prefetch.
4289 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4290 functions for TLB operation and data prefetch.
4291 * config/nds32/nds32-intrinsic.md: Likewise.
4292 * config/nds32/nds32_intrinsic.h: Likewise.
4293 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4294 (nds32_print_operand): Likewise.
4295 * config/nds32/nds32.h (nds32_builtins): Likewise.
4297 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4298 Andrew Pinski <pinsika@gcc.gnu.org>
4301 * match.pd: Use constant_boolean_node of correct type instead of
4302 boolean_true_node or boolean_false_node for simplifying
4303 pointer comparisons to zero.
4305 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4307 PR tree-optimization/80021
4308 * tree.c (verify_type_variant): Make error call in verify_variant_match
4309 translatable and remove final full stop.
4311 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4313 * config/nds32/constants.md (unspec_volatile_element): Add
4314 UNSPEC_VOLATILE_EH_RETURN.
4315 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4316 nds32_output_stack_pop): Support dwarf exception handling process.
4317 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4318 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4319 exception handling process.
4320 (nds32_compute_stack_frame): Likewise.
4321 (nds32_return_addr_rtx): Likewise.
4322 (nds32_initial_elimination_offset): Likewise.
4323 (nds32_expand_prologue): Likewise.
4324 (nds32_expand_epilogue): Likewise.
4325 (nds32_dynamic_chain_address): New function.
4326 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4328 (DYNAMIC_CHAIN_ADDRESS): Define.
4329 (EH_RETURN_DATA_REGNO): Define.
4330 (EH_RETURN_STACKADJ_RTX): Define.
4331 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4332 patterns for dwarf exception handling.
4334 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4336 * config/nds32/nds32.h: Clean up obsolete macros.
4338 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4340 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4341 Add enum values for particular instructions.
4342 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4343 particular intrinsic functions.
4344 * config/nds32/nds32-intrinsic.md: Likewise.
4345 * config/nds32/nds32_intrinsic.h: Likewise.
4346 * config/nds32/nds32.h (nds32_builtins): Likewise.
4347 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4348 (btst, ave): New patterns for particular instructions.
4350 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4352 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4353 Add enum values for atomic load/store and memory sync.
4354 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4356 * config/nds32/nds32-intrinsic.md: Likewise.
4357 * config/nds32/nds32_intrinsic.h: Likewise.
4358 * config/nds32/nds32.h (nds32_builtins): Likewise.
4360 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4362 PR tree-optimization/85257
4363 * fold-const.c (native_encode_vector): If not all elts could fit
4364 and off is -1, return 0 rather than offset.
4365 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4366 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4367 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4368 adjust buffer in native_interpret_expr call.
4370 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4372 * config/nds32/constants.md (unspec_volatile_element): Add cache
4373 control enum values.
4374 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4375 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4376 * config/nds32/nds32.c (nds32_cctl_names): New.
4377 (nds32_print_operand): Handle cache control register names.
4378 * config/nds32/nds32.h (nds32_builtins): New enum values.
4379 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4381 * config/nds32/nds32.md (type): Add mmu.
4382 * config/nds32/pipelines.md (simple_insn): Add mmu.
4384 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4386 * config/nds32/nds32.md (type): Remove call.
4387 * config/nds32/pipelines.md (simple_insn): Likewise.
4389 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4391 * config/nds32/constants.md (unspec_volatile_element): Add
4392 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4393 UNSPEC_VOLATILE_FMFCFG.
4394 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4395 description for fmfcfg and fmfcsr.
4396 (bdesc_1arg): Add fmtcsr.
4397 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4398 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4399 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4400 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4401 unspec_fmfcfg): New patterns.
4402 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4403 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4404 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4405 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4406 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4407 __nds32__fmfcfg): Define.
4409 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4411 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4412 intrinsic register names.
4413 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4414 intrinsic register enum values and macros.
4416 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4418 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4419 for load/store addressing form.
4420 (nds32_print_operand_address): Likewise.
4422 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4425 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4426 based on LABEL_REF. Remove useless assertion.
4427 (pic_address_needs_scratch): Fix formatting.
4428 (sparc_legitimize_pic_address): Minor tweaks.
4429 (sparc_delegitimize_address): Adjust assertion accordingly.
4430 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4431 into symbolic_operand.
4432 (movsi_high_pic_label_ref): Likewise.
4433 (movsi_lo_sum_pic_label_ref): Likewise.
4434 (movdi_pic_label_ref): Likewise.
4435 (movdi_high_pic_label_ref): Likewise.
4436 (movdi_lo_sum_pic_label_ref): Likewise.
4438 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4440 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4441 custom LIB_SPEC setup.
4443 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4444 Kito Cheng <kito.cheng@gmail.com>
4446 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4447 * config/riscv/freebsd.h: New.
4449 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4451 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4452 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4455 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4456 Kito Cheng <kito.cheng@gmail.com>
4458 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4459 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4460 * config/nds32/nds32-protos.h (nds32_output_call,
4461 nds32_output_return): Declare.
4462 * config/nds32/nds32.md: Refine all the call and return patterns.
4464 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4467 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4468 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4470 PR rtl-optimization/84872
4471 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4472 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4475 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4477 * expr.c (copy_blkmode_to_reg): Revert 254862.
4478 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4480 2018-04-06 Richard Biener <rguenther@suse.de>
4483 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4484 after seeing a component reference with an adjacent field. Treat
4485 refs to arrays at struct end of external decls similar to
4486 refs to unconstrained commons.
4488 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4491 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4492 look through SAVE_EXPRs with non-side-effects argument. Adjust
4494 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4497 2018-04-06 Richard Biener <rguenther@suse.de>
4500 * alias.c (find_base_term): New wrapper around find_base_term
4501 unwinding CSELIB_VAL_PTR changes.
4502 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4505 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4507 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4509 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4510 constant definitions.
4511 ("nop"): lr 0,0 -> nopr r0
4512 ("nop_lr0", "nop_lr1"): New insn definitions.
4514 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4516 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4517 NDS32_V3PUSH_AVAILABLE_P macro.
4519 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4520 Chung-Ju Wu <jasonwucj@gmail.com>
4522 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4523 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4524 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4525 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4526 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4527 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4528 * config/nds32/constraints.md: New constraints and checking for hard
4529 float configuration.
4530 * config/nds32/iterators.md: New mode iterator and attribute for hard
4531 float configuration.
4532 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4534 * config/nds32/nds32-fpu.md: New file.
4535 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4536 deal with hard float code generation.
4537 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4539 (abi_type, float_reg_number): New enum type.
4540 * config/nds32/nds32-predicates.c: New predicates for hard float.
4541 * config/nds32/nds32-protos.h: Declare functions for hard float.
4542 * config/nds32/nds32.c: Implementation for hard float configuration.
4543 * config/nds32/nds32.h: Definitions for hard float configuration.
4544 * config/nds32/nds32.md: Include hard float machine description and
4545 modify patterns for hard float configuration.
4546 * config/nds32/nds32.opt: New options for hard float configuration.
4547 * config/nds32/predicates.md: New predicates for hard float
4550 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4552 * common/config/nds32/nds32-common.c
4553 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4555 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4558 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4559 CONSTRUCTOR_ELT (ctor, ...)->value.
4561 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4564 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4566 2018-04-05 Tom de Vries <tom@codesourcery.com>
4569 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4572 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4573 Kito Cheng <kito.cheng@gmail.com>
4575 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4576 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4577 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4578 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4580 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4581 Kito Cheng <kito.cheng@gmail.com>
4583 * config/nds32/constraints.md (Ufe): New memory constraint.
4584 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4585 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4586 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4588 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4589 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4591 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4593 * config/nds32/nds32.md: Use optimize_size in the condition for
4594 alu-shift instructions.
4596 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4598 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4600 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4602 * config/nds32/nds32.md (negsi2): Refine pattern.
4604 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4605 Chung-Ju Wu <jasonwucj@gmail.com>
4607 * config/nds32/iterators.md (shift_rotate): New code iterator.
4608 (shift): New code attribute.
4609 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4610 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4611 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4612 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4613 bit-wise operations.
4614 (andsi3, *andsi3): Ditto.
4615 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4616 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4617 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4618 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4619 nds32_ior_operand, nds32_xor_operand): New predicates.
4621 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4623 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4624 (addsi3, subsi3): ... this.
4626 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4628 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4630 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4632 * config/nds32/nds32.md: Adjust indention.
4634 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4636 * config/nds32/nds32.md (feature): New attribute.
4638 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4640 * config/nds32/nds32.md (subtype): New attribute.
4642 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4645 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4646 expansion to perform a bitwise AND of the argument followed by a
4647 boolean negation of the result.
4649 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4651 PR rtl-optimization/84878
4652 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4653 the basic block. Assert the use reference is not artificial and that
4654 it has an associated insn.
4656 2018-04-04 Michael Matz <matz@suse.de>
4658 * builtins.c (compute_objsize): Pass correct operand
4659 to array_at_struct_end_p.
4661 2018-04-04 Richard Biener <rguenther@suse.de>
4664 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4665 from contexts for DINFO_LEVEL_TERSE and below.
4667 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4669 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4670 resiter_operand condition.
4671 * config/nds32/nds32.md (*move<mode>): Ditto.
4673 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4674 Monk Chiang <sh.chiang04@gmail.com>
4676 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4678 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4680 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4682 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4683 Kito Cheng <kito.cheng@gmail.com>
4685 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4686 nds32_cond_code_str, output_cond_branch,
4687 output_cond_branch_compare_zero, nds32_expand_cbranch,
4688 nds32_expand_cstore, nds32_expand_movcc,
4689 nds32_output_cbranchsi4_equality_zero,
4690 nds32_output_cbranchsi4_equality_reg,
4691 nds32_output_cbranchsi4_equality_reg_or_const_int,
4692 nds32_output_cbranchsi4_greater_less_zero: New functions.
4693 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4694 nds32_expand_cstore, nds32_expand_movcc,
4695 nds32_output_cbranchsi4_equality_zero,
4696 nds32_output_cbranchsi4_equality_reg,
4697 nds32_output_cbranchsi4_equality_reg_or_const_int,
4698 nds32_output_cbranchsi4_greater_less_zero): Declare.
4699 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4700 nds32_rimm11s_operand): New predicates.
4701 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4702 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4705 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4707 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4708 * config/nds32/nds32.md: Ditto.
4709 * config/nds32/pipelines.md: Ditto.
4711 2018-04-04 Richard Biener <rguenther@suse.de>
4713 PR tree-optimization/85168
4714 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4715 propagating abnormals.
4717 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4719 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4721 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4722 Kito Cheng <kito.cheng@gmail.com>
4724 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4725 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4726 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4727 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4728 * config/nds32/nds32.md (sibcall_internal): New.
4729 (sibcall_register): Remove.
4730 (sibcall_immediate): Remove.
4731 (sibcall_value_internal): New.
4732 (sibcall_value_register): Remove.
4733 (sibcall_value_immediate): Remove.
4734 * config/nds32/predicates.md (nds32_general_register_operand): New.
4735 (nds32_call_address_operand): New.
4737 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4739 PR rtl-optimization/85167
4740 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4741 bb_defs if *split_p, instead preinitialize it to NULL.
4743 PR tree-optimization/85156
4744 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4745 evaluating the argument multiple times.
4747 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4749 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4751 (_mm_cvtpd_ps): Likewise.
4752 (_mm_cvttpd_epi32): Likewise.
4753 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4754 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4755 vector, pixel, and bool following altivec.h include.
4757 2018-04-03 Martin Sebor <msebor@redhat.com>
4759 * doc/extend.texi (Common Function Attributes): Clarify.
4760 (const attribute): Likewise.
4761 (pure attribute): Likewise.
4763 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4766 * config/i386/i386.c (ix86_expand_vector_set): Use
4767 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4769 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4771 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4772 instructions when changing rounding bits to preserve precision bits
4773 in the x87 control word.
4775 2018-04-03 Martin Liska <mliska@suse.cz>
4777 PR tree-optimization/82491
4778 * rtl.h (strip_offset_and_add): Replace += suboffset with
4779 poly_uint64 () + suboffset.
4781 2018-03-29 Martin Liska <mliska@suse.cz>
4782 Martin Jambor <mjambor@suse.cz>
4785 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4786 param_type is not an integral or pointer type.
4788 2018-04-03 Richard Biener <rguenther@suse.de>
4790 * sese.h (recompute_all_dominators): Remove.
4792 2018-04-02 Martin Sebor <msebor@redhat.com>
4794 * doc/invoke.texi (-Wrestrict): Fix typos.
4796 2018-04-02 Jim Wilson <jimw@sifive.com>
4798 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4799 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4800 (<optab>di3, <optab>si3_extend): Likewise.
4801 (<optab>si3_mask, <optab>si3_mask_1): New.
4802 (<optab>di3_mask, <optab>di3_mask_1): New.
4803 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4804 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4805 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4807 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4809 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4812 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4814 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4815 (nds32_canonicalize_comparison): New function.
4817 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4818 Kito Cheng <kito.cheng@gmail.com>
4819 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4821 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4822 * config/nds32/constants.md (unspec_volatile_element): Add
4823 UNSPEC_VOLATILE_RELAX_GROUP.
4824 * config/nds32/nds32-relax-opt.c: New file.
4825 * config/nds32/nds32-predicates.c
4826 (nds32_symbol_load_store_p): New function.
4827 * config/nds32/nds32-protos.h
4828 (nds32_symbol_load_store_p): Declare function.
4829 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4830 * config/nds32/nds32.c
4831 (nds32_register_pass): New function to register pass.
4832 (nds32_register_passes): New function to register passes.
4833 * config/nds32/nds32.md (relax_group): New pattern.
4834 * config/nds32/nds32.opt (mrelax-hint): New option.
4835 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4837 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4839 * config/nds32/t-nds32: Modify files dependency.
4841 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4843 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4844 (PROFILE_HOOK): Define its implementation.
4846 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4848 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4849 type and 32-bit size.
4851 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4854 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4855 (V_128_256): New mode iterator.
4856 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4857 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4858 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4860 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4861 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4863 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4866 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4867 NaN inputs correctly.
4869 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4872 * config/rs6000/vsx.md (??r): New mode attribute.
4873 (*vsx_mov<mode>_64bit): Use it.
4874 (*vsx_mov<mode>_32bit): Likewise.
4876 2018-03-30 Martin Sebor <msebor@redhat.com>
4878 PR tree-optimization/84818
4879 * builtins.c (check_access): Use warning_n.
4881 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4884 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4886 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4889 2018-03-30 Julia Koval <julia.koval@intel.com>
4892 * x86-tune.def (movx, partial_reg_dependency): Enable for
4895 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4898 * lra-constraints.c (process_alt_operands): Move setting
4899 this_alternative_matches below.
4901 2018-03-29 Martin Liska <mliska@suse.cz>
4904 * doc/invoke.texi: Document how LTO works with debug info.
4905 Describe auto-load support of binutils. Mention 'x86-64'
4906 as valid option value of -march option.
4908 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4910 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4913 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4914 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4915 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4918 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4921 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4922 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4923 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4924 for RS6000_BTM_POWERPC64.
4925 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4926 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4927 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4932 2018-03-28 Carl Love <cel@us.ibm.com>
4935 2017-09-27 Carl Love <cel@us.ibm.com>
4937 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4938 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4939 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4942 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4944 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4945 instead of __vector bool.
4946 (_mm_max_pu8): Likewise.
4947 (_mm_min_pi16): Likewise.
4949 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4952 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4953 (DIVWEUO): Likewise.
4955 (DIVDEUO): Likewise.
4956 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4957 DIVWEUO and DIVDEUO.
4958 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4959 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4960 (div_extend): Likewise.
4961 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4963 (__builtin_divweuo): Likewise.
4964 (__builtin_divdeo): Likewise.
4965 (__builtin_divdeuo): Likewise.
4967 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4970 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4971 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4973 PR tree-optimization/82004
4974 * gimple-match-head.c (optimize_pow_to_exp): New function.
4975 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4976 Don't fold to exp if optimize_pow_to_exp is false.
4978 2018-03-28 Martin Liska <mliska@suse.cz>
4981 * calls.c (initialize_argument_information): Fix trailing space.
4982 * common.opt: Fix typo and provide better explanation for
4983 -fsanitize-coverage option.
4984 * config/i386/i386.opt: Fix typo.
4986 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4987 Martin Liska <mliska@suse.cz>
4990 * gimplify.c (asan_poison_variable): Don't do the check for
4991 gimplify_omp_ctxp here.
4992 (gimplify_decl_expr): Do it here.
4993 (gimplify_target_expr): Likewise.
4995 2018-03-28 Martin Liska <mliska@suse.cz>
4998 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4999 chkp_type_bounds_count if MPX is not enabled.
5001 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5003 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5005 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5008 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5009 function to create the function decl for complex long double
5010 multiply and divide for -mabi=ieeelongdouble.
5011 (init_float128_ieee): Call it.
5013 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5016 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5017 -fcf-protection=branch -mibt.
5018 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5020 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5023 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5025 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5028 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5029 extern array declarations.
5031 2018-03-27 Richard Biener <rguenther@suse.de>
5034 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5035 explicit single_use checks.
5037 2018-03-27 Richard Biener <rguenther@suse.de>
5039 PR tree-optimization/85082
5040 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5043 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5045 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5046 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5047 Turn on fasynchronous-unwind-tables and funwind-tables.
5049 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5052 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5053 (*bmi_blsr_<mode>_ccz): Ditto.
5055 2018-03-26 Tom de Vries <tom@codesourcery.com>
5057 PR tree-optimization/85063
5058 * omp-general.c (offloading_function_p): New function. Factor out
5060 * omp-offload.c (pass_omp_target_link::gate): ... here.
5061 * omp-general.h (offloading_function_p): Declare.
5062 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5063 with attribute omp declare target for offloading functions.
5065 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5067 PR tree-optimization/84005
5068 * tree-data-ref.h (get_base_for_alignment): Declare.
5069 * tree-data-ref.c (get_base_for_alignment_1): New function.
5070 (get_base_for_alignment): Likewise.
5071 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5072 get_base_for_alignment to find a suitable base object, instead
5073 of always using drb->base_address.
5075 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5078 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5079 known size by default.
5081 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5084 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5085 and non BLKmode operands.
5087 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5090 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5091 Clean up attributes.
5093 2018-03-23 Richard Biener <rguenther@suse.de>
5096 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5097 we are going to emit early debug for LTO.
5099 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5102 * function.c (match_asm_constraints_1): Don't optimize if input
5103 doesn't satisfy general_operand predicate for output's mode.
5106 * alias.c (write_dependence_p): Don't require for x_canonicalized
5107 non-VOIDmode if x has VOIDmode.
5110 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5111 just don't try to optimize it rather than assert it never happens.
5113 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5115 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5116 macro expansions for definition of ST_INTERNAL_<mode> and
5117 LD_INTERNAL_<mode> builtins.
5118 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5120 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5122 (altivec_expand_st_builtin): Likewise.
5123 (altivec_expand_builtin): Remove calls to deleted functions.
5124 (rs6000_address_for_altivec): Delete this function.
5125 * config/rs6000/vector.md: Remove expands for
5126 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5128 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5131 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5132 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5133 re-computing once computed.
5134 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5135 (arm_init_machine_status): Initialize
5136 machine->static_chain_stack_bytes.
5138 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5141 * doc/extend.texi: Add four new prototypes for vec_ld.
5142 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5143 definitions for more logical presentation.
5144 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5145 entries for V1TI variants of __builtin_altivec_ld builtin.
5146 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5147 handling of V1TI variant of LVX icode pattern.
5148 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5149 (rs6000_gimple_fold_builtin): Likewise.
5150 (altivec_init_builtins): Add code to define
5151 __builtin_altivec_lvx_v1ti function.
5153 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5156 * function.c (match_asm_constraints_1): Don't do the optimization
5157 if input isn't a REG, SUBREG, MEM or constant.
5159 2018-03-22 Tom de Vries <tom@codesourcery.com>
5161 PR tree-optimization/84956
5162 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5163 bb_has_abnormal_pred.
5165 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5168 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5169 DECL_INITIAL (decl) to decl at the end.
5170 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5173 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5175 * doc/extend.texi (__builtin_tgmath): Document when complex
5176 integer types are treated as _Complex _Float64.
5178 2018-03-21 Tom de Vries <tom@codesourcery.com>
5180 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5182 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5184 PR tree-optimization/84960
5185 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5186 if it is ENTRY block, move them into single succ of ENTRY in that case.
5188 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5190 PR tree-optimization/84811
5191 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5192 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5193 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5194 (known_subrange_p): Update accordingly. Cast each value involved
5195 in the size comparison, rather than casting the result of the
5198 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5200 PR tree-optimization/84982
5201 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5202 by flipping the least significant bit rather than all bits from
5203 bitpos to bitpos + bitsize - 1.
5205 2018-03-21 Nathan Sidwell <nathan@acm.org>
5207 * doc/extend.texi (Deprecated Features): Remove mention of
5208 long-deleted deprecations.
5210 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5213 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5214 * configure: Regenerate.
5216 2018-03-21 Tom de Vries <tom@codesourcery.com>
5218 PR tree-optimization/83126
5219 * tree-parloops.c (num_phis): New function.
5220 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5222 2018-03-21 Nathan Sidwell <nathan@acm.org>
5224 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5225 mention anon-struct/union members and trailing attributes.
5227 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5229 PR tree-optimization/84969
5230 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5231 builtin memset partitions if they set different rhs values.
5233 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5235 PR rtl-optimization/84989
5236 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5237 VEC_DUPLICATE with scalar result mode.
5239 2018-03-21 Martin Liska <mliska@suse.cz>
5242 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5243 not intended return statement.
5245 2018-03-21 Martin Liska <mliska@suse.cz>
5248 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5249 (chkp_find_bound_slots_1): Limit number of iterations.
5251 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5254 * Minor grammar fixes for x86 options.
5256 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5259 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5260 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5262 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5265 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5266 (altivec_lvx_<mode>_1op): Likewise.
5267 (altivec_stvx_<mode>_2op): Likewise.
5268 (altivec_stvx_<mode>_1op): Likewise.
5269 (altivec_lvx_<VM2:mode>): New define_expand.
5270 (altivec_stvx_<VM2:mode>): Likewise.
5271 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5272 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5273 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5274 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5275 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5276 (rs6000_gen_lvx): Likewise.
5277 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5278 (altivec_expand_stv_builtin): Likewise.
5279 (altivec_expand_builtin): Likewise.
5280 * config/rs6000/vector.md: Likewise.
5282 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5285 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5288 2018-03-20 Richard Biener <rguenther@suse.de>
5291 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5292 sign-conversions as zero, fall back to standard scalar_stmt
5295 2018-03-20 Martin Liska <mliska@suse.cz>
5298 * predict.c (rebuild_frequencies): Handle case when we have
5299 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5301 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5304 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5305 flag_section_anchors.
5306 * varasm.c (use_blocks_for_decl_p): Remove hack for
5307 dw2_force_const_mem.
5310 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5312 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5313 be created, use lowpart_subreg of operands[0] rather than operands[0]
5315 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5316 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5317 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5318 and n constraint instead of aarch64_shift_imm_di and Usd.
5319 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5320 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5322 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5325 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5326 to favor GPR over NEON registers.
5327 (<shift>di3_neon): Likewise.
5329 2018-03-20 Tom de Vries <tom@codesourcery.com>
5332 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5333 (nvptx_process_pars): Emit bar.sync asap and alap.
5335 2018-03-20 Tom de Vries <tom@codesourcery.com>
5338 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5339 seen_label if seen_label is already set.
5341 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5344 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5345 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5346 Use 1U instead of 1. Formatting fixes.
5349 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5350 instead of TREE_TYPE (s1) for the return value.
5352 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5354 PR tree-optimization/84946
5355 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5356 bitsize + bitsize in poly_uint64 rather than poly_int64.
5359 * dwarf2asm.c: Include fold-const.c.
5360 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5361 of decl rather than decl itself.
5363 PR rtl-optimization/84643
5364 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5366 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5369 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5370 calling assemble_variable.
5372 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5375 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5376 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5378 2018-03-19 Jim Wilson <jimw@sifive.com>
5381 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5382 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5383 (riscv_first_stack_step): Likewise.
5384 (riscv_option_override): Use STACK_BOUNDARY instead of
5386 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5388 (BIGGEST_ALIGNMENT): Set to 128.
5389 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5390 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5393 2018-03-19 Richard Biener <rguenther@suse.de>
5395 PR tree-optimization/84933
5396 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5397 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5399 2018-03-19 Richard Biener <rguenther@suse.de>
5401 PR tree-optimization/84859
5402 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5403 (cond_if_else_store_replacement): Perform sinking operation on
5404 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5405 Generalize what a BB with a single eligible store is.
5407 2018-03-19 Richard Biener <rguenther@suse.de>
5409 PR tree-optimization/84929
5410 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5411 chrec_is_positive against non-chrec arg.
5413 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5416 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5418 2018-03-18 Martin Liska <mliska@suse.cz>
5420 PR rtl-optimization/84635
5421 * regrename.c (build_def_use): Use matches_mode only when
5424 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5426 PR tree-optimization/84913
5427 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5428 vectorize chains of COND_EXPRs.
5430 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5432 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5434 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5436 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5438 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5440 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5442 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5443 Kito Cheng <kito.cheng@gmail.com>
5445 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5446 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5447 (nds32_adjust_reg_alloc_order): New function.
5448 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5450 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5452 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5453 nds32_print_operand, nds32_print_operand_address): Use
5454 HOST_WIDE_INT_PRINT_DEC instead.
5456 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5458 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5460 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5463 * config/i386/i386.c (initial_ix86_tune_features,
5464 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5466 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5467 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5468 rather than 1u << ix86_tune. Formatting fix.
5469 (ix86_option_override_internal): Change ix86_arch_mask from
5470 unsigned int to unsigned HOST_WIDE_INT, initialize to
5471 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5472 (ix86_function_specific_restore): Likewise.
5474 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5477 * postreload.c (reload_combine_recognize_pattern): Perform
5478 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5479 truncate_int_for_mode the result for the destination's mode.
5482 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5483 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5485 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5488 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5491 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5492 regno_allocno_class_array and sorted_pseudos.
5493 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5494 insns where regno is used.
5496 2018-03-16 Martin Liska <mliska@suse.cz>
5499 * multiple_target.c (create_dispatcher_calls): Redirect
5500 reference in the symbol table.
5502 2018-03-16 Martin Liska <mliska@suse.cz>
5505 * multiple_target.c (create_dispatcher_calls): Redirect also
5508 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5512 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5513 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5514 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5516 2018-03-16 Julia Koval <julia.koval@intel.com>
5518 * doc/invoke.texi (Skylake Server): Add CLWB.
5519 Cannonlake): Remove CLWB.
5521 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5523 PR tree-optimization/84841
5524 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5526 (FLOAT_ONE_CONST_TYPE): Define.
5527 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5528 (sort_by_operand_rank): Put entries with higher constant_type last
5529 rather than first to match comments.
5531 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5533 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5536 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5539 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5540 check and use address of target always.
5542 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5545 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5546 (indirect_thunk_bnd_needed): Likewise.
5547 (indirect_thunks_used): Likewise.
5548 (indirect_thunks_bnd_used): Likewise.
5549 (indirect_return_needed): New.
5550 (indirect_return_bnd_needed): Likewise.
5551 (output_indirect_thunk_function): Add a bool argument for
5553 (output_indirect_thunk_function): Don't generate alias for
5554 function return thunk.
5555 (ix86_code_end): Call output_indirect_thunk_function to generate
5556 function return thunks.
5557 (ix86_output_function_return): Set indirect_return_bnd_needed
5558 and indirect_return_needed instead of indirect_thunk_bnd_needed
5559 and indirect_thunk_needed.
5561 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5563 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5564 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5565 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5567 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5568 Paul Hua <paul.hua.gm@gmail.com>
5571 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5573 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5575 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5576 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5579 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5582 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5583 instead of GET_MODE_SIZE when comparing Units.
5585 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5588 * varasm.c (hash_section): Return an unchangeble hash value
5589 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5590 Return !aarch64_can_use_per_function_literal_pools_p ().
5592 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5595 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5596 rather than address of cmode as last argument to prepare_cmp_insn.
5598 2018-03-15 Julia Koval <julia.koval@intel.com>
5600 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5601 F_AVX512VNNI, F_AVX512BITALG): New.
5603 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5606 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5607 insn for floating-point loads and stores.
5609 2018-03-14 Carl Love <cel@us.ibm.com>
5611 * config/rs6000/rs6000-c.c: Add macro definitions for
5612 ALTIVEC_BUILTIN_VEC_PERMXOR.
5613 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5614 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5615 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5616 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5618 * config/doc/extend.texi: Add prototypes for vec_permxor.
5620 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5623 * diagnostic-show-locus.c (class layout_point): Convert m_line
5624 from int to linenum_type.
5625 (line_span::comparator): Use linenum "compare" function when
5626 comparing line numbers.
5627 (test_line_span): New function.
5628 (layout_range::contains_point): Convert param "row" from int to
5630 (layout_range::intersects_line_p): Likewise.
5631 (layout::will_show_line_p): Likewise.
5632 (layout::print_source_line): Likewise.
5633 (layout::should_print_annotation_line_p): Likewise.
5634 (layout::print_annotation_line): Likewise.
5635 (layout::print_leading_fixits): Likewise.
5636 (layout::annotation_line_showed_range_p): Likewise.
5637 (struct line_corrections): Likewise for field m_row.
5638 (line_corrections::line_corrections): Likewise for param "row".
5639 (layout::print_trailing_fixits): Likewise.
5640 (layout::get_state_at_point): Likewise.
5641 (layout::get_x_bound_for_row): Likewise.
5642 (layout::print_line): Likewise.
5643 (diagnostic_show_locus): Likewise for locals "last_line" and
5645 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5646 * input.c (selftest::test_linenum_comparisons): New function.
5647 (selftest::input_c_tests): Call it.
5648 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5649 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5650 * selftest.h (ASSERT_GT): New macro.
5651 (ASSERT_GT_AT): New macro.
5652 (ASSERT_LT): New macro.
5653 (ASSERT_LT_AT): New macro.
5655 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5657 PR rtl-optimization/84780
5658 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5660 2018-03-14 Martin Liska <mliska@suse.cz>
5662 * tree.c (record_node_allocation_statistics): Use
5663 get_stats_node_kind.
5664 (get_stats_node_kind): New function extracted from
5665 record_node_allocation_statistics.
5666 (free_node): Use get_stats_node_kind.
5668 2018-03-14 Richard Biener <rguenther@suse.de>
5670 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5671 that the value-set of ANTIC_IN doesn't grow.
5674 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5676 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5677 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5679 2018-03-14 Julia Koval <julia.koval@intel.com>
5681 * config.gcc (icelake-client, icelake-server): New.
5683 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5684 (initial_ix86_arch_features): Ditto.
5685 (PTA_SKYLAKE): Add SGX.
5686 (PTA_ICELAKE): Remove.
5687 (PTA_ICELAKE_CLIENT): New.
5688 (PTA_ICELAKE_SERVER): New.
5689 (ix86_option_override_internal): Split up icelake on icelake client and
5691 (get_builtin_code_for_version): Ditto.
5692 (fold_builtin_cpu): Ditto.
5693 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5694 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5695 * config/i386/i386.h (processor_type): Ditto.
5696 * doc/invoke.texi: Ditto.
5698 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5701 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5702 INTEGER_CST offset, add it together with bitpos / 8 and
5703 sign extend based on POINTER_SIZE.
5707 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5710 * config/i386/constraints.md (Yc): New register constraint.
5711 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5712 Use Yc constraint for alternative 2 of operand 0. Remove
5713 preferred_for_speed attribute.
5715 2018-03-14 Richard Biener <rguenther@suse.de>
5717 PR tree-optimization/84830
5718 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5719 with the old one to avoid oscillations.
5721 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5724 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5726 (assign_by_spills): Return a flag of reload assignment failure.
5727 Do not process the reload assignment failures. Do not spill other
5728 reload pseudos if they has the same reg class. Update n if
5730 (lra_assign): Add a return arg. Set up from the result of
5731 assign_by_spills call.
5732 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5733 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5734 usage_insns if it is not NULL.
5735 (spill_hard_reg_in_range): New function.
5736 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5737 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5738 function prototypes.
5739 (lra_assign): Change prototype.
5740 * lra.c (lra): Add code to deal with fails by splitting hard reg
5743 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5745 * config/riscv/riscv.opt (mrelax): New option.
5746 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5747 "norelax" when riscv_mrelax is disabled.
5748 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5750 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5753 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5754 reassociation for int modes.
5756 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5758 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5759 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5761 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5762 * config/aarch64/aarch64-sve.md
5763 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5764 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5765 (*extend<mode><Vwide>2): Rename to...
5766 (aarch64_sve_extend<mode><Vwide>2): ...this.
5767 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5768 renaming the old pattern to...
5769 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5771 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5772 define_expand, renaming the old pattern to...
5773 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5774 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5775 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5776 account when deciding which SVE instruction the optab should use.
5777 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5779 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5781 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5782 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5783 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5784 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5785 (tlsdesc_small_<mode>): Turn a define_expand and use
5786 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5787 (tlsdesc_small_advsimd_<mode>): ...this.
5788 (tlsdesc_small_sve_<mode>): New pattern.
5790 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5792 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5793 (UNSPEC_UMUL_HIGHPART): New constants.
5794 (MUL_HIGHPART): New int iteraor.
5795 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5796 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5798 (*<su>mul<mode>3_highpart): New define_insn.
5800 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5803 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5806 2018-03-13 Martin Liska <mliska@suse.cz>
5809 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5811 (sem_item_optimizer::~sem_item_optimizer): Release it.
5812 (sem_item_optimizer::merge_classes): Register variable aliases.
5813 (sem_item_optimizer::fixup_pt_set): New function.
5814 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5815 * ipa-icf.h: Declare new variables and functions.
5817 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5820 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5821 integer_pow2p@2 and test integer_pow2p in condition.
5822 (A < 0 ? C : 0): Similarly for @1.
5825 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5826 characters starting at p contain '\0' character, don't look beyond
5830 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5831 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5834 * reg-stack.c (change_stack): Change update_end var from int to
5835 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5836 also call set_block_for_insn on the newly added insns and rescan.
5839 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5840 on the last operand.
5843 * tree.c (stabilize_reference_1): Return save_expr (e) for
5844 STATEMENT_LIST even if it doesn't have side-effects.
5846 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5848 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5850 2018-03-12 Renlin Li <renlin.li@arm.com>
5852 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5853 aarch64_output_scalar_simd_mov_immediate.
5855 2018-03-12 Martin Sebor <msebor@redhat.com>
5857 PR tree-optimization/83456
5858 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5859 for perfectly overlapping calls to memcpy.
5860 (gimple_fold_builtin_memory_chk): Same.
5861 (gimple_fold_builtin_strcpy): Handle no-warning.
5862 (gimple_fold_builtin_stxcpy_chk): Same.
5863 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5865 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5867 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5868 parameter. Use it for SFmode.
5869 (rs6000_function_arg_advance_1): Adjust.
5870 (rs6000_function_arg): Adjust.
5871 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5873 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5875 PR rtl-optimization/84169
5876 PR rtl-optimization/84780
5877 * combine.c (can_combine_p): Check for a 2-insn combination whether
5878 the destination register is used between the two insns, too.
5880 2018-03-12 Richard Biener <rguenther@suse.de>
5882 PR tree-optimization/84803
5883 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5884 for refs DR analysis didn't process.
5886 2018-03-12 Richard Biener <rguenther@suse.de>
5888 PR tree-optimization/84777
5889 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5890 force-vectorize loops ignore whether we are optimizing for size.
5892 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5894 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5895 (TARGET_MD_ASM_ADJUST): Define.
5897 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5898 Kito Cheng <kito.cheng@gmail.com>
5899 Chung-Ju Wu <jasonwucj@gmail.com>
5901 * config/nds32/nds32.c (nds32_compute_stack_frame,
5902 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5903 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5904 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5905 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5906 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5907 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5908 * config/nds32/nds32.md (prologue, epilogue): Use macro
5909 NDS32_V3PUSH_AVAILABLE_P to do checking.
5911 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5914 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5915 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5916 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5917 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5918 addition of most attributes on !orig_type_die or the attribute not
5919 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5921 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5922 Chung-Ju Wu <jasonwucj@gmail.com>
5924 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5926 * config/nds32/nds32.opt (mvh): New option.
5928 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5929 Chung-Ju Wu <jasonwucj@gmail.com>
5931 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5933 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5934 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5937 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5938 Chung-Ju Wu <jasonwucj@gmail.com>
5940 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5942 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5943 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5945 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5946 Kito Cheng <kito.cheng@gmail.com>
5947 Chung-Ju Wu <jasonwucj@gmail.com>
5949 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5950 UNSPEC_FFMISM and UNSPEC_FLMISM.
5951 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5952 for ffb, ffmism and flmism.
5953 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5954 (unspec_ffmism): Ditto.
5955 (unspec_flmism): Ditto.
5956 (nds32_expand_builtin_impl): Check if string extension is available.
5957 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5958 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5960 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5963 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5966 * lra-assigns.c (assign_by_spills): Return a flag of reload
5967 assignment failure. Do not process the reload assignment
5968 failures. Do not spill other reload pseudos if they has the same
5970 (lra_assign): Add a return arg. Set up from the result of
5971 assign_by_spills call.
5972 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5973 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5974 usage_insns if it is not NULL.
5975 (spill_hard_reg_in_range): New function.
5976 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5977 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5978 function prototypes.
5979 (lra_assign): Change prototype.
5980 * lra.c (lra): Add code to deal with fails by splitting hard reg
5983 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5986 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5988 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5991 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5992 (dw_val_node): Add val_symbolic_view.
5993 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5994 (symview_upper_bound): New.
5995 (new_line_info_table): Initialize symviews_since_reset.
5996 (dwarf2out_source_line): Count symviews_since_reset and set
5997 symview_upper_bound.
5998 (dw_val_equal_p): Handle symview.
5999 (add_AT_symview): New.
6000 (print_dw_val): Handle symview.
6001 (attr_checksum, attr_checksum_ordered): Likewise.
6002 (same_dw_val_p, size_of_die): Likewise.
6003 (value_format, output_die): Likewise.
6004 (add_high_low_attributes): Use add_AT_symview for entry_view.
6005 (dwarf2out_finish): Reset symview_upper_bound, clear
6008 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6011 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6012 Add strict argument and use it.
6013 (rs6000_split_multireg_move): Update for new strict argument.
6014 (mem_operand_gpr): Disallow all non-offsettable addresses.
6015 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6017 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6020 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6021 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6022 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6025 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6026 decl, use remap_type if we want to use the type.
6028 2018-03-09 Martin Sebor <msebor@redhat.com>
6030 PR tree-optimization/84526
6031 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6033 (builtin_access::generic_overlap): Be prepared to handle non-array
6036 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6038 PR rtl-optimization/84682
6039 * lra-constraints.c (process_address_1): Check is_address flag
6040 for address constraints.
6041 (process_alt_operands): Likewise.
6042 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6043 preprocess_constraints.
6044 * recog.h (preprocess_constraints): Add oploc parameter.
6046 * recog.c (preprocess_constraints): Test address_operand for
6047 CT_ADDRESS constraints.
6049 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6052 * lra-assigns.c (assign_by_spills): Return a flag of reload
6053 assignment failure. Do not process the reload assignment
6054 failures. Do not spill other reload pseudos if they has the same
6056 (lra_assign): Add a return arg. Set up from the result of
6057 assign_by_spills call.
6058 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6059 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6060 usage_insns if it is not NULL.
6061 (spill_hard_reg_in_range): New function.
6062 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6063 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6064 function prototypes.
6065 (lra_assign): Change prototype.
6066 * lra.c (lra): Add code to deal with fails by splitting hard reg
6069 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6072 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6073 Accept complain bool parameter. Only emit errors if it is true.
6074 (arm_parse_cpu_option_name): Likewise.
6075 (arm_target_thumb_only): Adjust callers of the above.
6076 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6077 prototype to take a default true bool parameter.
6078 (arm_parse_arch_option_name): Likewise.
6080 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6081 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6085 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6086 * configure: Regenerate.
6087 * configure.ac ("linker --version-script option"): New.
6088 ("linker soname option"): New.
6090 2018-03-09 Richard Biener <rguenther@suse.de>
6092 PR tree-optimization/84775
6093 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6094 immediate uses of predicate stmts and mark them modified.
6097 PR tree-optimization/84178
6098 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6100 (version_loop_for_if_conversion): Delay update_ssa call.
6101 (tree_if_conversion): Delay update_ssa until after predicate
6104 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6107 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6108 when the function accesses prior frames.
6110 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6113 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6114 gen_llsym, otherwise call maybe_gen_llsym.
6117 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6118 has ',' character inside of it.
6120 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6123 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6124 as clobbering CC_REGNUM.
6126 2018-03-08 Richard Biener <rguenther@suse.de>
6129 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6130 (follow_copies_to_constant): Do not follow SSA names registered
6133 2018-03-08 Richard Biener <rguenther@suse.de>
6135 PR tree-optimization/84178
6136 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6138 (version_loop_for_if_conversion): Delay update_ssa call.
6139 (tree_if_conversion): Delay update_ssa until after predicate
6142 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6144 PR tree-optimization/84178
6145 * tree-if-conv.c (release_bb_predicate): Remove the
6146 the assertion that the stmts have NULL use_ops.
6147 Discard the statements, asserting that they haven't
6148 yet been added to a BB.
6150 2018-03-08 Richard Biener <rguenther@suse.de>
6152 PR tree-optimization/84746
6153 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6154 (phi_translate): Pass in destination ANTIC_OUT set.
6155 (phi_translate_1): Likewise. For a simplified result lookup
6156 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6157 (phi_translate_set): Adjust.
6158 (do_pre_regular_insertion): Likewise.
6159 (do_pre_partial_partial_insertion): Likewise.
6161 2018-03-08 Martin Liska <mliska@suse.cz>
6163 PR gcov-profile/84735
6164 * doc/gcov.texi: Document usage of profile files.
6165 * gcov-io.h: Document changes in the format.
6167 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6171 * dwarf2out.c (struct dw_line_info_table): Update comments for
6173 (FORCE_RESET_NEXT_VIEW): New.
6174 (FORCE_RESETTING_VIEW_P): New.
6175 (RESETTING_VIEW_P): Check for -1 too.
6176 (ZERO_VIEW_P): Likewise.
6177 (new_line_info_table): Force-reset next view.
6178 (dwarf2out_begin_function): Likewise.
6179 (dwarf2out_source_line): Simplify zero_view_p initialization.
6180 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6181 view directly. Omit view when omitting .loc at line 0.
6183 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6185 PR tree-optimization/84740
6186 * tree-switch-conversion.c (process_switch): Call build_constructors
6187 only if info.phi_count is non-zero.
6189 PR tree-optimization/84739
6190 * tree-tailcall.c (find_tail_calls): Check call arguments against
6191 DECL_ARGUMENTS (current_function_decl) rather than
6192 DECL_ARGUMENTS (func) when checking for tail recursion.
6194 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6196 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6197 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6198 Volker Reichelt's entry and add entries for people that perform
6199 GCC fuzzy testing and report numerous bugs.
6201 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6204 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6205 readonly data in sdata, if that is disabled.
6206 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6207 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6208 -mreadonly-in-sdata option.
6210 2018-03-07 Martin Sebor <msebor@redhat.com>
6212 PR tree-optimization/84468
6213 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6214 basic block when looking for nul assignment.
6216 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6219 * except.h (output_function_exception_table): Adjust prototype.
6220 * except.c (output_function_exception_table): Remove FNNAME parameter
6221 and add SECTION parameter. Ouput one part of the table at a time.
6222 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6223 the first part of the exception table and emit unwind directives.
6224 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6225 (i386_pe_seh_cold_init): Likewise.
6226 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6227 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6228 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6229 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6230 * config/i386/winnt.c: Include except.h.
6231 (struct seh_frame_state): Add reg_offset, after_prologue and
6232 in_cold_section fields.
6233 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6234 (i386_pe_seh_cold_init): New function.
6235 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6236 to seh->in_cold_section.
6237 (seh_emit_push): Record the offset of the push.
6238 (seh_emit_save): Record the offet of the save.
6239 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6240 Test seh->after_prologue to disregard the epilogue.
6241 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6242 (i386_pe_end_cold_function): New function.
6244 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6247 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6248 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6251 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6253 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6254 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6257 * multiple_target.c: Include tree-inline.h and intl.h.
6258 (expand_target_clones): Diagnose and fail if node->definition and
6259 !tree_versionable_function_p (node->decl).
6261 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6263 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6265 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6266 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6267 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6269 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6272 * combine.c (try_combine): Use reg_or_subregno instead of handling
6273 just paradoxical SUBREGs and REGs.
6275 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6277 * config/arc/arc.c (arc_finalize_pic): Remove function.
6278 (arc_must_save_register): We use single base PIC register, remove
6279 checks to save/restore the PIC register.
6280 (arc_expand_prologue): Likewise.
6281 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6283 (arc_verify_short): Likewise.
6284 (arc_attr_type): Likewise.
6285 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6286 (walk_stores): Likewise.
6287 (arc_address_cost): Make it static.
6288 (arc_verify_short): Likewise.
6289 (branch_dest): Likewise.
6290 (arc_attr_type): Likewise.
6291 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6292 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6293 (arc_final_prescan_insn): Remove inserting the nops due to
6294 hardware hazards. It is done in reorg step.
6295 (insn_length_variant_t): Remove.
6296 (insn_length_parameters_t): Likewise.
6297 (arc_insn_length_parameters): Likewise.
6298 (arc_get_insn_variants): Likewise.
6299 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6301 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6304 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6307 PR tree-optimization/84687
6308 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6310 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6312 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6314 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6315 Rename to ppc_speculation_barrier.
6316 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6317 __builtin_ppc_speculation_barrier.
6319 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6322 * combine.c (combine_simplify_rtx): Don't try to simplify if
6323 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6326 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6328 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6329 to 32 bytes when compiling for POWER9.
6331 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6334 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6335 regparm >= 3 with no arg reg available also for calls with
6336 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6339 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6341 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6343 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6346 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6348 2018-03-05 Richard Biener <rguenther@suse.de>
6350 PR tree-optimization/84486
6351 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6352 When inserting a __builtin_assume_aligned call set the LHS
6353 SSA name alignment info accordingly.
6355 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6357 PR tree-optimization/84114
6358 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6359 Avoid reassociation of FLOAT_MODE addition.
6361 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6363 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6364 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6365 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6366 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6367 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6368 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6369 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6371 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6372 __builtin_ia32_wbinvd): New builtins.
6373 (SPECIAL_ARGS2): New.
6374 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6375 (SPECIAL_ARGS2): New.
6376 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6377 (ix86_valid_target_attribute_inner_p): Ditto.
6378 (ix86_init_mmx_sse_builtins): Add special_args2.
6379 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6380 TARGET_WBNOINVD_P): New.
6381 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6382 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6383 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6384 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6385 * config/i386/pconfigintrin.h: New file.
6386 * config/i386/wbnoinvdintrin.h: Ditto.
6387 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6388 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6390 2018-03-05 Richard Biener <rguenther@suse.de>
6392 PR tree-optimization/84670
6393 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6395 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6396 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6397 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6398 doesn't grow if all successors (recursively) were visited at least
6401 2018-03-05 Richard Biener <rguenther@suse.de>
6403 PR tree-optimization/84650
6404 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6405 if executed in the loop pipeline.
6407 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6409 * doc/configfiles.texi (Configuration Files): Move info about
6410 conditionalizing $target-protos.h to...
6411 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6412 differs from $target-protos.h.
6414 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6415 Chung-Ju Wu <jasonwucj@gmail.com>
6417 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6418 * config/nds32/nds32-multiple.md (setmemsi): Define.
6419 * config/nds32/nds32-memory-manipulation.c
6420 (nds32_gen_dup_4_byte_to_word_value): New.
6421 (emit_setmem_word_loop): New.
6422 (emit_setmem_byte_loop): New.
6423 (nds32_expand_setmem_loop): New.
6424 (nds32_expand_setmem_loop_v3m): New.
6425 (nds32_expand_setmem_unroll): New.
6426 (nds32_expand_setmem): New.
6428 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6429 Chung-Ju Wu <jasonwucj@gmail.com>
6431 * config/nds32/nds32-memory-manipulation.c
6432 (nds32_emit_load_store): New.
6433 (nds32_emit_post_inc_load_store): New.
6434 (nds32_emit_mem_move): New.
6435 (nds32_emit_mem_move_block): New.
6436 (nds32_expand_movmemsi_loop_unknown_size): New.
6437 (nds32_expand_movmemsi_loop_known_size): New.
6438 (nds32_expand_movmemsi_loop): New.
6439 (nds32_expand_movmemsi_unroll): New.
6440 (nds32_expand_movmemqi): Rename ...
6441 (nds32_expand_movmemsi): ... to this.
6442 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6443 (movmemsi): ... to this.
6444 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6445 (nds32_expand_movmemsi): ... to this.
6447 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6448 Monk Chiang <sh.chiang04@gmail.com>
6449 Chung-Ju Wu <jasonwucj@gmail.com>
6451 * config/nds32/nds32-protos.h
6452 (nds32_expand_load_multiple): New arguments.
6453 (nds32_expand_store_multiple): Ditto.
6454 (nds32_valid_multiple_load_store): Rename ...
6455 (nds32_valid_multiple_load_store_p): ... to this.
6456 * config/nds32/nds32-memory-manipulation.c
6457 (nds32_expand_load_multiple): Refine implementation.
6458 (nds32_expand_store_multiple): Ditto.
6459 * config/nds32/nds32-multiple.md
6460 (load_multiple): Update nds32_expand_load_multiple interface.
6461 (store_multiple): Update nds32_expand_store_multiple interface.
6462 * config/nds32/nds32-predicates.c
6463 (nds32_valid_multiple_load_store): Rename ...
6464 (nds32_valid_multiple_load_store_p): ... to this and refine
6466 * config/nds32/predicates.md
6467 (nds32_load_multiple_and_update_address_operation): New predicate.
6468 (nds32_store_multiple_and_update_address_operation): New predicate.
6470 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6471 Chung-Ju Wu <jasonwucj@gmail.com>
6473 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6474 (combo): New attribute.
6475 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6477 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6479 * config/nds32/nds32.opt: Change -mcmodel= default value.
6481 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6482 Monk Chiang <sh.chiang04@gmail.com>
6483 Chung-Ju Wu <jasonwucj@gmail.com>
6485 * config/nds32/constants.md (unspec_element): New enum.
6486 * config/nds32/constraints.md (Umw): New constraint.
6487 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6488 * config/nds32/nds32-intrinsic.md: Likewise.
6489 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6490 (nds32_valid_smw_lwm_base_p): New.
6491 (nds32_output_smw_single_word): New.
6492 (nds32_output_lmw_single_word): New.
6493 (nds32_expand_unaligned_load): New.
6494 (nds32_expand_unaligned_store): New.
6495 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6496 (nds32_output_smw_single_word): Declare.
6497 (nds32_output_lmw_single_word): Declare.
6498 (nds32_expand_unaligned_load): Declare.
6499 (nds32_expand_unaligned_store): Declare.
6500 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6501 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6502 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6503 NDS32_BUILTIN_UASTORE_DW.
6504 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6507 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6508 Kito Cheng <kito.cheng@gmail.com>
6509 Chung-Ju Wu <jasonwucj@gmail.com>
6511 * config/nds32/nds32-intrinsic.c
6512 (nds32_expand_builtin_null_ftype_reg): Delete.
6513 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6514 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6515 (nds32_read_argument): New.
6516 (nds32_legitimize_target): Ditto.
6517 (nds32_legitimize_argument): Ditto.
6518 (nds32_check_constant_argument): Ditto.
6519 (nds32_expand_unop_builtin): Ditto.
6520 (nds32_expand_unopimm_builtin): Ditto.
6521 (nds32_expand_binop_builtin): Ditto.
6522 (nds32_builtin_decl_impl): Ditto.
6523 (builtin_description): Ditto.
6524 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6525 (nds32_init_builtins_impl): Ditto.
6526 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6527 (nds32_builtin_decl): New.
6528 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6529 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6531 2018-03-02 Jeff Law <law@redhat.com>
6533 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6534 (redundant_insn, fill_simple_delay_slots): Likewise.
6535 (fill_slots_from_thread): Likewise.
6536 * resource.c (mark_referenced_resources): Likewise.
6537 (mark_set_resources, find_dead_or_set_registers): Likewise.
6539 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6541 * substring-locations.h (format_warning_va): Formatting fix for
6543 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6545 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6546 * substring-locations.c: Include intl.h.
6547 (format_warning_va): Turned into small wrapper around
6548 format_warning_n_va, renamed to ...
6549 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6550 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6552 (format_warning_at_substring_n): New function.
6553 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6554 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6555 format_warning_at_substring with just a shorter name instead of
6556 const function pointer.
6557 (fmtwarn_n): New function.
6558 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6559 appropriate, get rid of all the fmtstr temporaries, move conditionals
6560 with G_() wrapped string literals directly into fmtwarn arguments,
6561 cast dir.len to (int), formatting fixes.
6563 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6565 * doc/invoke.texi: Remove "Cilk Plus" references.
6567 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6568 Richard Biener <rguenther@suse.de>
6571 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6572 for error or warning attributes if CALL_FROM_THUNK_P is set.
6575 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6578 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6579 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6582 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6583 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6585 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6586 instead of -1U in last predictors element's probability member.
6588 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6591 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6592 arguments if they are comparable.
6594 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6596 PR tree-optimization/84634
6597 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6598 masks and masked_loop_p with a single loop_masks, making sure it's
6599 null for bb vectorization.
6601 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6603 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6604 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6605 loop->force_vectorize to check whether there is no alias.
6607 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6610 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6612 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6614 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6615 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6616 instead of a loop around prev_real_insn.
6617 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6621 * config/i386/i386.c (ix86_print_operand): Use conditional
6622 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6625 2018-03-02 Richard Biener <rguenther@suse.de>
6627 PR tree-optimization/84427
6628 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6629 (bitmap_set_subtract_values): Rewrite to handle multiple
6632 (prune_clobbered_mems): Likewise.
6633 (phi_translate): Take edge instead of pred/phiblock.
6634 (phi_translate_1): Likewise.
6635 (phi_translate_set): Likewise. Insert all translated
6636 exprs for a value into the set, keeping possibly multiple
6637 expressions per value.
6638 (compute_antic_aux): Adjust for phi_translate changes.
6639 When intersecting union the expressions and prune those
6640 not in the final value set, keeping possibly multiple
6641 expressions per value. Do not use value-insertion
6642 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6643 all expressions. Add verification that the value-sets
6644 only shrink during iteration.
6645 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6646 (do_pre_regular_insertion): Likewise.
6647 (do_pre_partial_partial_insertion): Likewise.
6649 2018-03-02 Richard Biener <rguenther@suse.de>
6652 * config/darwin.c (saved_debug_info_level): New static global.
6653 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6654 (darwin_asm_lto_end): Restore debug info generation settings.
6656 2018-03-01 Martin Liska <mliska@suse.cz>
6659 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6662 2018-03-01 Richard Biener <rguenther@suse.de>
6665 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6666 type annotation with LTO when debug was disabled at compile-time.
6668 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6670 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6672 (mips_final_postscan_insn): Likewise.
6674 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6676 PR rtl-optimization/84528
6677 * alias.c (init_alias_target): Add commentary.
6678 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6679 a unique base value if the frame pointer is not eliminated
6680 to the stack pointer.
6682 2018-03-01 Tom de Vries <tom@codesourcery.com>
6684 PR rtl-optimization/83327
6685 * lra-int.h (hard_regs_spilled_into): Declare.
6686 * lra.c (hard_regs_spilled_into): Define.
6687 (init_reg_info): Init hard_regs_spilled_into.
6688 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6689 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6690 (process_bb_lives): Handle hard_regs_spilled_into.
6691 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6692 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6694 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6696 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6697 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6698 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6699 * config/rs6000/aix72.h: New file.
6701 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6703 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6704 instead of warning_at with conditional singular and plural messages
6708 * stor-layout.c (update_alignment_for_field): For
6709 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6710 && !DECL_PACKED (field), do the alignment update, just use
6711 only desired_align instead of MAX (type_align, desired_align)
6713 (place_field): Don't do known_align < desired_align handling
6714 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6715 is non-NULL, instead do it after rli->prev_field handling and
6716 only if not within a bitfield word. For DECL_PACKED (field)
6717 use type_align of BITS_PER_UNIT.
6719 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6721 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6722 superfluous parentheses and trailing spaces.
6724 2018-02-28 Richard Biener <rguenther@suse.de>
6726 PR tree-optimization/84584
6727 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6728 SCoPs with fake exit edge.
6730 2018-02-28 Martin Liska <mliska@suse.cz>
6733 * timevar.c (timer::print): Fix format to properly print 100%
6736 2018-02-28 Richard Biener <rguenther@suse.de>
6739 * genmatch.c (capture_info::walk_match): Do not mark
6740 captured expressions without operands as expr_p given
6741 they act more like predicates and should be subject to
6742 "lost tail" side-effect preserving.
6744 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6746 PR rtl-optimization/81611
6747 * auto-inc-dec.c (attempt_change): Move dead note from
6748 mem_insn if it's the next use of regno
6749 (find_address): Take address use of reg holding
6750 non-incremented value. Add parm to limit search to the named
6752 (merge_in_block): Attempt to use a mem insn that is the next
6753 use of the original regno.
6755 2018-02-27 Martin Sebor <msebor@redhat.com>
6758 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6759 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6761 2018-02-27 Martin Sebor <msebor@redhat.com>
6763 PR translation/84207
6764 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6765 n argument to unsigned HOST_WIDE_INT.
6766 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6767 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6768 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6769 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6771 2018-02-27 Richard Biener <rguenther@suse.de>
6773 PR tree-optimization/84512
6774 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6775 Do not use the estimate returned from record_stmt_cost for
6776 the scalar iteration cost but sum properly using add_stmt_cost.
6778 2018-02-27 Richard Biener <rguenther@suse.de>
6780 PR tree-optimization/84466
6781 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6782 Adjust last change to less strictly validate use operands.
6784 2018-02-27 Martin Liska <mliska@suse.cz>
6786 PR gcov-profile/84548
6787 * gcov.c (process_file): Allow partial overlap and consider it
6788 also as group functions.
6789 (output_lines): Properly calculate range of lines for a group.
6791 2018-02-27 Martin Liska <mliska@suse.cz>
6793 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6794 'ggc' suffixes. Change first column width.
6795 (timer::print): Fix formatting of the column.
6797 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6799 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6800 preserve inline entry blocks for the sake of debug inline
6801 entry point markers alone.
6802 (remove_unused_locals): Suggest in comments a better place to
6803 force the preservation of inline entry blocks that are
6804 otherwise unused, but do not preserve them.
6806 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6808 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6810 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6813 * config/i386/constraints.md (Bs): Replace
6814 ix86_indirect_branch_register with
6815 TARGET_INDIRECT_BRANCH_REGISTER.
6817 * config/i386/i386.md (indirect_jump): Likewise.
6818 (tablejump): Likewise.
6819 (*sibcall_memory): Likewise.
6820 (*sibcall_value_memory): Likewise.
6821 Peepholes of indirect call and jump via memory: Likewise.
6822 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6823 (*sibcall_value_GOT_32): Likewise.
6824 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6825 (GOT_memory_operand): Likewise.
6826 (call_insn_operand): Likewise.
6827 (sibcall_insn_operand): Likewise.
6828 (GOT32_symbol_operand): Likewise.
6829 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6831 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6833 PR rtl-optimization/83496
6834 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6835 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6836 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6837 redundant insn, if any.
6838 (relax_delay_slots): Likewise.
6839 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6841 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6843 PR tree-optimization/83965
6844 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6845 that grouped statements are part of a reduction chain. Return
6846 true if the statement is not marked as a reduction itself but
6848 (vect_recog_dot_prod_pattern): Don't check whether the statement
6849 is part of a group here.
6850 (vect_recog_sad_pattern): Likewise.
6851 (vect_recog_widen_sum_pattern): Likewise.
6853 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6856 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6858 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6861 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6863 (ix86_output_indirect_function_return): New prototype.
6864 (ix86_split_simple_return_pop_internal): Likewise.
6865 * config/i386/i386.c (indirect_return_via_cx): New.
6866 (indirect_return_via_cx_bnd): Likewise.
6867 (indirect_thunk_name): Handle return va CX_REG.
6868 (output_indirect_thunk_function): Create alias for
6869 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6870 (ix86_output_indirect_jmp): Remove the bool argument.
6871 (ix86_output_indirect_function_return): New function.
6872 (ix86_split_simple_return_pop_internal): Likewise.
6873 * config/i386/i386.md (*indirect_jump): Don't pass false
6874 to ix86_output_indirect_jmp.
6875 (*tablejump_1): Likewise.
6876 (simple_return_pop_internal): Change it to define_insn_and_split.
6877 Call ix86_split_simple_return_pop_internal to split it for
6879 (simple_return_indirect_internal): Call
6880 ix86_output_indirect_function_return instead of
6881 ix86_output_indirect_jmp.
6883 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6886 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6887 memset and value initialization afterwards.
6889 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6891 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6893 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6896 * common/config/aarch64/aarch64-common.c
6897 (aarch_option_optimization_table[]): Switch
6898 off fomit-frame-pointer
6900 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6901 Chung-Ju Wu <jasonwucj@gmail.com>
6903 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6905 (store_multiple): Ditto.
6907 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6909 * config.gcc: Add --with-cpu support for nds32 target.
6910 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6911 * config/nds32/nds32.opt: Add -mcpu= option.
6913 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6915 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6916 isel=yes): Warn for these deprecated options.
6918 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6920 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6921 ISA_2_5_MASKS_EMBEDDED.
6923 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6925 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6926 p->max as pointers rather than using iterative_hash_expr.
6928 2018-02-23 Carl Love <cel@us.ibm.com>
6930 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6931 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6933 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6934 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6935 P8V_BUILTIN_VEC_VUNSIGNED2.
6937 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6940 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6941 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6943 * lra-constraints.c (curr_insn_transform): Set up
6944 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6946 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6947 * lra-eliminations.c (spill_pseudos): Ditto.
6948 (process_insn_for_elimination): Ditto.
6949 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6950 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6951 LRA_NON_CLOBBERED_ALT.
6953 2018-02-22 Martin Sebor <msebor@redhat.com>
6955 PR tree-optimization/84480
6956 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6957 to maybe_diag_stxncpy_trunc. Call it.
6958 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6959 from gimple_fold_builtin_strcpy. Print inlining stack.
6960 (handle_builtin_stxncpy): Print inlining stack.
6961 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6963 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6966 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6967 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6968 and -fcheck-pointer-bounds are used together.
6969 (indirect_thunk_prefix): New enum.
6970 (indirect_thunk_need_prefix): New function.
6971 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6972 "_nt" instead of "_bnd" for NOTRACK prefix.
6973 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6974 (output_indirect_thunk_function): Likewise.
6976 (ix86_code_end): Update output_indirect_thunk_function calls.
6977 (ix86_output_indirect_branch_via_reg): Replace
6978 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6979 (ix86_output_indirect_branch_via_push): Likewise.
6980 (ix86_output_function_return): Likewise.
6981 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6982 incompatible with -fcf-protection=branch and
6983 -fcheck-pointer-bounds.
6985 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6988 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6989 Change gcc_assert call to output_operand_lossage.
6991 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6993 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6995 2018-02-22 DJ Delorie <dj@redhat.com>
6996 Sebastian Perta <sebastian.perta@renesas.com>
6997 Oleg Endo <olegendo@gcc.gnu.org>
6999 * config/rx/rx.c (rx_rtx_costs): New function.
7000 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7002 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7004 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7006 2018-02-22 Martin Liska <mliska@suse.cz>
7009 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7010 Add "native" as a possible value.
7012 2018-02-22 Martin Liska <mliska@suse.cz>
7015 * config/i386/i386.c (ix86_option_override_internal):
7016 Add "native" as a possible value for -march and -mtune.
7018 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7021 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7022 to all type variants.
7024 PR tree-optimization/84503
7025 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7026 width as info->bitpos + info->bitsize - start.
7027 (merged_store_group::merge_overlapping): Simplify width computation.
7028 (check_no_overlap): New function.
7029 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7030 start + width and last_order of the group, fail if check_no_overlap
7032 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7033 to group if check_no_overlap fails.
7035 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7037 * config/rs6000/altivec.md: Delete contraint arguments to
7038 define_expand, define_split, and define_peephole2, and in
7039 define_insn_and_split if always unused.
7040 * config/rs6000/darwin.md: Ditto.
7041 * config/rs6000/dfp.md: Ditto.
7042 * config/rs6000/rs6000.md: Ditto.
7043 * config/rs6000/sync.md: Ditto.
7044 * config/rs6000/vector.md: Ditto.
7045 * config/rs6000/vsx.md: Ditto.
7047 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7049 * config/rs6000/altivec.md: Write output control strings as braced
7050 blocks instead of double-quoted strings.
7051 * config/rs6000/darwin.md: Ditto.
7052 * config/rs6000/rs6000.md: Ditto.
7053 * config/rs6000/vector.md: Ditto.
7054 * config/rs6000/vsx.md: Ditto.
7056 2018-02-21 Jason Merrill <jason@redhat.com>
7058 PR c++/84314 - ICE with templates and fastcall attribute.
7059 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7061 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7063 * ipa-cp.c (determine_versionability): Fix comment typos.
7065 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7068 * ipa-cp.c (determine_versionability): Do not version functions caling
7071 2018-02-21 Martin Liska <mliska@suse.cz>
7074 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7075 Add "native" as a possible value.
7076 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7077 the macro when native cpu detection is available.
7079 2018-02-21 Martin Liska <mliska@suse.cz>
7082 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7083 Add "native" as a possible value.
7084 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7085 when native cpu detection is available.
7087 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7088 Martin Sebor <msebor@redhat.com>
7090 PR tree-optimization/84478
7091 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7093 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7094 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7095 support which is conservatively correct, for 2 only stay conservative
7096 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7097 argument to the 2 argument get_range_strlen, adjust 6 arg
7098 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7100 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7101 (gimple_fold_builtin_strlen): Pass true as last argument to
7104 2018-02-20 Martin Sebor <msebor@redhat.com>
7107 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7108 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7109 (builtin_memref::builtin_memref): Factor out parts into
7110 set_base_and_offset and call it.
7112 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7115 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7116 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7117 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7118 search at the associated MODE_INT.
7120 2018-02-20 Jeff Law <law@redhat.com>
7123 PR tree-optimization/81592
7125 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7126 for range data rather than using global data.
7127 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7128 range data rather than using global data.
7129 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7130 pass it to children as needed.
7131 (struct directive::fmtresult): Similarly.
7132 (struct directive::set_width): Similarly.
7133 (struct directive::set_precision): Similarly.
7134 (format_integer, format_directive, parse_directive): Similarly.
7135 (format_none): Accept unnamed vr_values parameter.
7136 (format_percent, format_floating, format_character): Similarly.
7137 (format_string, format_plain): Similarly.
7138 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7139 the EVRP range analyzer for range data rather than using global data.
7140 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7141 gimple-ssa-evrp-analyze.h
7142 (class sprintf_dom_walker): Add after_dom_children member function.
7143 Add evrp_range_analyzer member.
7144 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7145 range analyzer as needed.
7146 (sprintf_dom_walker::after_dom_children): New member function.
7147 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7149 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7150 (evrp_range_analyzer::pop_to_marker): Likewise.
7152 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7154 PR tree-optimization/84419
7155 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7156 with the required type if its current type is compatible but
7159 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7162 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7163 after vectorization.
7165 2018-02-20 Martin Liska <mliska@suse.cz>
7168 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7169 possible values if we don't have a hint.
7171 2018-02-20 Martin Liska <mliska@suse.cz>
7175 * final.c (shorten_branches): Build align_tab array with one
7177 * opts.c (finish_options): Add alignment option limit check.
7178 (MAX_CODE_ALIGN): Likewise.
7179 (MAX_CODE_ALIGN_VALUE): Likewise.
7180 * doc/invoke.texi: Document maximum allowed option value for
7181 all -falign-* options.
7183 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7186 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7187 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7188 * var-tracking.c (emit_note_insn_var_location): Remove all references
7189 to NOTE_INSN_CALL_ARG_LOCATION.
7190 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7191 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7192 Use copy_rtx_if_shared.
7193 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7194 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7195 (dwarf2out_var_location): Remove handling of
7196 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7198 * final.c (final_scan_insn): Remove all references to
7199 NOTE_INSN_CALL_ARG_LOCATION.
7200 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7201 before dumping final insns.
7202 * except.c (emit_note_eh_region_end): Remove all references to
7203 NOTE_INSN_CALL_ARG_LOCATION.
7204 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7205 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7206 * config/arc/arc.c (hwloop_optimize): Likewise.
7207 * config/arm/arm.c (create_fix_barrier): Likewise.
7208 * config/s390/s390.c (s390_chunkify_start): Likewise.
7209 * config/sh/sh.c (find_barrier): Likewise.
7210 * config/i386/i386.c (rest_of_insert_endbranch,
7211 ix86_seh_fixup_eh_fallthru): Likewise.
7212 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7213 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7214 * config/frv/frv.c (frv_function_prologue): Likewise.
7215 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7217 (note_outside_basic_block_p): Remove all references to
7218 NOTE_INSN_CALL_ARG_LOCATION.
7219 * gengtype.c (adjust_field_rtx_def): Likewise.
7220 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7222 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7223 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7226 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7229 PR tree-optimization/84452
7230 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7231 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7234 2018-02-19 Martin Liska <mliska@suse.cz>
7237 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7239 2018-02-19 Martin Liska <mliska@suse.cz>
7240 Richard Sandiford <richard.sandiford@linaro.org>
7242 PR tree-optimization/82491
7243 * gimple-fold.c (get_base_constructor): Make earlier bail out
7246 2018-02-19 Carl Love <cel@us.ibm.com>
7248 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7249 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7251 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7252 P8V_BUILTIN_VEC_NEG.
7254 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7256 * config/rl78/rl78.md (movdf): New define expand.
7258 2018-02-19 Martin Liska <mliska@suse.cz>
7261 * doc/invoke.texi: Fix typo.
7262 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7264 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7266 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7267 handle rs6000_single_float and rs6000_double_float specially for
7270 2018-02-16 Jeff Law <law@redhat.com>
7272 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7275 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7278 * ipa-inline.c (inline_small_functions): Fix a typo.
7280 2018-02-16 Nathan Sidwell <nathan@acm.org>
7282 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7284 2018-02-16 Carl Love <cel@us.ibm.com>
7286 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7287 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7288 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7289 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7290 expansion to P8V_BUILTIN_VEC_FLOAT2.
7292 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7294 PR rtl-optimization/70023
7295 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7296 src_regno into account.
7298 2018-02-16 Carl Love <cel@us.ibm.com>
7300 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7301 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7302 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7303 * config/rs6000/rs6000.c: Remove case statements for
7304 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7305 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7306 and P9V_BUILTIN_VEC_VINSERT4B.
7307 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7308 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7309 * config/rs6000/vsx.md:
7310 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7313 2018-02-16 Carl Love <cel@us.ibm.com>
7315 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7317 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7319 * config/rs6000/rs6000-c.c: Add the definitions for
7320 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7321 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7322 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7323 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7324 definition for insert4b and define insn *insert3b_internal.
7325 * doc/extend.texi: Add documentation for vec_extract4b.
7327 2018-02-16 Nathan Sidwell <nathan@acm.org>
7329 * doc/extend.texi (Backwards Compatibility): Mention friend
7330 injection. Note for-scope is deprecated.
7331 * doc/invoke.texi (-ffriend-injection): Deprecate.
7333 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7335 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7336 that moved to I2, also allow destinations that are a paradoxical
7337 subreg (instead of a normal reg).
7339 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7342 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7345 2018-02-16 Richard Biener <rguenther@suse.de>
7347 PR tree-optimization/84037
7348 PR tree-optimization/84016
7350 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7351 Adjust vec_construct for the fact we need additional higher latency
7352 128bit inserts for AVX256 and AVX512 vector builds.
7353 (ix86_add_stmt_cost): Scale vector construction cost for
7356 2018-02-16 Richard Biener <rguenther@suse.de>
7358 PR tree-optimization/84417
7359 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7360 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7361 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7363 2018-02-16 Martin Liska <mliska@suse.cz>
7366 * internal-fn.def (ASAN_CHECK): Set proper flags.
7367 (ASAN_MARK): Likewise.
7369 2018-02-16 Julia Koval <julia.koval@intel.com>
7371 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7372 from PTA_CANNONLAKE.
7374 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7377 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7378 Use ++iter rather than iter++ for std::list iterators.
7379 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7380 defer deleting them until all nodes in the forest are processed. Do
7381 free even leaf nodes. Change to_process into auto_vec.
7384 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7385 * vec.h (vec_default_construct): Use memset instead of placement new
7386 if BROKEN_VALUE_INITIALIZATION is defined.
7387 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7388 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7391 PR rtl-optimization/83723
7392 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7393 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7394 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7396 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7398 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7400 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7402 PR rtl-optimization/81443
7403 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7404 from inner REGs to paradoxical SUBREGs.
7406 2018-02-16 Richard Biener <rguenther@suse.de>
7408 PR tree-optimization/84399
7409 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7410 For operands we can analyze at their definition make sure we can
7411 analyze them at each use as well.
7413 2018-02-16 Richard Biener <rguenther@suse.de>
7415 PR tree-optimization/84190
7416 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7417 volatile accesses if the decl isn't volatile.
7419 2018-02-15 Jason Merrill <jason@redhat.com>
7421 PR c++/84314 - ICE with templates and fastcall attribute.
7422 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7423 TYPE_CANONICAL on an existing type.
7425 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7427 PR tree-optimization/84383
7428 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7429 dstoff nor call operand_equal_p if dstbase is NULL.
7431 PR tree-optimization/84334
7432 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7433 also a CONSTANT_CLASS_P, punt.
7435 2018-02-14 Jim Wilson <jimw@sifive.com>
7437 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7438 first SMALL_OPERAND check. New local min_second_step. Move assert
7439 to where locals are set. Add TARGET_RVC support.
7440 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7442 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7444 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7446 2018-02-14 Martin Sebor <msebor@redhat.com>
7448 PR tree-optimization/83698
7449 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7450 arrays constrain the offset range to their bounds.
7451 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7452 (builtin_access::overlap): Avoid setting the size of overlap if it's
7454 (maybe_diag_overlap): Also consider arrays when deciding what values
7455 of offsets to include in diagnostics.
7457 2018-02-14 Martin Sebor <msebor@redhat.com>
7460 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7461 that correspond to the kind of a declaration.
7463 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7466 * config/pa/pa.md: Load address of PIC label using the linkage table
7467 if the label is nonlocal.
7469 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7471 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7472 warning message if user requests -maltivec=be.
7473 * doc/invoke.texi: Document deprecation of -maltivec=be.
7475 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7478 * config/rs6000/rs6000-c.c: Update definitions for
7479 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7480 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7482 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7485 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7486 add _get_ssp intrinsics. Remove argument from
7487 __builtin_ia32_rdssp[d|q].
7488 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7489 * config/i386/i386-builtin.def: Remove argument from
7490 __builtin_ia32_rdssp[d|q].
7491 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7492 ix86_expand_special_args_builtin for _rdssp[d|q].
7493 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7494 Clear register before usage.
7495 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7496 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7498 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7500 PR tree-optimization/84357
7501 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7502 operand 1 of an ARRAY_REF too.
7504 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7507 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7508 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7510 (set_of_reg): New struct.
7511 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7512 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7513 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7515 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7516 Split into bitclr, bitset, bitinvert patterns if appropriate.
7517 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7518 use rx_fuse_in_memory_bitop.
7519 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7520 to named insn, correct maximum insn length.
7522 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7525 * machmode.def: Define a complex mode for PARTIAL_INT.
7526 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7528 * doc/rtl.texi: Document CSPImode.
7529 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7531 (msp430_hard_regno_nregs_with_padding): Likewise.
7533 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7536 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7538 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7540 PR rtl-optimization/84169
7541 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7542 we generated a parallel as new i3 and we split that to new i2 and i3
7543 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7544 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7545 those to i2, not i1. Partially rewrite this scan code.
7547 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7550 * stor-layout.c (place_field): For variable length fields, adjust
7551 offset_align afterwards not just based on the field's alignment,
7552 but also on the size.
7555 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7556 of exps and logs in the use_exp2 case.
7558 2018-02-13 Jeff Law <law@redhat.com>
7560 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7563 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7566 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7570 * common.opt (gas-loc-support, gas-locview-support): New.
7571 (ginline-points, ginternal-reset-location-views): New.
7572 * doc/invoke.texi: Document them. Use @itemx where intended.
7573 (gvariable-location-views): Adjust.
7574 * target.def (reset_location_view): New.
7575 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7576 (TARGET_RESET_LOCATION_VIEW): New.
7577 * doc/tm.texi: Rebuilt.
7578 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7579 (dwarf2out_default_as_locview_support): New.
7580 (output_asm_line_debug_info): Use option variables.
7581 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7582 (output_loc_list): Likewise.
7583 (add_high_low_attributes): Check option variables.
7584 Don't output entry view attribute in strict mode.
7585 (gen_inlined_subroutine_die): Check option variables.
7586 (dwarf2out_inline_entry): Likewise.
7587 (init_sections_and_labels): Likewise.
7588 (dwarf2out_early_finish): Likewise.
7589 (maybe_reset_location_view): New, from...
7590 (dwarf2out_var_location): ... here. Call it.
7591 * debug.h (dwarf2out_default_as_loc_support): Declare.
7592 (dwarf2out_default_as_locview_support): Declare.
7593 * hooks.c (hook_int_rtx_insn_0): New.
7594 * hooks.h (hook_int_rtx_insn_0): Declare.
7595 * toplev.c (process_options): Take -gas-loc-support and
7596 -gas-locview-support from dwarf2out. Enable
7597 -gvariable-location-views by default only with locview
7598 assembler support. Enable -ginternal-reset-location-views by
7599 default only if the target defines the corresponding hook.
7600 Enable -ginline-points by default if location views are
7601 enabled; force it disabled if statement frontiers are
7603 * tree-inline.c (expand_call_inline): Check option variables.
7604 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7606 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7608 PR tree-optimization/84321
7609 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7610 handling. Also check whether the anti-range contains any values
7611 that satisfy the mask; switch to a VR_RANGE if not.
7613 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7616 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7618 2018-02-13 Martin Jambor <mjambor@suse.cz>
7621 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7622 of call statements, also set location of a load to a temporary.
7624 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7626 * config/rl78/rl78.c (add_vector_labels): New function.
7627 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7628 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7629 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7630 which checks that no arguments are passed.
7631 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7632 * doc/extend.texi: Documentation for the new attribute.
7634 2018-02-13 Andreas Schwab <schwab@suse.de>
7636 * config/riscv/linux.h (CPP_SPEC): Define.
7638 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7641 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7642 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7643 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7644 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7645 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7646 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7647 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7648 temporarily for AES and PCLMUL builtins.
7650 PR tree-optimization/84339
7651 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7652 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7656 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7657 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7658 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7660 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7662 * omp-simd-clone.h: New file.
7663 * omp-simd-clone.c: Include omp-simd-clone.h.
7664 (expand_simd_clones): No longer static.
7665 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7666 cgraph.h and omp-simd-clone.h.
7667 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7668 (vect_recog_widen_shift_pattern): Formatting fix.
7669 (vect_pattern_recog_1): Don't check optab for calls.
7672 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7673 operands[2] into a REG before using gen_lowpart on it.
7675 2018-02-12 Jeff Law <law@redhat.com>
7678 * config/sh/sh.c (find_barrier): Consider a sibling call
7681 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7682 successfully back substituting a reg.
7684 2018-02-12 Richard Biener <rguenther@suse.de>
7686 PR tree-optimization/84037
7687 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7688 parameter, move visited init to caller.
7689 (vect_slp_analyze_operations): Separate cost from validity
7690 check, initialize visited once for all instances.
7691 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7693 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7695 (vect_model_promotion_demotion_cost): Likewise.
7696 (vectorizable_bswap): Guard cost modeling with !slp_node
7697 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7699 (vectorizable_call): Likewise.
7700 (vectorizable_conversion): Likewise.
7701 (vectorizable_assignment): Likewise.
7702 (vectorizable_shift): Likewise.
7703 (vectorizable_operation): Likewise.
7704 (vectorizable_store): Likewise.
7705 (vectorizable_load): Likewise.
7706 (vectorizable_condition): Likewise.
7707 (vectorizable_comparison): Likewise.
7709 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7712 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7713 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7714 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7716 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7719 * params.def (inline-min-speedup): Increase from 8 to 15.
7720 (max-inline-insns-auto): Decrease from 40 to 30.
7721 * ipa-split.c (consider_split): Add some buffer for function to
7722 be considered inlining candidate.
7723 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7726 2018-02-12 Richard Biener <rguenther@suse.de>
7728 PR tree-optimization/84037
7729 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7730 matched stmts if we cannot swap the non-matched ones.
7732 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7734 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7735 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7736 _mm_maskz_scalef_round_ss): New intrinsics.
7737 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7738 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7739 __builtin_ia32_scalefss_round): Remove.
7740 (__builtin_ia32_scalefsd_mask_round,
7741 __builtin_ia32_scalefss_mask_round): New intrinsics.
7742 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7743 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7744 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7745 "<round_constraint>")): Changed to ...
7746 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7747 "<round_scalar_constraint>")): ... this.
7748 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7749 %0, %1, %2<round_op3>}"): Changed to ...
7750 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7751 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7752 %2<round_scalar_mask_op3>}"): ... this.
7753 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7755 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7757 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7758 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7759 (_mm_maskz_sqrt_round_ss): New intrinsics.
7760 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7761 (__builtin_ia32_sqrtsd_mask_round)
7762 (__builtin_ia32_sqrtss_mask_round): New builtins.
7763 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7764 (__builtin_ia32_sqrtss_round): Remove.
7765 (__builtin_ia32_sqrtsd_mask_round)
7766 (__builtin_ia32_sqrtss_mask_round): New builtins.
7767 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7768 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7769 ((match_operand:VF_128 1 "vector_operand"
7770 "xBm,<round_constraint>")): Changed to ...
7771 ((match_operand:VF_128 1 "vector_operand"
7772 "xBm,<round_scalar_constraint>")): ... this.
7773 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7774 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7775 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7776 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7777 %<iptr>1<round_scalar_mask_op3>}): ... this.
7778 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7779 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7781 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7784 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7785 Cast vec_cmpeq result to correct type.
7786 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7787 Cast vec_cmpgt result to correct type.
7789 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7791 * final.c (final_scan_insn_1): Renamed from...
7792 (final_scan_insn): ... this. New wrapper, to recover
7793 seen from the outermost call in recursive ones.
7794 * config/sparc/sparc.c (output_return): Drop seen from call.
7795 (output_sibcall): Likewise.
7796 * config/visium/visium.c (output_branch): Likewise.
7798 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7800 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7803 2018-02-10 Alan Modra <amodra@gmail.com>
7806 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7807 Specify LR as an input.
7809 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7812 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7813 remove_member_access_dummy_vars): New functions.
7814 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7815 lower_omp_1, execute_lower_omp): Use them.
7817 PR rtl-optimization/84308
7818 * shrink-wrap.c (spread_components): Release todo vector.
7820 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7822 PR rtl-optimization/57193
7823 * ira-color.c (struct allocno_color_data): Add member
7824 conflict_allocno_hard_prefs.
7825 (update_conflict_allocno_hard_prefs): New.
7826 (bucket_allocno_compare_func): Add a preference based on
7827 conflict_allocno_hard_prefs.
7828 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7829 (color_allocnos): Remove a dead code. Initiate
7830 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7832 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7835 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7836 constraint from =wa to wa. Avoid a subreg on the output operand,
7837 instead use a pseudo and subreg it in a move.
7838 (p9_xxbrd_<mode>): Changed to ...
7839 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7840 (p9_xxbrd_v2df): New expander.
7841 (p9_xxbrw_<mode>): Changed to ...
7842 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7843 (p9_xxbrw_v4sf): New expander.
7845 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7847 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7849 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7852 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7853 multiply in 32-bit mode.
7854 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7855 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7858 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7860 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7861 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7862 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7863 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7865 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7868 * dwarf2out.c (is_trivial_indirect_ref): New function.
7869 (dwarf2out_late_global_decl): Do not generate a location
7870 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7871 and that are not defined in the current unit.
7873 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7875 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7876 instead of a libcall for UNORDERED.
7878 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7881 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7882 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7884 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7887 * config/s390/s390.c (s390_set_current_function): Invoke
7888 s390_indirect_branch_settings also if fndecl didn't change.
7890 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7892 * config/rs6000/rs6000.md (blockage): Set length to zero.
7894 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7896 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7898 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7901 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7902 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7906 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7907 PARALLEL incoming that failed vt_get_decl_and_offset check.
7910 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7911 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7913 (get_variable_section): For decls in named .bss* sections pass true as
7914 second argument to bss_initializer_p.
7916 2018-02-09 Marek Polacek <polacek@redhat.com>
7917 Jakub Jelinek <jakub@redhat.com>
7920 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7921 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7922 Sync some changes from cxx_fold_indirect_ref.
7924 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7926 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7928 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7929 (BLOCK_INLINE_ENTRY_LABEL): New.
7930 (dwarf2out_var_location): Disregard inline entry markers.
7931 (inline_entry_data): New struct.
7932 (inline_entry_data_hasher): New hashtable type.
7933 (inline_entry_data_hasher::hash): New.
7934 (inline_entry_data_hasher::equal): New.
7935 (inline_entry_data_table): New variable.
7936 (add_high_low_attributes): Add DW_AT_entry_pc and
7937 DW_AT_GNU_entry_view attributes if a pending entry is found
7938 in inline_entry_data_table. Add old entry_pc attribute only
7939 if debug nonbinding markers are disabled.
7940 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7941 markers are enabled.
7942 (block_within_block_p, dwarf2out_inline_entry): New.
7943 (dwarf2out_finish): Check that no entries remained in
7944 inline_entry_data_table.
7945 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7946 (final_scan_insn, notice_source_line): Likewise.
7947 (rest_of_clean_state): Skip inline entry markers.
7948 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7950 * gimple.c (gimple_build_debug_inline_entry): New.
7951 * gimple.h (enum gimple_debug_subcode): Add
7952 GIMPLE_DEBUG_INLINE_ENTRY.
7953 (gimple_build_debug_inline_entry): Declare.
7954 (gimple_debug_inline_entry_p): New.
7955 (gimple_debug_nonbind_marker_p): Adjust.
7956 * insn-notes.def (INLINE_ENTRY): New.
7957 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7958 inline entry marker notes.
7959 (print_insn): Likewise.
7960 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7961 (INSN_DEBUG_MARKER_KIND): Likewise.
7962 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7963 * tree-inline.c (expand_call_inline): Build and insert
7964 debug_inline_entry stmt.
7965 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7966 inline entry blocks early, if nonbind markers are enabled.
7967 (dump_scope_block): Dump fragment info.
7968 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7969 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7970 (gimple_build_debug_inline_entry): New.
7971 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7972 Enable/disable inline entry points too.
7973 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7974 (DEBUG_INSN): Describe inline entry markers.
7976 * common.opt (gvariable-location-views): New.
7977 (gvariable-location-views=incompat5): New.
7978 * config.in: Rebuilt.
7979 * configure: Rebuilt.
7980 * configure.ac: Test assembler for view support.
7981 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7982 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7983 * dwarf2out.c (var_loc_view): New typedef.
7984 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7985 (dwarf2out_locviews_in_attribute): New.
7986 (dwarf2out_locviews_in_loclist): New.
7987 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7988 (enum dw_line_info_opcode): Add LI_adv_address.
7989 (struct dw_line_info_table): Add view.
7990 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7991 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7992 (zero_view_p): New variable.
7993 (ZERO_VIEW_P): New macro.
7994 (output_asm_line_debug_info): New.
7995 (struct var_loc_node): Add view.
7996 (add_AT_view_list, AT_loc_list): New.
7997 (add_var_loc_to_decl): Add view param. Test it against last.
7998 (new_loc_list): Add view params. Record them.
7999 (AT_loc_list_ptr): Handle loc and view lists.
8000 (view_list_to_loc_list_val_node): New.
8001 (print_dw_val): Handle dw_val_class_view_list.
8002 (size_of_die): Likewise.
8003 (value_format): Likewise.
8004 (loc_list_has_views): New.
8005 (gen_llsym): Set vl_symbol too.
8006 (maybe_gen_llsym, skip_loc_list_entry): New.
8007 (dwarf2out_maybe_output_loclist_view_pair): New.
8008 (output_loc_list): Output view list or entries too.
8009 (output_view_list_offset): New.
8010 (output_die): Handle dw_val_class_view_list.
8011 (output_dwarf_version): New.
8012 (output_compilation_unit_header): Use it.
8013 (output_skeleton_debug_sections): Likewise.
8014 (output_rnglists, output_line_info): Likewise.
8015 (output_pubnames, output_aranges): Update version comments.
8016 (output_one_line_info_table): Output view numbers in asm comments.
8017 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8018 Call maybe_gen_llsym.
8019 (loc_list_from_tree_1): Adjust.
8020 (add_AT_location_description): Create view list attribute if
8021 needed, check it's absent otherwise.
8022 (convert_cfa_to_fb_loc_list): Adjust.
8023 (maybe_emit_file): Call output_asm_line_debug_info for test.
8024 (dwarf2out_var_location): Reset views as needed. Precompute
8025 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8026 attribute. Set view.
8027 (new_line_info_table): Reset next view.
8028 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8029 (dwarf2out_source_line): Likewise. Output view resets and labels to
8030 the assembler, or select appropriate line info opcodes.
8031 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8032 (optimize_string_length): Catch it. Adjust.
8033 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8034 dw_val_class_view_list, and remove it if no longer needed.
8035 (hash_loc_list): Hash view numbers.
8036 (loc_list_hasher::equal): Compare them.
8037 (optimize_location_lists): Check whether a view list symbol is
8038 needed, and whether the locview attribute is present, and
8039 whether they match. Remove the locview attribute if no longer
8041 (index_location_lists): Call skip_loc_list_entry for test.
8042 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8043 Use output_dwarf_version.
8044 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8045 (struct dw_val_node): Add val_view_list.
8046 * final.c (SEEN_NEXT_VIEW): New.
8047 (set_next_view_needed): New.
8048 (clear_next_view_needed): New.
8049 (maybe_output_next_view): New.
8050 (final_start_function): Rename to...
8051 (final_start_function_1): ... this. Take pointer to FIRST,
8052 add SEEN parameter. Emit param bindings in the initial view.
8053 (final_start_function): Reintroduce SEEN-less interface.
8054 (final): Rename to...
8055 (final_1): ... this. Take SEEN parameter. Output final pending
8056 next view at the end.
8057 (final): Reintroduce seen-less interface.
8058 (final_scan_insn): Output pending next view before switching
8059 sections or ending a block. Mark the next view as needed when
8060 outputting variable locations. Notify debug backend of section
8061 changes, and of location view changes.
8062 (rest_of_handle_final): Adjust.
8063 * toplev.c (process_options): Autodetect value for debug variable
8064 location views option. Warn on incompat5 without -gdwarf-5.
8065 * doc/invoke.texi (gvariable-location-views): New.
8066 (gvariable-location-views=incompat5): New.
8067 (gno-variable-location-views): New.
8069 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8071 PR tree-optimization/84136
8072 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8073 that the result of find_edge is non-NULL.
8075 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8078 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8079 storing integer register in SImode. Fix cost of 256 and 512
8080 byte aligned SSE register store.
8082 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8084 * config/i386/i386.c (ix86_multiplication_cost): Fix
8085 multiplication cost for TARGET_AVX512DQ.
8087 2018-02-08 Marek Polacek <polacek@redhat.com>
8089 PR tree-optimization/84238
8090 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8093 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8095 PR tree-optimization/84265
8096 * tree-vect-stmts.c (vectorizable_store): Don't treat
8097 VMAT_CONTIGUOUS accesses as grouped.
8098 (vectorizable_load): Likewise.
8100 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8102 PR tree-optimization/81635
8103 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8104 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8105 (test_round_for_mask): New functions.
8106 (wide_int_cc_tests): Call test_round_for_mask.
8107 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8108 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8109 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8110 range returned by get_range_info.
8112 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8115 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8116 * symtab.c: Include builtins.h
8117 (symtab_node::output_to_lto_symbol_table_p): Move here
8118 from lto-streamer-out.c:output_symbol_p.
8119 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8120 (output_symbol_p): Move all logic to symtab.c
8121 (produce_symtab): Update.
8123 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8125 * config/s390/s390-opts.h (enum indirect_branch): Define.
8126 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8127 (s390_indirect_branch_via_thunk)
8128 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8129 (enum s390_indirect_branch_type): Define.
8130 * config/s390/s390.c (struct s390_frame_layout, struct
8131 machine_function): Remove.
8132 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8133 (indirect_branch_table_label_no, indirect_branch_table_name):
8135 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8136 (enum s390_indirect_branch_option): Define.
8137 (s390_return_addr_from_memory): New function.
8138 (s390_handle_string_attribute): New function.
8139 (s390_attribute_table): Add new attribute handler.
8140 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8141 (s390_indirect_branch_via_thunk): New function.
8142 (s390_indirect_branch_via_inline_thunk): New function.
8143 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8144 sibling call optimization for non z10 compiles.
8145 (s390_emit_call): Force indirect branch target to be a single
8146 register. Add r1 clobber for non-z10 compiles.
8147 (s390_emit_epilogue): Emit return jump via return_use expander.
8148 (s390_reorg): Handle JUMP_INSNs as execute targets.
8149 (s390_option_override_internal): Perform validity checks for the
8150 new command line options.
8151 (s390_indirect_branch_attrvalue): New function.
8152 (s390_indirect_branch_settings): New function.
8153 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8154 (s390_output_indirect_thunk_function): New function.
8155 (s390_code_end): Implement target hook.
8156 (s390_case_values_threshold): Implement target hook.
8157 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8159 * config/s390/s390.h (struct s390_frame_layout)
8160 (struct machine_function): Move here from s390.c.
8161 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8162 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8163 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8164 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8165 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8166 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8167 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8168 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8169 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8170 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8171 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8172 (mnemonic attribute): Add values which aren't recognized
8174 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8175 pattern for branch conversion. Fix mnemonic attribute.
8176 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8177 indirect branch via thunk if requested.
8178 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8179 ("*indirect_jump"): Disable for branch conversion using out of
8181 ("indirect_jump_via_thunk<mode>_z10")
8182 ("indirect_jump_via_thunk<mode>")
8183 ("indirect_jump_via_inlinethunk<mode>_z10")
8184 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8185 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8186 ("casesi_jump_via_inlinethunk<mode>_z10")
8187 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8188 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8189 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8190 ("*indirect2_jump"): Disable for branch conversion.
8191 ("casesi_jump"): Turn into expander and expand patterns for branch
8193 ("return_use"): New expander.
8194 ("*return"): Emit return via thunk and rename it to ...
8195 ("*return<mode>"): ... this one.
8196 * config/s390/s390.opt: Add new options and and enum for the
8199 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8201 * lra-constraints.c (match_reload): Unconditionally use
8202 gen_lowpart_SUBREG, rather than selecting between that
8203 and equivalent gen_rtx_SUBREG code.
8205 2018-02-08 Richard Biener <rguenther@suse.de>
8207 PR tree-optimization/84233
8208 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8209 changed flag instead of boguously re-using phi_inserted.
8211 2018-02-08 Martin Jambor <mjambor@suse.cz>
8213 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8214 static local variables.
8216 2018-02-08 Richard Biener <rguenther@suse.de>
8218 PR tree-optimization/84278
8219 * tree-vect-stmts.c (vectorizable_store): When looking for
8220 smaller vector types to perform grouped strided loads/stores
8221 make sure the mode is supported by the target.
8222 (vectorizable_load): Likewise.
8224 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8226 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8227 Increase LDP/STP opportunities by adding adjacent callee-saves.
8229 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8231 PR rtl-optimization/84068
8232 PR rtl-optimization/83459
8233 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8235 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8237 PR tree-optimization/84224
8238 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8239 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8242 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8245 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8246 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8247 count, remove one USE.
8249 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8251 * doc/install.texi (Configuration): Document the
8252 --with-long-double-format={ibm,ieee} PowerPC configuration
8256 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8257 Convert from define_expand to be define_insn_and_split. Rework
8258 float/double/_Float128 conversions to QI/HI/SImode to work with
8259 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8260 conversions to QI/HImode types did a store and then a load to
8261 truncate the value. For conversions to VSX registers, don't split
8262 the insn, instead emit the code directly. Use the code iterator
8263 any_fix to combine signed and unsigned conversions.
8264 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8265 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8266 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8267 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8268 (fix_<mode>di2_hw): Likewise.
8269 (fixuns_<mode>di2_hw): Likewise.
8270 (fix_<mode>si2_hw): Likewise.
8271 (fixuns_<mode>si2_hw): Likewise.
8272 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8273 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8274 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8275 fix<uns>_trunc<SFDF:mode>si2_p8.
8276 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8278 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8279 (fix<uns>_<mode>_mem): Likewise.
8280 (fctiw<u>z_<mode>_mem): Likewise.
8281 (fix<uns>_<mode>_mem): Likewise.
8282 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8283 the register allocator from doing a direct move to the GPRs to do
8284 a store, and instead use the ISA 3.0 store byte/half-word from
8285 vector register instruction. For IEEE 128-bit floating point,
8286 also optimize stores of 32-bit ints.
8287 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8289 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8291 * genextract.c (push_pathstr_operand): New function to support
8293 (walk_rtx): Call push_pathstr_operand.
8294 (print_path): Support [a-zA-Z].
8296 2018-02-07 Richard Biener <rguenther@suse.de>
8298 PR tree-optimization/84037
8299 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8300 (cse_and_gimplify_to_preheader): Declare.
8301 (vect_get_place_in_interleaving_chain): Likewise.
8302 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8304 (_loop_vec_info::~_loop_vec_info): Delete it.
8305 (cse_and_gimplify_to_preheader): New function.
8306 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8307 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8308 (vectorizable_load): Likewise. For grouped stores always base
8309 the IV on the first element.
8310 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8311 condition before gimplifying.
8313 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8315 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8316 *DIV_EXPR and *MOD_EXPR.
8318 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8321 * config/i386/i386.c (ix86_option_override_internal): Mask out
8322 the CF_SET bit when checking -fcf-protection.
8324 2018-02-07 Tom de Vries <tom@codesourcery.com>
8327 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8330 2018-02-07 Richard Biener <rguenther@suse.de>
8332 PR tree-optimization/84204
8333 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8336 PR tree-optimization/84205
8337 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8338 special-case isl_ast_op_zdiv_r.
8340 PR tree-optimization/84223
8341 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8342 Only add conditions from within the region.
8343 (gather_bbs::after_dom_children): Adjust.
8345 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8348 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8349 * config/avr/avr.md: Only post-reload split REG-REG moves if
8350 either register is GENERAL_REG_P.
8352 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8354 PR tree-optimization/84235
8355 * tree-ssa-scopedtables.c
8356 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8357 if the subtraction is performed in floating point type where NaNs are
8358 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8359 build 1. Formatting fix.
8361 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8364 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8365 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8366 and skip it regardless of bb boundaries. Use CALL_P macro,
8367 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8368 unnecessarily, formatting fix.
8370 2018-02-06 Michael Collison <michael.collison@arm.com>
8372 * config/arm/thumb2.md:
8373 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8374 (*thumb_mov_notscc): Ditto.
8376 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8379 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8380 unsigned_fix, not "s".
8382 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8384 * configure.ac (gcc_fn_eh_frame_ro): New function.
8385 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8386 correct .eh_frame permissions.
8387 * configure: Regenerate.
8389 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8391 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8394 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8396 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8397 Display warning message for -mno-speculate-indirect-jumps.
8399 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8401 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8403 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8405 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8407 PR tree-optimization/84225
8408 * tree-eh.c (find_trapping_overflow): Only call
8409 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8411 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8414 * config/i386/i386.c: Reimplement the check of possible options
8415 -mibt/-mshstk conbination. Change error messages.
8416 * doc/invoke.texi: Fix a typo: remove extra '='.
8418 2018-02-06 Marek Polacek <polacek@redhat.com>
8420 PR tree-optimization/84228
8421 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8423 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8426 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8427 emitted arch directives.
8428 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8429 __ARM_FEATURE_COPROC before changing architectures.
8431 2018-02-06 Richard Biener <rguenther@suse.de>
8433 * config/i386/i386.c (print_reg): Fix typo.
8434 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8436 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8438 * configure: Regenerate.
8440 2018-02-05 Martin Sebor <msebor@redhat.com>
8442 PR tree-optimization/83369
8443 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8446 2018-02-05 Martin Liska <mliska@suse.cz>
8448 * doc/invoke.texi: Cherry-pick upstream r323995.
8450 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8452 * ira.c (ira_init_register_move_cost): Adjust comment.
8454 2018-02-05 Martin Liska <mliska@suse.cz>
8456 PR gcov-profile/84137
8457 * doc/gcov.texi: Fix typo in documentation.
8459 2018-02-05 Martin Liska <mliska@suse.cz>
8461 PR gcov-profile/83879
8462 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8463 using dlopen functionality.
8465 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8467 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8468 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8469 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8470 _mm_maskz_range_round_ss): New intrinsics.
8471 (__builtin_ia32_rangesd128_round)
8472 (__builtin_ia32_rangess128_round): Remove.
8473 (__builtin_ia32_rangesd128_mask_round,
8474 __builtin_ia32_rangess128_mask_round): New builtins.
8475 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8476 __builtin_ia32_rangess128_round): Remove.
8477 (__builtin_ia32_rangesd128_mask_round,
8478 __builtin_ia32_rangess128_mask_round): New builtins.
8479 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8480 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8481 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8482 "<round_saeonly_constraint>")): Changed to ...
8483 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8484 "<round_saeonly_scalar_constraint>")): ... this.
8485 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8486 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8487 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8488 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8489 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8491 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8493 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8495 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8496 Remove all values except native, 8540 and 8548.
8498 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8500 * config/i386/i386.c (ix86_output_function_return): Pass
8501 INVALID_REGNUM, instead of -1, as invalid register number to
8502 indirect_thunk_name and output_indirect_thunk.
8504 2018-02-02 Julia Koval <julia.koval@intel.com>
8506 * config.gcc: Add -march=icelake.
8507 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8508 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8509 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8510 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8511 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8512 (processor_target_table): Add icelake.
8513 (ix86_option_override_internal): Handle new PTAs.
8514 (get_builtin_code_for_version): Handle icelake.
8515 (M_INTEL_COREI7_ICELAKE): New.
8516 (fold_builtin_cpu): Handle icelake.
8517 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8518 * doc/invoke.texi: Add -march=icelake.
8520 2018-02-02 Julia Koval <julia.koval@intel.com>
8522 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8523 to wide_int_bitmask.
8524 * wide-int-bitmask.h: New.
8526 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8529 * config/i386/i386.md: Replace Pmode with word_mode in
8530 builtin_setjmp_setup and builtin_longjmp to support x32.
8532 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8536 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8538 (rs6000_supported_cpu_names): New static variable.
8539 (linux_cpu_translation_table): Likewise.
8540 (elf_platform) <cpu>: Define new static variable and use it.
8541 Translate kernel AT_PLATFORM name to canonical name if needed.
8542 Error if platform name is unknown.
8544 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8547 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8549 2018-02-01 Jeff Law <law@redhat.com>
8552 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8553 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8554 the scratch if RELEASE_VIA_POP is false.
8555 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8556 If we have to save a temporary register, decrement SIZE appropriately.
8557 Pass new arguments to release_scratch_register_on_entry.
8558 (ix86_adjust_stack_and_probe): Likewise.
8559 (ix86_emit_probe_stack_range): Pass new arguments to
8560 release_scratch_register_on_entry.
8562 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8564 PR rtl-optimization/84157
8565 * combine.c (change_zero_ext): Use REG_P predicate in
8566 front of HARD_REGISTER_P predicate.
8568 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8570 * config/avr/avr.c (avr_option_override): Move disabling of
8571 -fdelete-null-pointer-checks to...
8572 * common/config/avr/avr-common.c (avr_option_optimization_table):
8575 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8577 PR tree-optimization/81635
8578 * tree-data-ref.c (split_constant_offset_1): For types that
8579 wrap on overflow, try to use range info to prove that wrapping
8582 2018-02-01 Renlin Li <renlin.li@arm.com>
8585 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8587 (aarch64_register_move_cost): Likewise.
8588 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8590 (REG_CLASS_NAMES): Likewise.
8591 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8592 TAILCALL_ADDR_REGS. Remove IP registers.
8593 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8595 2018-02-01 Richard Biener <rguenther@suse.de>
8597 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8598 for specifying RPO order and allow NULL for that.
8599 * domwalk.c (dom_walker::dom_walker): Likewise.
8600 (dom_walker::walk): Handle NULL RPO order.
8601 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8603 (rewrite_update_dom_walker): Likewise.
8604 (mark_def_dom_walker): Likewise.
8606 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8608 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8609 (aarch64_maybe_expand_sve_subreg_move): Declare.
8610 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8611 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8613 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8614 that are semantically a reverse operation.
8615 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8616 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8617 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8619 (aarch64_can_change_mode_class): For big-endian, forbid changes
8620 between two SVE modes if they have different element sizes.
8622 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8624 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8625 the TImode handling for big-endian targets.
8627 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8629 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8630 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8632 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8633 Remove BSWAP handing for big-endian targets and use the form of
8634 LD1RQ appropariate for the mode.
8636 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8638 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8639 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8642 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8645 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8646 check for operands that need to go through aarch64_sve_reload_be.
8648 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8650 PR tree-optimization/81661
8651 PR tree-optimization/84117
8652 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8653 * tree-eh.c: Include gimplify.h.
8654 (find_trapping_overflow, replace_trapping_overflow,
8655 rewrite_to_non_trapping_overflow): New functions.
8656 * tree-vect-loop.c: Include tree-eh.h.
8657 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8658 * tree-data-ref.c: Include tree-eh.h.
8659 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8661 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8663 PR rtl-optimization/84123
8664 * combine.c (change_zero_ext): Check if hard register satisfies
8665 can_change_dest_mode before calling gen_lowpart_SUBREG.
8667 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8670 * ira.c (ira_init_register_move_cost): Remove assert.
8672 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8674 PR rtl-optimization/84071
8675 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8676 * doc/tm.texi: Regenerate.
8678 2018-01-31 Richard Biener <rguenther@suse.de>
8680 PR tree-optimization/84132
8681 * tree-data-ref.c (analyze_miv_subscript): Properly
8682 check whether evolution_function_is_affine_multivariate_p
8683 before calling gcd_of_steps_may_divide_p.
8685 2018-01-31 Julia Koval <julia.koval@intel.com>
8688 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8689 * config/i386/i386.md (rdpid_rex64) New.
8690 (rdpid): Make 32bit only.
8692 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8695 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8696 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8698 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8701 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8703 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8705 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8707 PR rtl-optimization/84071
8708 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8709 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8711 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8713 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8714 (arc_attribute_table): Add 'aux' attribute.
8715 (arc_in_small_data_p): Consider aux like variables.
8716 (arc_is_aux_reg_p): New function.
8717 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8718 (arc_get_aux_arg): New function.
8719 (prepare_move_operands): Handle aux-register access.
8720 (arc_handle_aux_attribute): New function.
8721 * doc/extend.texi (ARC Variable attributes): Add subsection.
8723 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8725 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8726 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8727 (arc_attribute_table): Add 'uncached' attribute.
8728 (arc_print_operand): Print '.di' flag for uncached memory
8730 (arc_in_small_data_p): Do not consider for small data the uncached
8732 (arc_is_uncached_mem_p): New function.
8733 * config/arc/predicates.md (compact_store_memory_operand): Check
8734 for uncached memory accesses.
8735 (nonvol_nonimm_operand): Likewise.
8736 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8738 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8741 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8742 falign-loops=): Add Optimization flag.
8744 2018-01-30 Jeff Law <law@redhat.com>
8747 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8748 INT_REGISTERS_SAVED. Check it prior to calling
8749 get_scratch_register_on_entry.
8750 (ix86_adjust_stack_and_probe): Similarly.
8751 (ix86_emit_probe_stack_range): Similarly.
8752 (ix86_expand_prologue): Corresponding changes.
8754 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8757 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8758 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8760 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8763 * lra-constraints.c (curr_insn_transform): Process AND in the
8766 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8768 PR rtl-optimization/83986
8769 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8770 dependence against last_pending_memory_flush in addition to
8773 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8775 PR tree-optimization/81611
8776 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8779 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8782 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8785 2018-01-30 Richard Biener <rguenther@suse.de>
8786 Jakub Jelinek <jakub@redhat.com>
8788 PR tree-optimization/84111
8789 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8790 inner loops added during recursion, as they don't have up-to-date
8793 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8796 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8797 (can_inline_edge_by_limits_p): ... here.
8798 (can_early_inline_edge_p, check_callers,
8799 update_caller_keys, update_callee_keys, recursive_inlining,
8800 add_new_edges_to_heap, speculation_useful_p,
8801 inline_small_functions,
8802 inline_small_functions, flatten_function,
8803 inline_to_all_callers_1): Update.
8805 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8807 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8810 2018-01-30 Richard Biener <rguenther@suse.de>
8812 PR tree-optimization/83008
8813 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8814 invariant and constant vector uses in stmts when they need
8817 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8820 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8821 * configure: Regenerate.
8823 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8825 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8827 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8828 Use gen_rtx_REG rather than gen_lowpart.
8830 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8832 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8833 rather than 0 when creating partial subregs.
8835 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8837 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8840 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8843 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8844 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8845 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8846 flags. This restores the settings used before the 2017-07-24.
8847 Turning off pre increment/decrement/modify allows IVOPTS to
8848 optimize DF/SF loops where the index is an int.
8850 2018-01-29 Richard Biener <rguenther@suse.de>
8851 Kelvin Nilsen <kelvin@gcc.gnu.org>
8854 * tree-vect-stmts.c (vectorizable_call): Don't call
8855 targetm.vectorize_builtin_md_vectorized_function if callee is
8858 2018-01-22 Carl Love <cel@us.ibm.com>
8860 * doc/extend.tex: Fix typo in second arg in
8861 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8863 2018-01-29 Richard Biener <rguenther@suse.de>
8865 PR tree-optimization/84086
8866 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8867 (flush_ssaname_freelist): When SSA names were released reset
8868 the SCEV hash table.
8870 2018-01-29 Richard Biener <rguenther@suse.de>
8872 PR tree-optimization/84057
8873 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8874 removed paths when removing edges.
8876 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8878 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8879 -mfunction-return=@var{choice}.
8881 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8884 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8886 (layout::print_source_line): Likewise.
8887 (test_get_line_width_without_trailing_whitespace): Add test cases.
8889 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8892 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8895 2018-01-26 Jim Wilson <jimw@sifive.com>
8897 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8899 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8902 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8904 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8905 and CMP + SUB-immediate -> SUBS.
8907 2018-01-26 Martin Sebor <msebor@redhat.com>
8909 PR tree-optimization/83896
8910 * tree-ssa-strlen.c (get_string_len): Rename...
8911 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8912 Avoid assuming length is constant.
8913 (handle_char_store): Use HOST_WIDE_INT for string length.
8915 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8918 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8919 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8921 2018-01-26 Richard Biener <rguenther@suse.de>
8923 PR rtl-optimization/84003
8924 * dse.c (record_store): Only record redundant stores when
8925 the earlier store aliases at least all accesses the later one does.
8927 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8929 PR rtl-optimization/83985
8930 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8931 REG_CFA_RESTORE insns.
8932 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8933 REG_CFA_RESTORE insns here.
8936 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8937 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8939 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8941 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8942 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8943 (arc_init): Likewise.
8944 (arc_override_options): Likewise.
8945 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8947 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8949 * config/arc/arc.h (TARGET_DBNZ): Define.
8950 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8951 properly set the tune attribute.
8952 (dbnz): Use TARGET_DBNZ guard.
8953 * config/arc/arc.opt (mtune): Add core3 option.
8955 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8957 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8958 recognize new pic like addresses.
8959 (arc_delegitimize_address): Clean up.
8961 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8963 * config/arc/arc-arches.def: Option mrf16 valid for all
8965 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8966 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8967 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8968 * config/arc/arc-tables.opt: Regenerate.
8969 * config/arc/arc.c (arc_conditional_register_usage): Handle
8970 reduced register file case.
8971 (arc_file_start): Set must have build attributes.
8972 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8974 * config/arc/arc.opt (mrf16): Add new option.
8975 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8976 * config/arc/genmultilib.awk: Handle new mrf16 option.
8977 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8978 * config/arc/t-multilib: Regenerate.
8979 * doc/invoke.texi (ARC Options): Document mrf16 option.
8981 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8983 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8984 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8985 (arc_attribute_table): Add 'secure_call' attribute.
8986 (arc_print_operand): Print secure call operand.
8987 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8989 (arc_is_secure_call_p): New function. * config/arc/arc.md
8990 (call_i): Add support for sjli instruction.
8991 (call_value_i): Likewise.
8992 * config/arc/constraints.md (Csc): New constraint.
8994 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8995 John Eric Martin <John.Martin@emmicro-us.com>
8997 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8998 * config/arc/arc.c (_arc_jli_section): New struct.
8999 (arc_jli_section): New type.
9000 (rc_jli_sections): New static variable.
9001 (arc_handle_jli_attribute): New function.
9002 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9003 (arc_file_end): New function.
9004 (TARGET_ASM_FILE_END): Define.
9005 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9006 (arc_add_jli_section): New function.
9007 (jli_call_scan): Likewise.
9008 (arc_reorg): Call jli_call_scan.
9009 (arc_output_addsi): Remove 'S' from printing asm operand.
9010 (arc_is_jli_call_p): New function.
9011 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9013 (movhi_insn): Likewise.
9014 (movsi_insn): Likewise.
9015 (movsi_set_cc_insn): Likewise.
9016 (loadqi_update): Likewise.
9017 (load_zeroextendqisi_update): Likewise.
9018 (load_signextendqisi_update): Likewise.
9019 (loadhi_update): Likewise.
9020 (load_zeroextendhisi_update): Likewise.
9021 (load_signextendhisi_update): Likewise.
9022 (loadsi_update): Likewise.
9023 (loadsf_update): Likewise.
9024 (movsicc_insn): Likewise.
9025 (bset_insn): Likewise.
9026 (bxor_insn): Likewise.
9027 (bclr_insn): Likewise.
9028 (bmsk_insn): Likewise.
9029 (bicsi3_insn): Likewise.
9030 (cmpsi_cc_c_insn): Likewise.
9031 (movsi_ne): Likewise.
9032 (movsi_cond_exec): Likewise.
9033 (clrsbsi2): Likewise.
9044 (call_i): Remove 'S' asm letter, add jli instruction.
9045 (call_value_i): Likewise.
9046 * config/arc/arc.op (mjli-always): New option.
9047 * config/arc/constraints.md (Cji): New constraint.
9048 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9050 (subsf3_fpx): Likewise.
9051 (mulsf3_fpx): Likewise.
9052 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9054 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9056 * doc/invoke.texi (ARC): Document mjli-always option.
9058 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9060 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9061 avoid addition with 0 and use incw and decw where possible.
9063 2018-01-26 Richard Biener <rguenther@suse.de>
9065 PR tree-optimization/81082
9066 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9067 association if it requires casting to unsigned.
9068 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9069 from fold_plusminus_mult_expr to catch important cases late when
9070 range info is available.
9072 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9074 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9075 * configure.ac (hidden_linkonce): New test.
9076 * configure: Regenerate.
9077 * config.in: Regenerate.
9079 2018-01-26 Julia Koval <julia.koval@intel.com>
9081 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9082 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9083 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9084 _mm_mask_bitshuffle_epi64_mask): Fix type.
9085 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9086 USI_FTYPE_V4DI_V4DI_USI): Remove.
9087 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9088 __builtin_ia32_vpshufbitqmb256_mask,
9089 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9090 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9091 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9093 2018-01-26 Alan Modra <amodra@gmail.com>
9096 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9097 UNSPEC_VBPERMQ. Sort other unspecs.
9099 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9101 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9103 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9106 * predict.c (drop_profile): Do not push/pop cfun; update also
9108 (handle_missing_profiles): Fix logic looking for zero profiles.
9110 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9113 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9114 on functions with #pragma omp declare simd or functions with simd
9116 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9117 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9118 Remove trailing \n from warning_at calls.
9120 2018-01-25 Tom de Vries <tom@codesourcery.com>
9123 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9124 for neutered workers.
9126 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9129 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9130 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9132 2018-01-24 Jeff Law <law@redhat.com>
9135 * i386.c (get_probe_interval): Move to earlier point.
9136 (ix86_compute_frame_layout): If -fstack-clash-protection and
9137 the frame is larger than the probe interval, then use pushes
9138 to save registers rather than reg->mem moves.
9139 (ix86_expand_prologue): Remove conditional for int_registers_saved
9142 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9145 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9146 min/max for never referenced object.
9148 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9151 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9153 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9154 attributes from DECL_ATTRIBUTES (decl) without affecting
9155 DECL_ATTRIBUTES (current_function_decl).
9156 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9157 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9159 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9161 PR tree-optimization/83979
9162 * fold-const.c (fold_comparison): Use constant_boolean_node
9163 instead of boolean_{true,false}_node.
9165 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9167 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9170 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9172 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9173 Simplify the clause that sets the length attribute.
9174 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9175 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9176 clause that sets the length attribute.
9177 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9179 2018-01-24 Tom de Vries <tom@codesourcery.com>
9182 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9183 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9184 Add strict parameter.
9185 (prevent_branch_around_nothing): Insert dummy insn between branch to
9186 label and label with no ptx insn inbetween.
9187 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9189 2018-01-24 Tom de Vries <tom@codesourcery.com>
9192 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9193 for neutered threads in warp.
9194 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9196 2018-01-24 Richard Biener <rguenther@suse.de>
9198 PR tree-optimization/83176
9199 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9202 2018-01-24 Richard Biener <rguenther@suse.de>
9204 PR tree-optimization/82819
9205 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9206 code generating pluses that are no-ops in the target precision.
9208 2018-01-24 Richard Biener <rguenther@suse.de>
9211 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9213 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9215 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9216 to merge probabilities.
9217 * predict.c (probably_never_executed): Also mark as cold functions
9218 with global 0 profile and guessed local profile.
9219 * profile-count.c (profile_probability::combine_with_count): New
9221 * profile-count.h (profile_probability::operator*,
9222 profile_probability::operator*=, profile_probability::operator/,
9223 profile_probability::operator/=): Reduce precision to adjusted
9224 and set value to guessed on contradictory divisions.
9225 (profile_probability::combine_with_freq): Remove.
9226 (profile_probability::combine_wiht_count): Declare.
9227 (profile_count::force_nonzero):: Set to adjusted.
9228 (profile_count::probability_in):: Set quality to adjusted.
9229 * tree-ssa-tail-merge.c (replace_block_by): Use
9232 2018-01-23 Andrew Waterman <andrew@sifive.com>
9233 Jim Wilson <jimw@sifive.com>
9235 * config/riscv/riscv.c (riscv_stack_boundary): New.
9236 (riscv_option_override): Set riscv_stack_boundary. Handle
9237 riscv_preferred_stack_boundary_arg.
9238 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9239 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9240 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9241 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9242 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9243 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9245 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9248 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9249 of struct ix86_frame.
9250 (ix86_expand_epilogue): Likewise. Add a local variable for
9251 the reg_save_offset field in struct ix86_frame.
9253 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9255 PR tree-optimization/82604
9256 * tree-loop-distribution.c (enum partition_kind): New enum item
9257 PKIND_PARTIAL_MEMSET.
9258 (partition_builtin_p): Support above new enum item.
9259 (generate_code_for_partition): Ditto.
9260 (compute_access_range): Differentiate cases that equality can be
9261 proven at all loops, the innermost loops or no loops.
9262 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9263 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9264 (finalize_partitions, distribute_loop): Don't fuse partition of
9265 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9266 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9269 2018-01-23 Martin Liska <mliska@suse.cz>
9271 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9272 order to ignore the predictor.
9273 (PRED_POLYMORPHIC_CALL): Likewise.
9274 (PRED_RECURSIVE_CALL): Likewise.
9276 2018-01-23 Martin Liska <mliska@suse.cz>
9278 * tree-profile.c (tree_profiling): Print function header to
9279 aware reader which function we are working on.
9280 * value-prof.c (gimple_find_values_to_profile): Do not print
9281 not interesting value histograms.
9283 2018-01-23 Martin Liska <mliska@suse.cz>
9285 * profile-count.h (enum profile_quality): Add
9286 profile_uninitialized as the first value. Do not number values
9287 as they are zero based.
9288 (profile_count::verify): Update sanity check.
9289 (profile_probability::verify): Likewise.
9291 2018-01-23 Nathan Sidwell <nathan@acm.org>
9293 * doc/invoke.texi (ffor-scope): Deprecate.
9295 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9297 PR tree-optimization/83510
9298 * domwalk.c (set_all_edges_as_executable): New function.
9299 (dom_walker::dom_walker): Convert bool param
9300 "skip_unreachable_blocks" to enum reachability. Move setup of
9301 edge flags to set_all_edges_as_executable and only do it when
9302 reachability is REACHABLE_BLOCKS.
9303 * domwalk.h (enum dom_walker::reachability): New enum.
9304 (dom_walker::dom_walker): Convert bool param
9305 "skip_unreachable_blocks" to enum reachability.
9306 (set_all_edges_as_executable): New decl.
9307 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9308 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9310 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9311 but converting true to REACHABLE_BLOCKS.
9312 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9314 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9315 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9316 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9318 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9319 if check_all_array_refs will be called.
9321 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9323 * tree.c (selftest::test_location_wrappers): Add more test
9326 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9328 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9329 (selftest::test_bit_in_range): Likewise.
9331 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9334 * doc/sourcebuild.texi (vect_float): Say that the selector
9335 only describes the situation when -funsafe-math-optimizations is on.
9336 (vect_float_strict): Document.
9338 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9340 PR tree-optimization/83965
9341 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9342 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9343 instead of checking only for a reduction.
9344 (vect_recog_widen_sum_pattern): Likewise.
9346 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9348 * predict.c (probably_never_executed): Only use precise profile info.
9349 (compute_function_frequency): Skip after inlining hack since we now
9350 have quality checking.
9352 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9354 * profile-count.h (profile_probability::very_unlikely,
9355 profile_probability::unlikely, profile_probability::even): Set
9356 precision to guessed.
9358 2018-01-23 Richard Biener <rguenther@suse.de>
9360 PR tree-optimization/83963
9361 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9362 Properly terminate dominator walk when crossing the exit edge not
9363 when visiting its source block.
9365 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9368 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9369 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9371 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9373 PR tree-optimization/83957
9374 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9375 semicolon after for body surrounded by braces.
9377 PR tree-optimization/83081
9378 * profile-count.h (profile_probability::split): New method.
9379 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9380 Use profile_probability::split.
9381 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9382 when splitting a single conditional jump into 2.
9384 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9386 PR tree-optimization/69452
9387 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9390 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9392 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9393 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9394 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9396 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9398 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9399 * config/rl78/rl78.md (movdi): New define_expand.
9400 * config/rl78/rl78.c (rl78_split_movdi): New function.
9402 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9405 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9407 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9408 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9409 128-bit to produce an UNSPEC move to get the double word with the
9410 signbit and then a shift directly to do signbit.
9411 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9412 implementation with a new version that just does either a direct
9413 move or a regular move. Move memory interface to separate insns.
9414 Move insns so they are next to the expander.
9415 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9416 with signbit move. Split big and little endian case.
9417 (signbit<mode>2_dm_mem_le): Likewise.
9418 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9419 (signbit<mode>2_dm2): Likewise.
9421 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9423 * config/rl78/rl78.md (anddi3): New define_expand.
9425 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9427 * config/rl78/rl78.md (umindi3): New define_expand.
9429 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9431 * config/rl78/rl78.md (smindi3): New define_expand.
9433 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9435 * config/rl78/rl78.md (smaxdi3): New define_expand.
9437 2018-01-22 Carl Love <cel@us.ibm.com>
9439 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9440 LVX_V1TI): Add macro expansion.
9441 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9442 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9443 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9444 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9445 Change check to determine if the instruction is a byte reversing
9446 entry. Fix typo in comment.
9447 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9448 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9449 Add def_builtin calls for new builtins.
9450 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9451 Add define_insn expansion.
9453 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9455 * config/rl78/rl78.md (umaxdi3): New define_expand.
9457 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9459 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9460 for non-QImode registers.
9462 2018-01-22 Richard Biener <rguenther@suse.de>
9464 PR tree-optimization/83963
9465 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9466 including the loop exit block.
9467 (scop_detection::merge_sese): Likewise.
9468 (scop_detection::add_scop): Do it here instead.
9470 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9472 * doc/sourcebuild.texi (arm_softfloat): Document.
9474 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9477 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9478 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9479 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9481 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9482 David Edelsohn <dje.gcc@gmail.com>
9485 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9486 Change "crset eq" to "crset 2".
9487 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9488 (*call_indirect_aix<mode>_nospec): Likewise.
9489 (*call_value_indirect_aix<mode>_nospec): Likewise.
9490 (*call_indirect_elfv2<mode>_nospec): Likewise.
9491 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9492 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9493 change assembly output from . to $.
9494 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9495 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9496 (*tablejump<mode>_internal1_nospec): Likewise.
9498 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9501 * config/sh/sh_optimize_sett_clrt.cc:
9502 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9504 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9506 PR tree-optimization/83940
9507 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9508 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9509 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9510 use it to pass back the definition type.
9511 (vect_check_store_rhs): Likewise rhs_dt_out.
9512 (vect_build_gather_load_calls): Add a mask_dt argument and use
9513 it instead of a call to vect_is_simple_use.
9514 (vectorizable_store): Update calls to vect_check_load_store_mask
9515 and vect_check_store_rhs. Use the dt returned by the latter instead
9516 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9517 instead of calls to vect_is_simple_use. Pass the scalar rather
9518 than the vector operand to vect_is_simple_use when handling
9519 second and subsequent copies of an rhs value.
9520 (vectorizable_load): Update calls to vect_check_load_store_mask
9521 and vect_build_gather_load_calls. Use the cached mask_dt and
9522 gs_info.offset_dt instead of calls to vect_is_simple_use.
9524 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9527 * tree-emutls.c: Include gimplify.h.
9528 (lower_emutls_2): New function.
9529 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9530 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9531 it before further processing.
9534 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9535 UINTVAL (trueop1) instead of INTVAL (op1).
9537 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9541 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9542 INCOMING_FRAME_SP_OFFSET if not defined.
9543 (scan_trace): Add ENTRY argument. If true and
9544 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9545 emit a note to adjust the CFA offset.
9546 (create_cfi_notes): Adjust scan_trace callers.
9547 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9548 INCOMING_FRAME_SP_OFFSET in the CIE.
9549 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9550 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9552 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9553 * doc/tm.texi: Regenerated.
9555 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9557 PR rtl-optimization/83147
9558 * lra-constraints.c (remove_inheritance_pseudos): Use
9559 lra_substitute_pseudo_within_insn.
9561 2018-01-19 Tom de Vries <tom@codesourcery.com>
9562 Cesar Philippidis <cesar@codesourcery.com>
9565 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9567 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9570 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9571 spaces for function labels.
9573 2018-01-19 Martin Liska <mliska@suse.cz>
9575 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9576 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9577 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9578 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9579 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9580 (PRED_CONST_RETURN): Change from 69 to 65.
9581 (PRED_NULL_RETURN): Change from 91 to 71.
9582 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9583 (PRED_LOOP_GUARD): Change from 66 to 73.
9585 2018-01-19 Martin Liska <mliska@suse.cz>
9587 * predict.c (predict_insn_def): Add new assert.
9588 (struct branch_predictor): Change type to signed integer.
9589 (test_prediction_value_range): Amend test to cover
9591 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9592 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9593 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9594 (PRED_LOOP_IV_COMPARE): Likewise.
9595 * predict.h (PROB_UNINITIALIZED): Define new constant.
9597 2018-01-19 Martin Liska <mliska@suse.cz>
9599 * predict.c (dump_prediction): Add new format for
9600 analyze_brprob.py script which is enabled with -details
9602 * profile-count.h (precise_p): New function.
9604 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9606 PR tree-optimization/83922
9607 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9608 there are no statements that need masking.
9609 (vect_active_double_reduction_p): New function.
9610 (vect_analyze_loop_operations): Use it when handling phis that
9611 are not in the loop header.
9613 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9615 PR tree-optimization/83914
9616 * tree-vect-loop.c (vectorizable_induction): Don't convert
9617 init_expr or apply the peeling adjustment for inductions
9618 that are nested within the vectorized loop.
9620 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9622 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9625 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9629 * function.h (gimplify_parameters): Add gimple_seq * argument.
9630 * function.c: Include gimple.h and options.h.
9631 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9632 for the added local temporaries if needed.
9633 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9634 if there are any parameter cleanups, wrap whole body into a
9635 try/finally with the cleanups.
9637 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9640 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9641 Use GET_MODE_CLASS for scalar floating point.
9643 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9647 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9648 Fix call of call_cgraph_insertion_hooks.
9650 2018-01-18 Martin Sebor <msebor@redhat.com>
9652 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9654 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9657 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9660 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9663 * common.opt: (-ffile-prefix-map): New option.
9664 * opts.c (common_handle_option): Defer it.
9665 * opts-global.c (handle_common_deferred_options): Handle it.
9666 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9667 * file-prefix-map.h: New file.
9668 (remap_debug_filename, add_debug_prefix_map): ...here.
9669 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9670 * final.c (debug_prefix_map, add_debug_prefix_map
9671 remap_debug_filename): Move to...
9672 * file-prefix-map.c: New file.
9673 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9674 generalize, get rid of alloca(), use strrchr() instead of strchr().
9675 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9676 Implement in terms of add_prefix_map().
9677 (remap_macro_filename, remap_debug_filename): Implement in term of
9679 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9680 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9681 * dbxout.c: Include file-prefix-map.h.
9682 * varasm.c: Likewise.
9683 * vmsdbgout.c: Likewise.
9684 * xcoffout.c: Likewise.
9685 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9686 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9687 * doc/invoke.texi (-ffile-prefix-map): Document.
9688 (-fdebug-prefix-map): Update description.
9690 2018-01-18 Martin Liska <mliska@suse.cz>
9692 * config/i386/i386.c (indirect_thunk_name): Document that also
9694 (output_indirect_thunk): Document why both instructions
9695 (pause and lfence) are generated.
9697 2018-01-18 Richard Biener <rguenther@suse.de>
9699 PR tree-optimization/83887
9700 * graphite-scop-detection.c
9701 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9702 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9703 (scop_detection::merge_sese): Re-implement with a flood-fill
9704 algorithm that properly finds a SESE region if it exists.
9706 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9709 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9710 pointer_diff optimizations use view_convert instead of convert.
9712 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9714 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9715 Generate different code for -mno-speculate-indirect-jumps.
9716 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9717 (*call_indirect_aix<mode>): Disable for
9718 -mno-speculate-indirect-jumps.
9719 (*call_indirect_aix<mode>_nospec): New define_insn.
9720 (*call_value_indirect_aix<mode>): Disable for
9721 -mno-speculate-indirect-jumps.
9722 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9723 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9724 -mno-speculate-indirect-jumps.
9725 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9727 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9729 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9730 long double type, set the flags for noting the default long double
9731 type, even if we don't pass or return a long double type.
9733 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9736 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9739 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9741 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9742 support for merge[hl].
9743 (fold_mergehl_helper): New helper function.
9744 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9745 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9746 (altivec_vmrglw_direct): Add xxmrglw insn.
9748 2018-01-17 Andrew Waterman <andrew@sifive.com>
9750 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9751 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9753 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9756 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9757 call the lto_location_cache before reading the
9758 DECL_SOURCE_LOCATION of the types.
9760 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9761 Richard Sandiford <richard.sandiford@linaro.org>
9763 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9764 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9765 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9766 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9767 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9769 * config/aarch64/constraints.md (aarch64_movti_operand):
9771 * config/aarch64/predicates.md (Uti): Add new constraint.
9773 2018-01-17 Carl Love <cel@us.ibm.com>
9775 * config/rs6000/vsx.md (define_expand xl_len_r,
9776 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9777 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9779 (define_expand, define_insn): Move the shift left from the
9780 define_insn to the define_expand for lxvl and stxvl instructions.
9781 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9782 and XL_LEN_R definitions to PURE.
9784 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9786 * config/i386/i386.c (indirect_thunk_name): Declare regno
9787 as unsigned int. Compare regno with INVALID_REGNUM.
9788 (output_indirect_thunk): Ditto.
9789 (output_indirect_thunk_function): Ditto.
9790 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9791 in the call to output_indirect_thunk_function.
9793 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9796 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9797 rather than the size of inner_type to determine the stack slot size
9798 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9800 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9803 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9806 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9808 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9809 endian Linux systems to optionally enable multilibs for selecting
9810 the long double type if the user configured an explicit type.
9811 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9812 have no long double multilibs if not defined.
9813 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9814 warn if the user used -mabi={ieee,ibm}longdouble and we built
9815 multilibs for long double.
9816 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9817 appropriate multilib option.
9818 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9820 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9821 for building long double multilibs.
9822 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9824 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9826 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9829 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9831 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9834 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9837 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9840 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9842 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9843 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9845 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9847 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9848 different rtl trees depending on TARGET_64BIT.
9849 (rs6000_gen_lvx): Likewise.
9851 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9853 * config/visium/visium.md (nop): Tweak comment.
9854 (hazard_nop): Likewise.
9856 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9858 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9859 -mspeculate-indirect-jumps.
9860 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9861 for -mno-speculate-indirect-jumps.
9862 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9863 (*call_value_indirect_elfv2<mode>): Disable for
9864 -mno-speculate-indirect-jumps.
9865 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9866 (indirect_jump): Emit different RTL for
9867 -mno-speculate-indirect-jumps.
9868 (*indirect_jump<mode>): Disable for
9869 -mno-speculate-indirect-jumps.
9870 (*indirect_jump<mode>_nospec): New define_insn.
9871 (tablejump): Emit different RTL for
9872 -mno-speculate-indirect-jumps.
9873 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9874 (tablejumpsi_nospec): New define_expand.
9875 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9876 (tablejumpdi_nospec): New define_expand.
9877 (*tablejump<mode>_internal1): Disable for
9878 -mno-speculate-indirect-jumps.
9879 (*tablejump<mode>_internal1_nospec): New define_insn.
9880 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9883 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9885 * caller-save.c (insert_save): Drop unnecessary parameter. All
9888 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9889 Richard Biener <rguenth@suse.de>
9892 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9893 return early, inline manually is_gimple_sizepos. Make sure if we
9894 call gimplify_expr we don't end up with a gimple constant.
9895 * tree.c (variably_modified_type_p): Don't return true for
9896 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9897 * gimplify.h (is_gimple_sizepos): Remove.
9899 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9901 PR tree-optimization/83857
9902 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9903 vectorizable_live_operation for pure SLP statements.
9904 (vectorizable_live_operation): Handle PHIs.
9906 2018-01-16 Richard Biener <rguenther@suse.de>
9908 PR tree-optimization/83867
9909 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9910 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9912 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9915 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9916 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9917 If off is not INTEGER_CST, issue a may not be aligned warning
9918 rather than isn't aligned. Use isn%'t rather than isn't.
9919 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9921 <case MULT_EXPR>: Improve the case when bottom and one of the
9922 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9923 operand, in that case check if the other operand is multiple of
9924 bottom divided by the INTEGER_CST operand.
9926 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9929 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9930 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9931 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9932 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9933 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9934 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9935 (pa_function_arg_size): New function.
9937 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9939 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9940 in a separate statement.
9942 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9944 PR tree-optimization/83847
9945 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9946 group gathers and scatters.
9948 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9950 PR rtl-optimization/86620
9951 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9953 PR rtl-optimization/83213
9954 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9955 to last if both are JUMP_INSNs.
9957 PR tree-optimization/83843
9958 * gimple-ssa-store-merging.c
9959 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9960 store_immediate_info for bswap/nop orig_stores.
9962 2018-01-15 Andrew Waterman <andrew@sifive.com>
9964 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9966 <UDIV>: Increase cost if !TARGET_DIV.
9968 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9970 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9971 (define_attr "cr_logical_3op"): New.
9972 (cceq_ior_compare): Adjust.
9973 (cceq_ior_compare_complement): Adjust.
9974 (*cceq_rev_compare): Adjust.
9975 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9976 (is_cracked_insn): Adjust.
9977 (insn_must_be_first_in_group): Adjust.
9978 * config/rs6000/40x.md: Adjust.
9979 * config/rs6000/440.md: Adjust.
9980 * config/rs6000/476.md: Adjust.
9981 * config/rs6000/601.md: Adjust.
9982 * config/rs6000/603.md: Adjust.
9983 * config/rs6000/6xx.md: Adjust.
9984 * config/rs6000/7450.md: Adjust.
9985 * config/rs6000/7xx.md: Adjust.
9986 * config/rs6000/8540.md: Adjust.
9987 * config/rs6000/cell.md: Adjust.
9988 * config/rs6000/e300c2c3.md: Adjust.
9989 * config/rs6000/e500mc.md: Adjust.
9990 * config/rs6000/e500mc64.md: Adjust.
9991 * config/rs6000/e5500.md: Adjust.
9992 * config/rs6000/e6500.md: Adjust.
9993 * config/rs6000/mpc.md: Adjust.
9994 * config/rs6000/power4.md: Adjust.
9995 * config/rs6000/power5.md: Adjust.
9996 * config/rs6000/power6.md: Adjust.
9997 * config/rs6000/power7.md: Adjust.
9998 * config/rs6000/power8.md: Adjust.
9999 * config/rs6000/power9.md: Adjust.
10000 * config/rs6000/rs64.md: Adjust.
10001 * config/rs6000/titan.md: Adjust.
10003 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10005 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10006 ix86_indirect_branch_register logic.
10008 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10010 * config/i386/constraints.md (Bs): Update
10011 ix86_indirect_branch_register check. Don't check
10012 ix86_indirect_branch_register with GOT_memory_operand.
10014 * config/i386/predicates.md (GOT_memory_operand): Don't check
10015 ix86_indirect_branch_register here.
10016 (GOT32_symbol_operand): Likewise.
10018 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10020 * config/i386/predicates.md (constant_call_address_operand):
10021 Rewrite ix86_indirect_branch_register logic.
10022 (sibcall_insn_operand): Likewise.
10024 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10026 * config/i386/constraints.md (Bs): Replace
10027 ix86_indirect_branch_thunk_register with
10028 ix86_indirect_branch_register.
10030 * config/i386/i386.md (indirect_jump): Likewise.
10031 (tablejump): Likewise.
10032 (*sibcall_memory): Likewise.
10033 (*sibcall_value_memory): Likewise.
10034 Peepholes of indirect call and jump via memory: Likewise.
10035 * config/i386/i386.opt: Likewise.
10036 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10037 (GOT_memory_operand): Likewise.
10038 (call_insn_operand): Likewise.
10039 (sibcall_insn_operand): Likewise.
10040 (GOT32_symbol_operand): Likewise.
10042 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10044 PR middle-end/83837
10045 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10046 type rather than type addr's type points to.
10047 (expand_omp_atomic_mutex): Likewise.
10048 (expand_omp_atomic): Likewise.
10050 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10053 * config/i386/i386.c (output_indirect_thunk_function): Use
10054 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10055 for __x86_return_thunk.
10057 2018-01-15 Richard Biener <rguenther@suse.de>
10059 PR middle-end/83850
10060 * expmed.c (extract_bit_field_1): Fix typo.
10062 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10065 * config/arm/iterators.md (VF): New mode iterator.
10066 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10067 Remove integer-related logic from pattern.
10068 (neon_vabd<mode>_3): Likewise.
10070 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10072 PR middle-end/82694
10073 * common.opt (fstrict-overflow): No longer an alias.
10074 (fwrapv-pointer): New option.
10075 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10076 also for pointer types based on flag_wrapv_pointer.
10077 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10078 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10079 opts->x_flag_wrapv got set.
10080 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10081 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10082 POINTER_TYPE_OVERFLOW_UNDEFINED.
10083 * match.pd: Likewise in address comparison pattern.
10084 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10086 2018-01-15 Richard Biener <rguenther@suse.de>
10089 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10090 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10091 Reset type names to their identifier if their TYPE_DECL doesn't
10092 have linkage (and thus is used for ODR and devirt).
10093 (save_debug_info_for_decl): Remove.
10094 (save_debug_info_for_type): Likewise.
10095 (add_tree_to_fld_list): Adjust.
10096 * tree-pretty-print.c (dump_generic_node): Make dumping of
10097 type names more robust.
10099 2018-01-15 Richard Biener <rguenther@suse.de>
10101 * BASE-VER: Bump to 8.0.1.
10103 2018-01-14 Martin Sebor <msebor@redhat.com>
10106 * builtins.c (check_access): Avoid warning when the no-warning bit
10109 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10111 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10112 * ira-color (allocno_hard_regs_compare): Likewise.
10114 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10117 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10118 Use .pushsection/.popsection.
10120 2018-01-14 Martin Sebor <msebor@redhat.com>
10123 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10125 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10127 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10128 entry from extra_headers.
10129 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10130 extra_headers, make the list bitwise identical to the i?86-*-* one.
10132 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10134 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10135 -mcmodel=large with -mindirect-branch=thunk,
10136 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10137 -mfunction-return=thunk-extern.
10138 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10139 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10140 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10142 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10144 * config/i386/i386.c (print_reg): Print the name of the full
10145 integer register without '%'.
10146 (ix86_print_operand): Handle 'V'.
10147 * doc/extend.texi: Document 'V' modifier.
10149 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10151 * config/i386/constraints.md (Bs): Disallow memory operand for
10152 -mindirect-branch-register.
10154 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10155 (GOT_memory_operand): Likewise.
10156 (call_insn_operand): Likewise.
10157 (sibcall_insn_operand): Likewise.
10158 (GOT32_symbol_operand): Likewise.
10159 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10160 for -mindirect-branch-register.
10161 (tablejump): Likewise.
10162 (*sibcall_memory): Likewise.
10163 (*sibcall_value_memory): Likewise.
10164 Disallow peepholes of indirect call and jump via memory for
10165 -mindirect-branch-register.
10166 (*call_pop): Replace m with Bw.
10167 (*call_value_pop): Likewise.
10168 (*sibcall_pop_memory): Replace m with Bs.
10169 * config/i386/i386.opt (mindirect-branch-register): New option.
10170 * doc/invoke.texi: Document -mindirect-branch-register option.
10172 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10174 * config/i386/i386-protos.h (ix86_output_function_return): New.
10175 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10176 set function_return_type.
10177 (indirect_thunk_name): Add ret_p to indicate thunk for function
10179 (output_indirect_thunk_function): Pass false to
10180 indirect_thunk_name.
10181 (ix86_output_indirect_branch_via_reg): Likewise.
10182 (ix86_output_indirect_branch_via_push): Likewise.
10183 (output_indirect_thunk_function): Create alias for function
10184 return thunk if regno < 0.
10185 (ix86_output_function_return): New function.
10186 (ix86_handle_fndecl_attribute): Handle function_return.
10187 (ix86_attribute_table): Add function_return.
10188 * config/i386/i386.h (machine_function): Add
10189 function_return_type.
10190 * config/i386/i386.md (simple_return_internal): Use
10191 ix86_output_function_return.
10192 (simple_return_internal_long): Likewise.
10193 * config/i386/i386.opt (mfunction-return=): New option.
10194 (indirect_branch): Mention -mfunction-return=.
10195 * doc/extend.texi: Document function_return function attribute.
10196 * doc/invoke.texi: Document -mfunction-return= option.
10198 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10200 * config/i386/i386-opts.h (indirect_branch): New.
10201 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10202 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10203 with local indirect jump when converting indirect call and jump.
10204 (ix86_set_indirect_branch_type): New.
10205 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10206 (indirectlabelno): New.
10207 (indirect_thunk_needed): Likewise.
10208 (indirect_thunk_bnd_needed): Likewise.
10209 (indirect_thunks_used): Likewise.
10210 (indirect_thunks_bnd_used): Likewise.
10211 (INDIRECT_LABEL): Likewise.
10212 (indirect_thunk_name): Likewise.
10213 (output_indirect_thunk): Likewise.
10214 (output_indirect_thunk_function): Likewise.
10215 (ix86_output_indirect_branch_via_reg): Likewise.
10216 (ix86_output_indirect_branch_via_push): Likewise.
10217 (ix86_output_indirect_branch): Likewise.
10218 (ix86_output_indirect_jmp): Likewise.
10219 (ix86_code_end): Call output_indirect_thunk_function if needed.
10220 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10222 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10223 (ix86_attribute_table): Add indirect_branch.
10224 * config/i386/i386.h (machine_function): Add indirect_branch_type
10225 and has_local_indirect_jump.
10226 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10228 (tablejump): Likewise.
10229 (*indirect_jump): Use ix86_output_indirect_jmp.
10230 (*tablejump_1): Likewise.
10231 (simple_return_indirect_internal): Likewise.
10232 * config/i386/i386.opt (mindirect-branch=): New option.
10233 (indirect_branch): New.
10236 (thunk-inline): Likewise.
10237 (thunk-extern): Likewise.
10238 * doc/extend.texi: Document indirect_branch function attribute.
10239 * doc/invoke.texi: Document -mindirect-branch= option.
10241 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10244 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10246 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10248 * ipa-inline.c (want_inline_small_function_p): Return false if
10249 inlining has already failed with CIF_FINAL_ERROR.
10250 (update_caller_keys): Call want_inline_small_function_p before
10252 (update_callee_keys): Likewise.
10254 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10256 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10258 (rs6000_quadword_masked_address_p): Likewise.
10259 (quad_aligned_load_p): Likewise.
10260 (quad_aligned_store_p): Likewise.
10261 (const_load_sequence_p): Add comment to describe the outer-most loop.
10262 (mimic_memory_attributes_and_flags): New function.
10263 (rs6000_gen_stvx): Likewise.
10264 (replace_swapped_aligned_store): Likewise.
10265 (rs6000_gen_lvx): Likewise.
10266 (replace_swapped_aligned_load): Likewise.
10267 (replace_swapped_load_constant): Capitalize argument name in
10268 comment describing this function.
10269 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10270 and stores that access quad-word aligned addresses and replace
10271 with stvx or lvx instructions when appropriate.
10272 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10273 New function prototype.
10274 (rs6000_quadword_masked_address_p): Likewise.
10275 (rs6000_gen_lvx): Likewise.
10276 (rs6000_gen_stvx): Likewise.
10277 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10278 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10279 when memory address is aligned.
10280 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10281 this split to select lvx instruction when memory address is aligned.
10282 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10283 instruction when memory address is aligned.
10284 (*vsx_le_perm_load_v16qi): Likewise.
10285 (four unnamed splitters): Modify to select the stvx instruction
10286 when memory is aligned.
10288 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10290 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10291 which appears in the queue multiple times.
10293 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10294 Alan Hayward <alan.hayward@arm.com>
10295 David Sherwood <david.sherwood@arm.com>
10297 * tree-vectorizer.h (vec_lower_bound): New structure.
10298 (_loop_vec_info): Add check_nonzero and lower_bounds.
10299 (LOOP_VINFO_CHECK_NONZERO): New macro.
10300 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10301 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10302 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10303 fields. Make seg_len the distance travelled, not including the
10305 (dr_direction_indicator): Declare.
10306 (dr_zero_step_indicator): Likewise.
10307 (dr_known_forward_stride_p): Likewise.
10308 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10310 (runtime_alias_check_p): Allow runtime alias checks with
10312 (operator ==): Compare access_size and align.
10313 (prune_runtime_alias_test_list): Rework for new distinction between
10314 the access_size and seg_len.
10315 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10317 (get_segment_min_max): New function.
10318 (create_intersect_range_checks): Use it.
10319 (dr_step_indicator): New function.
10320 (dr_direction_indicator): Likewise.
10321 (dr_zero_step_indicator): Likewise.
10322 (dr_known_forward_stride_p): Likewise.
10323 * tree-loop-distribution.c (data_ref_segment_size): Return
10324 DR_STEP * (niters - 1).
10325 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10327 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10328 (vect_preserves_scalar_order_p): New function, split out from...
10329 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10330 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10331 (vect_vfa_access_size): New function.
10332 (vect_vfa_align): Likewise.
10333 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10334 (dump_lower_bound): New function.
10335 (vect_check_lower_bound): Likewise.
10336 (vect_small_gap_p): Likewise.
10337 (vectorizable_with_step_bound_p): Likewise.
10338 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10339 depencies if the vectorization factor is 1. Convert the checks
10340 for nonzero steps into checks on the bounds of DR_STEP. Try using
10341 a bunds check for variable steps if the minimum required step is
10342 relatively small. Update calls to the dr_with_seg_len
10343 constructor and to vect_compile_time_alias.
10344 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10346 (vect_loop_versioning): Call it.
10347 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10349 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10351 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10352 Alan Hayward <alan.hayward@arm.com>
10353 David Sherwood <david.sherwood@arm.com>
10355 * doc/sourcebuild.texi (vect_scatter_store): Document.
10356 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10358 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10360 * genopinit.c (main): Add supports_vec_scatter_store and
10361 supports_vec_scatter_store_cached to target_optabs.
10362 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10363 IFN_MASK_SCATTER_STORE.
10364 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10366 * internal-fn.h (internal_store_fn_p): Declare.
10367 (internal_fn_stored_value_index): Likewise.
10368 * internal-fn.c (scatter_store_direct): New macro.
10369 (expand_scatter_store_optab_fn): New function.
10370 (direct_scatter_store_optab_supported_p): New macro.
10371 (internal_store_fn_p): New function.
10372 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10373 IFN_MASK_SCATTER_STORE.
10374 (internal_fn_mask_index): Likewise.
10375 (internal_fn_stored_value_index): New function.
10376 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10377 for scatter stores.
10378 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10379 * optabs-query.c (supports_vec_scatter_store_p): New function.
10380 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10381 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10382 true for scatter stores.
10383 (vect_gather_scatter_fn_p): Handle scatter stores too.
10384 (vect_check_gather_scatter): Consider using scatter stores if
10385 supports_vec_scatter_store_p.
10386 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10387 scatter stores too.
10388 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10389 internal_fn_stored_value_index.
10390 (check_load_store_masking): Handle scatter stores too.
10391 (vect_get_store_rhs): Make public.
10392 (vectorizable_call): Use internal_store_fn_p.
10393 (vectorizable_store): Handle scatter store internal functions.
10394 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10395 when deciding whether the end of the group has been reached.
10396 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10397 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10398 (mask_scatter_store<mode>): New insns.
10400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10401 Alan Hayward <alan.hayward@arm.com>
10402 David Sherwood <david.sherwood@arm.com>
10404 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10405 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10406 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10408 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10409 Use vect_truncate_gather_scatter_offset if we can't treat the
10410 operation as a normal gather load or scatter store.
10411 (get_group_load_store_type): Take the gather_scatter_info
10412 as argument. Try using a gather load or scatter store for
10413 single-element groups.
10414 (get_load_store_type): Update calls to get_group_load_store_type
10415 and vect_use_strided_gather_scatters_p.
10417 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10418 Alan Hayward <alan.hayward@arm.com>
10419 David Sherwood <david.sherwood@arm.com>
10421 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10422 optional tree argument.
10423 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10425 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10426 but continue to use the current value as a fallback.
10427 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10428 to compare the updates.
10429 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10430 (get_load_store_type): Use it when handling a strided access.
10431 (vect_get_strided_load_store_ops): New function.
10432 (vect_get_data_ptr_increment): Likewise.
10433 (vectorizable_load): Handle strided gather loads. Always pass
10434 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10436 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10437 Alan Hayward <alan.hayward@arm.com>
10438 David Sherwood <david.sherwood@arm.com>
10440 * doc/md.texi (gather_load@var{m}): Document.
10441 (mask_gather_load@var{m}): Likewise.
10442 * genopinit.c (main): Add supports_vec_gather_load and
10443 supports_vec_gather_load_cached to target_optabs.
10444 * optabs-tree.c (init_tree_optimization_optabs): Use
10445 ggc_cleared_alloc to allocate target_optabs.
10446 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10447 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10449 * internal-fn.h (internal_load_fn_p): Declare.
10450 (internal_gather_scatter_fn_p): Likewise.
10451 (internal_fn_mask_index): Likewise.
10452 (internal_gather_scatter_fn_supported_p): Likewise.
10453 * internal-fn.c (gather_load_direct): New macro.
10454 (expand_gather_load_optab_fn): New function.
10455 (direct_gather_load_optab_supported_p): New macro.
10456 (direct_internal_fn_optab): New function.
10457 (internal_load_fn_p): Likewise.
10458 (internal_gather_scatter_fn_p): Likewise.
10459 (internal_fn_mask_index): Likewise.
10460 (internal_gather_scatter_fn_supported_p): Likewise.
10461 * optabs-query.c (supports_at_least_one_mode_p): New function.
10462 (supports_vec_gather_load_p): Likewise.
10463 * optabs-query.h (supports_vec_gather_load_p): Declare.
10464 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10465 and memory_type field.
10466 (NUM_PATTERNS): Bump to 15.
10467 * tree-vect-data-refs.c: Include internal-fn.h.
10468 (vect_gather_scatter_fn_p): New function.
10469 (vect_describe_gather_scatter_call): Likewise.
10470 (vect_check_gather_scatter): Try using internal functions for
10471 gather loads. Recognize existing calls to a gather load function.
10472 (vect_analyze_data_refs): Consider using gather loads if
10473 supports_vec_gather_load_p.
10474 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10475 (vect_get_gather_scatter_offset_type): Likewise.
10476 (vect_convert_mask_for_vectype): Likewise.
10477 (vect_add_conversion_to_patterm): Likewise.
10478 (vect_try_gather_scatter_pattern): Likewise.
10479 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10480 (vect_vect_recog_func_ptrs): Add it.
10481 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10482 internal_fn_mask_index and internal_gather_scatter_fn_p.
10483 (check_load_store_masking): Take the gather_scatter_info as an
10484 argument and handle gather loads.
10485 (vect_get_gather_scatter_ops): New function.
10486 (vectorizable_call): Check internal_load_fn_p.
10487 (vectorizable_load): Likewise. Handle gather load internal
10489 (vectorizable_store): Update call to check_load_store_masking.
10490 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10491 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10492 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10493 (aarch64_gather_scale_operand_d): New predicates.
10494 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10495 (mask_gather_load<mode>): New insns.
10497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10498 Alan Hayward <alan.hayward@arm.com>
10499 David Sherwood <david.sherwood@arm.com>
10501 * optabs.def (fold_left_plus_optab): New optab.
10502 * doc/md.texi (fold_left_plus_@var{m}): Document.
10503 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10504 * internal-fn.c (fold_left_direct): Define.
10505 (expand_fold_left_optab_fn): Likewise.
10506 (direct_fold_left_optab_supported_p): Likewise.
10507 * fold-const-call.c (fold_const_fold_left): New function.
10508 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10509 * tree-parloops.c (valid_reduction_p): New function.
10510 (gather_scalar_reductions): Use it.
10511 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10512 (vect_finish_replace_stmt): Declare.
10513 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10514 (needs_fold_left_reduction_p): New function, split out from...
10515 (vect_is_simple_reduction): ...here. Accept reductions that
10516 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10517 (vect_force_simple_reduction): Also store the reduction type in
10518 the assignment's STMT_VINFO_REDUC_TYPE.
10519 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10520 (merge_with_identity): New function.
10521 (vect_expand_fold_left): Likewise.
10522 (vectorize_fold_left_reduction): Likewise.
10523 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10524 scalar phi in place for it. Check for target support and reject
10525 cases that would reassociate the operation. Defer the transform
10526 phase to vectorize_fold_left_reduction.
10527 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10528 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10529 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10531 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10533 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10534 call to ifc_temp_var.
10536 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10537 Alan Hayward <alan.hayward@arm.com>
10538 David Sherwood <david.sherwood@arm.com>
10540 * target.def (legitimize_address_displacement): Take the original
10541 offset as a poly_int.
10542 * targhooks.h (default_legitimize_address_displacement): Update
10544 * targhooks.c (default_legitimize_address_displacement): Likewise.
10545 * doc/tm.texi: Regenerate.
10546 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10547 as an argument, moving assert of ad->disp == ad->disp_term to...
10548 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10549 Try calling targetm.legitimize_address_displacement before expanding
10550 the address rather than afterwards, and adjust for the new interface.
10551 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10552 Match the new hook interface. Handle SVE addresses.
10553 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10554 new hook interface.
10556 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10558 * Makefile.in (OBJS): Add early-remat.o.
10559 * target.def (select_early_remat_modes): New hook.
10560 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10561 * doc/tm.texi: Regenerate.
10562 * targhooks.h (default_select_early_remat_modes): Declare.
10563 * targhooks.c (default_select_early_remat_modes): New function.
10564 * timevar.def (TV_EARLY_REMAT): New timevar.
10565 * passes.def (pass_early_remat): New pass.
10566 * tree-pass.h (make_pass_early_remat): Declare.
10567 * early-remat.c: New file.
10568 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10570 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10573 Alan Hayward <alan.hayward@arm.com>
10574 David Sherwood <david.sherwood@arm.com>
10576 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10577 vfm1 with a bound_epilog parameter.
10578 (vect_do_peeling): Update calls accordingly, and move the prologue
10579 call earlier in the function. Treat the base bound_epilog as 0 for
10580 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10581 this base when peeling for gaps.
10582 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10583 with fully-masked loops.
10584 (vect_estimate_min_profitable_iters): Handle the single peeled
10585 iteration in that case.
10587 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10588 Alan Hayward <alan.hayward@arm.com>
10589 David Sherwood <david.sherwood@arm.com>
10591 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10592 single-element interleaving even if the size is not a power of 2.
10593 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10594 accesses for single-element interleaving if the group size is
10597 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10598 Alan Hayward <alan.hayward@arm.com>
10599 David Sherwood <david.sherwood@arm.com>
10601 * doc/md.texi (fold_extract_last_@var{m}): Document.
10602 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10603 * optabs.def (fold_extract_last_optab): New optab.
10604 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10605 * internal-fn.c (fold_extract_direct): New macro.
10606 (expand_fold_extract_optab_fn): Likewise.
10607 (direct_fold_extract_optab_supported_p): Likewise.
10608 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10609 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10610 EXTRACT_LAST_REDUCTION.
10611 (get_initial_def_for_reduction): Do not create an initial vector
10612 for EXTRACT_LAST_REDUCTION reductions.
10613 (vectorizable_reduction): Leave the scalar phi in place for
10614 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10615 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10616 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10617 transform phase to vectorizable_condition.
10618 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10620 (vect_finish_stmt_generation): ...here.
10621 (vect_finish_replace_stmt): New function.
10622 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10623 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10625 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10627 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10628 Alan Hayward <alan.hayward@arm.com>
10629 David Sherwood <david.sherwood@arm.com>
10631 * doc/md.texi (extract_last_@var{m}): Document.
10632 * optabs.def (extract_last_optab): New optab.
10633 * internal-fn.def (EXTRACT_LAST): New internal function.
10634 * internal-fn.c (cond_unary_direct): New macro.
10635 (expand_cond_unary_optab_fn): Likewise.
10636 (direct_cond_unary_optab_supported_p): Likewise.
10637 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10638 loops using EXTRACT_LAST.
10639 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10640 (extract_last_<mode>): ...this optab.
10641 (vec_extract<mode><Vel>): Update accordingly.
10643 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10644 Alan Hayward <alan.hayward@arm.com>
10645 David Sherwood <david.sherwood@arm.com>
10647 * target.def (empty_mask_is_expensive): New hook.
10648 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10649 * doc/tm.texi: Regenerate.
10650 * targhooks.h (default_empty_mask_is_expensive): Declare.
10651 * targhooks.c (default_empty_mask_is_expensive): New function.
10652 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10653 if the target says that empty masks are expensive.
10654 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10656 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10659 Alan Hayward <alan.hayward@arm.com>
10660 David Sherwood <david.sherwood@arm.com>
10662 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10663 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10664 (vect_use_loop_mask_for_alignment_p): New function.
10665 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10666 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10667 niters_skip argument. Make sure that the first niters_skip elements
10668 of the first iteration are inactive.
10669 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10670 Update call to vect_set_loop_masks_directly.
10671 (get_misalign_in_elems): New function, split out from...
10672 (vect_gen_prolog_loop_niters): ...here.
10673 (vect_update_init_of_dr): Take a code argument that specifies whether
10674 the adjustment should be added or subtracted.
10675 (vect_update_init_of_drs): Likewise.
10676 (vect_prepare_for_masked_peels): New function.
10677 (vect_do_peeling): Skip prologue peeling if we're using a mask
10678 instead. Update call to vect_update_inits_of_drs.
10679 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10681 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10682 alignment. Do not include the number of peeled iterations in
10683 the minimum threshold in that case.
10684 (vectorizable_induction): Adjust the start value down by
10685 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10686 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10687 Take the number of skipped iterations into account when calculating
10689 * tree-vect-stmts.c (vect_gen_while_not): New function.
10691 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10692 Alan Hayward <alan.hayward@arm.com>
10693 David Sherwood <david.sherwood@arm.com>
10695 * doc/sourcebuild.texi (vect_fully_masked): Document.
10696 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10697 default value to 0.
10698 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10700 (vect_analyze_loop_2): ...here. Don't check the vectorization
10701 factor against the number of loop iterations if the loop is
10704 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10705 Alan Hayward <alan.hayward@arm.com>
10706 David Sherwood <david.sherwood@arm.com>
10708 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10709 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10710 (dump_groups): Update accordingly.
10711 (iv_use::mem_type): New member variable.
10712 (address_p): New function.
10713 (record_use): Add a mem_type argument and initialize the new
10715 (record_group_use): Add a mem_type argument. Use address_p.
10716 Remove obsolete null checks of base_object. Update call to record_use.
10717 (find_interesting_uses_op): Update call to record_group_use.
10718 (find_interesting_uses_cond): Likewise.
10719 (find_interesting_uses_address): Likewise.
10720 (get_mem_type_for_internal_fn): New function.
10721 (find_address_like_use): Likewise.
10722 (find_interesting_uses_stmt): Try find_address_like_use before
10723 calling find_interesting_uses_op.
10724 (addr_offset_valid_p): Use the iv mem_type field as the type
10725 of the addressed memory.
10726 (add_autoinc_candidates): Likewise.
10727 (get_address_cost): Likewise.
10728 (split_small_address_groups_p): Use address_p.
10729 (split_address_groups): Likewise.
10730 (add_iv_candidate_for_use): Likewise.
10731 (autoinc_possible_for_pair): Likewise.
10732 (rewrite_groups): Likewise.
10733 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10734 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10735 (get_alias_ptr_type_for_ptr_address): New function.
10736 (rewrite_use_address): Rewrite address uses in calls that were
10737 identified by find_address_like_use.
10739 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10740 Alan Hayward <alan.hayward@arm.com>
10741 David Sherwood <david.sherwood@arm.com>
10743 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10745 * gimple-expr.h (is_gimple_addressable: Likewise.
10746 * gimple-expr.c (is_gimple_address): Likewise.
10747 * internal-fn.c (expand_call_mem_ref): New function.
10748 (expand_mask_load_optab_fn): Use it.
10749 (expand_mask_store_optab_fn): Likewise.
10751 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10752 Alan Hayward <alan.hayward@arm.com>
10753 David Sherwood <david.sherwood@arm.com>
10755 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10756 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10757 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10758 (cond_umax@var{mode}): Document.
10759 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10760 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10761 (cond_umin_optab, cond_umax_optab): New optabs.
10762 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10763 (COND_IOR, COND_XOR): New internal functions.
10764 * internal-fn.h (get_conditional_internal_fn): Declare.
10765 * internal-fn.c (cond_binary_direct): New macro.
10766 (expand_cond_binary_optab_fn): Likewise.
10767 (direct_cond_binary_optab_supported_p): Likewise.
10768 (get_conditional_internal_fn): New function.
10769 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10770 Cope with reduction statements that are vectorized as calls rather
10772 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10773 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10774 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10775 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10776 (UNSPEC_COND_EOR): New unspecs.
10777 (optab): Add mappings for them.
10778 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10779 (sve_int_op, sve_fp_op): New int attributes.
10781 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10782 Alan Hayward <alan.hayward@arm.com>
10783 David Sherwood <david.sherwood@arm.com>
10785 * optabs.def (while_ult_optab): New optab.
10786 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10787 * internal-fn.def (WHILE_ULT): New internal function.
10788 * internal-fn.h (direct_internal_fn_supported_p): New override
10789 that takes two types as argument.
10790 * internal-fn.c (while_direct): New macro.
10791 (expand_while_optab_fn): New function.
10792 (convert_optab_supported_p): Likewise.
10793 (direct_while_optab_supported_p): New macro.
10794 * wide-int.h (wi::udiv_ceil): New function.
10795 * tree-vectorizer.h (rgroup_masks): New structure.
10796 (vec_loop_masks): New typedef.
10797 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10798 and fully_masked_p.
10799 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10800 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10801 (vect_max_vf): New function.
10802 (slpeel_make_loop_iterate_ntimes): Delete.
10803 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10804 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10805 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10806 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10807 internal-fn.h, stor-layout.h and optabs-query.h.
10808 (vect_set_loop_mask): New function.
10809 (add_preheader_seq): Likewise.
10810 (add_header_seq): Likewise.
10811 (interleave_supported_p): Likewise.
10812 (vect_maybe_permute_loop_masks): Likewise.
10813 (vect_set_loop_masks_directly): Likewise.
10814 (vect_set_loop_condition_masked): Likewise.
10815 (vect_set_loop_condition_unmasked): New function, split out from
10816 slpeel_make_loop_iterate_ntimes.
10817 (slpeel_make_loop_iterate_ntimes): Rename to..
10818 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10819 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10820 (vect_do_peeling): Update call accordingly.
10821 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10823 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10824 mask_compare_type, can_fully_mask_p and fully_masked_p.
10825 (release_vec_loop_masks): New function.
10826 (_loop_vec_info): Use it to free the loop masks.
10827 (can_produce_all_loop_masks_p): New function.
10828 (vect_get_max_nscalars_per_iter): Likewise.
10829 (vect_verify_full_masking): Likewise.
10830 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10831 retries, and free the mask rgroups before retrying. Check loop-wide
10832 reasons for disallowing fully-masked loops. Make the final decision
10833 about whether use a fully-masked loop or not.
10834 (vect_estimate_min_profitable_iters): Do not assume that peeling
10835 for the number of iterations will be needed for fully-masked loops.
10836 (vectorizable_reduction): Disable fully-masked loops.
10837 (vectorizable_live_operation): Likewise.
10838 (vect_halve_mask_nunits): New function.
10839 (vect_double_mask_nunits): Likewise.
10840 (vect_record_loop_mask): Likewise.
10841 (vect_get_loop_mask): Likewise.
10842 (vect_transform_loop): Handle the case in which the final loop
10843 iteration might handle a partial vector. Call vect_set_loop_condition
10844 instead of slpeel_make_loop_iterate_ntimes.
10845 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10846 (check_load_store_masking): New function.
10847 (prepare_load_store_mask): Likewise.
10848 (vectorizable_store): Handle fully-masked loops.
10849 (vectorizable_load): Likewise.
10850 (supportable_widening_operation): Use vect_halve_mask_nunits for
10852 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10853 (vect_gen_while): New function.
10854 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10855 (aarch64_uqdec<mode>): New insn.
10857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10858 Alan Hayward <alan.hayward@arm.com>
10859 David Sherwood <david.sherwood@arm.com>
10861 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10862 (reduc_xor_scal_optab): New optabs.
10863 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10864 (reduc_xor_scal_@var{m}): Document.
10865 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10866 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10867 internal functions.
10868 * fold-const-call.c (fold_const_call): Handle them.
10869 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10870 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10871 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10872 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10873 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10874 (UNSPEC_XORV): New unspecs.
10875 (optab): Add entries for them.
10876 (BITWISEV): New int iterator.
10877 (bit_reduc_op): New int attributes.
10879 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10880 Alan Hayward <alan.hayward@arm.com>
10881 David Sherwood <david.sherwood@arm.com>
10883 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10884 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10885 * optabs.def (vec_shl_insert_optab): New optab.
10886 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10887 (duplicate_and_interleave): Likewise.
10888 * tree-vect-loop.c: Include internal-fn.h.
10889 (neutral_op_for_slp_reduction): New function, split out from
10890 get_initial_defs_for_reduction.
10891 (get_initial_def_for_reduction): Handle option 2 for variable-length
10892 vectors by loading the neutral value into a vector and then shifting
10893 the initial value into element 0.
10894 (get_initial_defs_for_reduction): Replace the code argument with
10895 the neutral value calculated by neutral_op_for_slp_reduction.
10896 Use gimple_build_vector for constant-length vectors.
10897 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10898 but the first group_size elements have a neutral value.
10899 Use duplicate_and_interleave otherwise.
10900 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10901 Update call to get_initial_defs_for_reduction. Handle SLP
10902 reductions for variable-length vectors by creating one vector
10903 result for each scalar result, with the elements associated
10904 with other scalar results stubbed out with the neutral value.
10905 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10906 Require IFN_VEC_SHL_INSERT for double reductions on
10907 variable-length vectors, or SLP reductions that have
10908 a neutral value. Require can_duplicate_and_interleave_p
10909 support for variable-length unchained SLP reductions if there
10910 is no neutral value, such as for MIN/MAX reductions. Also require
10911 the number of vector elements to be a multiple of the number of
10912 SLP statements when doing variable-length unchained SLP reductions.
10913 Update call to vect_create_epilog_for_reduction.
10914 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10915 and remove initial values.
10916 (duplicate_and_interleave): Make public.
10917 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10918 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10920 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10921 Alan Hayward <alan.hayward@arm.com>
10922 David Sherwood <david.sherwood@arm.com>
10924 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10925 (can_duplicate_and_interleave_p): New function.
10926 (vect_get_and_check_slp_defs): Take the vector of statements
10927 rather than just the current one. Remove excess parentheses.
10928 Restriction rejectinon of vect_constant_def and vect_external_def
10929 for variable-length vectors to boolean types, or types for which
10930 can_duplicate_and_interleave_p is false.
10931 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10932 (duplicate_and_interleave): New function.
10933 (vect_get_constant_vectors): Use gimple_build_vector for
10934 constant-length vectors and suitable variable-length constant
10935 vectors. Use duplicate_and_interleave for other variable-length
10936 vectors. Don't defer the update when inserting new statements.
10938 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10939 Alan Hayward <alan.hayward@arm.com>
10940 David Sherwood <david.sherwood@arm.com>
10942 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10943 min_profitable_iters doesn't go negative.
10945 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10946 Alan Hayward <alan.hayward@arm.com>
10947 David Sherwood <david.sherwood@arm.com>
10949 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10950 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10951 * optabs.def (vec_mask_load_lanes_optab): New optab.
10952 (vec_mask_store_lanes_optab): Likewise.
10953 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10954 (MASK_STORE_LANES): Likewise.
10955 * internal-fn.c (mask_load_lanes_direct): New macro.
10956 (mask_store_lanes_direct): Likewise.
10957 (expand_mask_load_optab_fn): Handle masked operations.
10958 (expand_mask_load_lanes_optab_fn): New macro.
10959 (expand_mask_store_optab_fn): Handle masked operations.
10960 (expand_mask_store_lanes_optab_fn): New macro.
10961 (direct_mask_load_lanes_optab_supported_p): Likewise.
10962 (direct_mask_store_lanes_optab_supported_p): Likewise.
10963 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10965 (vect_load_lanes_supported): Likewise.
10966 * tree-vect-data-refs.c (strip_conversion): New function.
10967 (can_group_stmts_p): Likewise.
10968 (vect_analyze_data_ref_accesses): Use it instead of checking
10969 for a pair of assignments.
10970 (vect_store_lanes_supported): Take a masked_p parameter.
10971 (vect_load_lanes_supported): Likewise.
10972 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10973 vect_store_lanes_supported and vect_load_lanes_supported.
10974 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10975 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10976 parameter. Don't allow gaps for masked accesses.
10977 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10978 and vect_load_lanes_supported.
10979 (get_load_store_type): Take a masked_p parameter and update
10980 call to get_group_load_store_type.
10981 (vectorizable_store): Update call to get_load_store_type.
10982 Handle IFN_MASK_STORE_LANES.
10983 (vectorizable_load): Update call to get_load_store_type.
10984 Handle IFN_MASK_LOAD_LANES.
10986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10987 Alan Hayward <alan.hayward@arm.com>
10988 David Sherwood <david.sherwood@arm.com>
10990 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10992 * config/aarch64/aarch64-protos.h
10993 (aarch64_sve_struct_memory_operand_p): Declare.
10994 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10995 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10996 (VPRED, vpred): Handle SVE structure modes.
10997 * config/aarch64/constraints.md (Utx): New constraint.
10998 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10999 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11000 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11001 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11002 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11003 structure modes. Split into pieces after RA.
11004 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11005 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11007 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11008 SVE structure modes.
11009 (aarch64_classify_address): Likewise.
11010 (sizetochar): Move earlier in file.
11011 (aarch64_print_operand): Handle SVE register lists.
11012 (aarch64_array_mode): New function.
11013 (aarch64_sve_struct_memory_operand_p): Likewise.
11014 (TARGET_ARRAY_MODE): Redefine.
11016 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11017 Alan Hayward <alan.hayward@arm.com>
11018 David Sherwood <david.sherwood@arm.com>
11020 * target.def (array_mode): New target hook.
11021 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11022 * doc/tm.texi: Regenerate.
11023 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11024 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11025 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11026 targetm.array_mode.
11027 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11030 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11031 Alan Hayward <alan.hayward@arm.com>
11032 David Sherwood <david.sherwood@arm.com>
11034 * fold-const.c (fold_binary_loc): Check the argument types
11035 rather than the result type when testing for a vector operation.
11037 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11039 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11040 * doc/tm.texi: Regenerate.
11042 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11043 Alan Hayward <alan.hayward@arm.com>
11044 David Sherwood <david.sherwood@arm.com>
11046 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11047 (sve): Document new AArch64 extension.
11048 * doc/md.texi (w): Extend the description of the AArch64
11049 constraint to include SVE vectors.
11050 (Upl, Upa): Document new AArch64 predicate constraints.
11051 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11053 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11054 (msve-vector-bits=): New option.
11055 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11056 SVE when these are disabled.
11057 (sve): New extension.
11058 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11059 modes. Adjust their number of units based on aarch64_sve_vg.
11060 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11061 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11062 aarch64_addr_query_type.
11063 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11064 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11065 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11066 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11067 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11068 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11069 (aarch64_simd_imm_zero_p): Delete.
11070 (aarch64_check_zero_based_sve_index_immediate): Declare.
11071 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11072 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11073 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11074 (aarch64_sve_float_mul_immediate_p): Likewise.
11075 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11076 rather than an rtx.
11077 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11078 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11079 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11080 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11081 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11082 (aarch64_regmode_natural_size): Likewise.
11083 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11084 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11086 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11087 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11088 for VG and the SVE predicate registers.
11089 (V_ALIASES): Add a "z"-prefixed alias.
11090 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11091 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11092 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11093 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11094 (REG_CLASS_NAMES): Add entries for them.
11095 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11096 and the predicate registers.
11097 (aarch64_sve_vg): Declare.
11098 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11099 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11100 (REGMODE_NATURAL_SIZE): Define.
11101 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11103 * config/aarch64/aarch64.c: Include cfgrtl.h.
11104 (simd_immediate_info): Add a constructor for series vectors,
11105 and an associated step field.
11106 (aarch64_sve_vg): New variable.
11107 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11108 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11109 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11110 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11111 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11112 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11113 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11114 (aarch64_get_mask_mode): New functions.
11115 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11116 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11117 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11118 predicate modes and predicate registers. Explicitly restrict
11119 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11120 to store a vector mode if it is recognized by
11121 aarch64_classify_vector_mode.
11122 (aarch64_regmode_natural_size): New function.
11123 (aarch64_hard_regno_caller_save_mode): Return the original mode
11125 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11126 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11127 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11128 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11130 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11131 does not overlap dest if the function is frame-related. Handle
11133 (aarch64_split_add_offset): New function.
11134 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11135 them aarch64_add_offset.
11136 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11137 and update call to aarch64_sub_sp.
11138 (aarch64_add_cfa_expression): New function.
11139 (aarch64_expand_prologue): Pass extra temporary registers to the
11140 functions above. Handle the case in which we need to emit new
11141 DW_CFA_expressions for registers that were originally saved
11142 relative to the stack pointer, but now have to be expressed
11143 relative to the frame pointer.
11144 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11146 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11147 IP0 and IP1 values for SVE frames.
11148 (aarch64_expand_vec_series): New function.
11149 (aarch64_expand_sve_widened_duplicate): Likewise.
11150 (aarch64_expand_sve_const_vector): Likewise.
11151 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11152 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11153 into the register, rather than emitting a SET directly.
11154 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11155 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11156 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11157 (offset_9bit_signed_scaled_p): New functions.
11158 (aarch64_replicate_bitmask_imm): New function.
11159 (aarch64_bitmask_imm): Use it.
11160 (aarch64_cannot_force_const_mem): Reject expressions involving
11161 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11162 (aarch64_classify_index): Handle SVE indices, by requiring
11163 a plain register index with a scale that matches the element size.
11164 (aarch64_classify_address): Handle SVE addresses. Assert that
11165 the mode of the address is VOIDmode or an integer mode.
11166 Update call to aarch64_classify_symbol.
11167 (aarch64_classify_symbolic_expression): Update call to
11168 aarch64_classify_symbol.
11169 (aarch64_const_vec_all_in_range_p): New function.
11170 (aarch64_print_vector_float_operand): Likewise.
11171 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11172 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11173 and the FP immediates 1.0 and 0.5.
11174 (aarch64_print_address_internal): Handle SVE addresses.
11175 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11176 (aarch64_regno_regclass): Handle predicate registers.
11177 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11179 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11180 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11181 (aarch64_convert_sve_vector_bits): New function.
11182 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11183 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11184 rather than an rtx.
11185 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11186 Handle SVE vector and predicate modes. Accept VL-based constants
11187 that need only one temporary register, and VL offsets that require
11188 no temporary registers.
11189 (aarch64_conditional_register_usage): Mark the predicate registers
11190 as fixed if SVE isn't available.
11191 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11192 Return true for SVE vector and predicate modes.
11193 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11194 rather than an unsigned int. Handle SVE modes.
11195 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11197 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11199 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11200 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11201 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11202 (aarch64_sve_float_mul_immediate_p): New functions.
11203 (aarch64_sve_valid_immediate): New function.
11204 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11205 Explicitly reject structure modes. Check for INDEX constants.
11206 Handle PTRUE and PFALSE constants.
11207 (aarch64_check_zero_based_sve_index_immediate): New function.
11208 (aarch64_simd_imm_zero_p): Delete.
11209 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11210 vector modes. Accept constants in the range of CNT[BHWD].
11211 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11212 ask for an Advanced SIMD mode.
11213 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11214 (aarch64_simd_vector_alignment): Handle SVE predicates.
11215 (aarch64_vectorize_preferred_vector_alignment): New function.
11216 (aarch64_simd_vector_alignment_reachable): Use it instead of
11218 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11219 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11221 (MAX_VECT_LEN): Delete.
11222 (expand_vec_perm_d): Add a vec_flags field.
11223 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11224 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11225 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11227 (aarch64_evpc_rev): Rename to...
11228 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11229 (aarch64_evpc_rev_global): New function.
11230 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11231 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11233 (aarch64_evpc_sve_tbl): New function.
11234 (aarch64_expand_vec_perm_const_1): Update after rename of
11235 aarch64_evpc_rev. Handle SVE permutes too, trying
11236 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11237 than aarch64_evpc_tbl.
11238 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11239 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11240 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11241 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11242 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11243 (aarch64_expand_sve_vcond): New functions.
11244 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11245 of aarch64_vector_mode_p.
11246 (aarch64_dwarf_poly_indeterminate_value): New function.
11247 (aarch64_compute_pressure_classes): Likewise.
11248 (aarch64_can_change_mode_class): Likewise.
11249 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11250 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11251 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11252 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11253 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11254 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11255 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11256 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11258 (Dn, Dl, Dr): Accept const as well as const_vector.
11259 (Dz): Likewise. Compare against CONST0_RTX.
11260 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11261 of "vector" where appropriate.
11262 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11263 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11264 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11265 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11266 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11267 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11268 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11269 (v_int_equiv): Extend to SVE modes.
11270 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11272 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11273 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11274 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11275 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11276 (SVE_COND_FP_CMP): New int iterators.
11277 (perm_hilo): Handle the new unpack unspecs.
11278 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11280 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11281 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11282 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11283 (aarch64_equality_operator, aarch64_constant_vector_operand)
11284 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11285 (aarch64_sve_nonimmediate_operand): Likewise.
11286 (aarch64_sve_general_operand): Likewise.
11287 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11288 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11289 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11290 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11291 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11292 (aarch64_sve_float_arith_immediate): Likewise.
11293 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11294 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11295 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11296 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11297 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11298 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11299 (aarch64_sve_float_arith_operand): Likewise.
11300 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11301 (aarch64_sve_float_mul_operand): Likewise.
11302 (aarch64_sve_vec_perm_operand): Likewise.
11303 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11304 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11305 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11306 as well as const_vector.
11307 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11308 in file. Use CONST0_RTX and CONSTM1_RTX.
11309 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11310 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11311 Use aarch64_simd_imm_zero.
11312 * config/aarch64/aarch64-sve.md: New file.
11313 * config/aarch64/aarch64.md: Include it.
11314 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11315 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11316 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11317 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11318 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11319 (sve): New attribute.
11320 (enabled): Disable instructions with the sve attribute unless
11322 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11323 aarch64_expand_mov_immediate.
11324 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11325 CNT[BHSD] immediates.
11326 (movti): Split CONST_POLY_INT moves into two halves.
11327 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11328 Split additions that need a temporary here if the destination
11329 is the stack pointer.
11330 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11331 (*add<mode>3_poly_1): New instruction.
11332 (set_clobber_cc): New expander.
11334 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11336 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11337 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11338 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11339 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11340 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11341 Change innermode from fixed_mode_size to machine_mode.
11342 (simplify_subreg): Update call accordingly. Handle a constant-sized
11343 subreg of a variable-length CONST_VECTOR.
11345 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11346 Alan Hayward <alan.hayward@arm.com>
11347 David Sherwood <david.sherwood@arm.com>
11349 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11350 (add_offset_to_base): New function, split out from...
11351 (create_mem_ref): ...here. When handling a scale other than 1,
11352 check first whether the address is valid without the offset.
11353 Add it into the base if so, leaving the index and scale as-is.
11355 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11358 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11359 fold_for_warn before checking if arg2 is INTEGER_CST.
11361 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11363 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11364 (store_multiple_operation): Delete.
11365 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11366 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11367 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11368 guarded by TARGET_STRING.
11369 (rs6000_output_load_multiple): Delete.
11370 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11371 OPTION_MASK_STRING / TARGET_STRING handling.
11372 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11373 (const rs6000_opt_masks) <"string">: Change mask to 0.
11374 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11375 (MASK_STRING): Delete.
11376 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11378 (load_multiple): Delete.
11385 (store_multiple): Delete.
11392 (movmemsi_8reg): Delete.
11393 (corresponding unnamed define_insn): Delete.
11394 (movmemsi_6reg): Delete.
11395 (corresponding unnamed define_insn): Delete.
11396 (movmemsi_4reg): Delete.
11397 (corresponding unnamed define_insn): Delete.
11398 (movmemsi_2reg): Delete.
11399 (corresponding unnamed define_insn): Delete.
11400 (movmemsi_1reg): Delete.
11401 (corresponding unnamed define_insn): Delete.
11402 * config/rs6000/rs6000.opt (mno-string): New.
11403 (mstring): Replace by deprecation warning stub.
11404 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11406 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11408 * regrename.c (regrename_do_replace): If replacing the same
11409 reg multiple times, try to reuse last created gen_raw_REG.
11412 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11413 main to workaround a bug in GDB.
11415 2018-01-12 Tom de Vries <tom@codesourcery.com>
11418 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11420 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11422 PR rtl-optimization/80481
11423 * ira-color.c (get_cap_member): New function.
11424 (allocnos_conflict_by_live_ranges_p): Use it.
11425 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11426 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11428 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11431 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11432 (*saddl_se_1): Ditto.
11433 (*ssubsi_1): Ditto.
11434 (*ssubl_se_1): Ditto.
11436 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11438 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11439 rather than wi::to_widest for DR_INITs.
11440 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11441 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11442 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11444 (vect_analyze_group_access_1): Note that here.
11446 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11448 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11449 polynomial type sizes.
11451 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11453 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11454 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11455 (gimple_add_tmp_var): Likewise.
11457 2018-01-12 Martin Liska <mliska@suse.cz>
11459 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11460 (gimple_alloc_sizes): Likewise.
11461 (dump_gimple_statistics): Use PRIu64 in printf format.
11462 * gimple.h: Change uint64_t to int.
11464 2018-01-12 Martin Liska <mliska@suse.cz>
11466 * tree-core.h: Use uint64_t instead of int.
11467 * tree.c (tree_node_counts): Likewise.
11468 (tree_node_sizes): Likewise.
11469 (dump_tree_statistics): Use PRIu64 in printf format.
11471 2018-01-12 Martin Liska <mliska@suse.cz>
11473 * Makefile.in: As qsort_chk is implemented in vec.c, add
11474 vec.o to linkage of gencfn-macros.
11475 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11476 passing the info to record_node_allocation_statistics.
11477 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11479 * ggc-common.c (struct ggc_usage): Add operator== and use
11480 it in operator< and compare function.
11481 * mem-stats.h (struct mem_usage): Likewise.
11482 * vec.c (struct vec_usage): Remove operator< and compare
11483 function. Can be simply inherited.
11485 2018-01-12 Martin Jambor <mjambor@suse.cz>
11488 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11489 * tree-ssa-math-opts.c: Include domwalk.h.
11490 (convert_mult_to_fma_1): New function.
11491 (fma_transformation_info): New type.
11492 (fma_deferring_state): Likewise.
11493 (cancel_fma_deferring): New function.
11494 (result_of_phi): Likewise.
11495 (last_fma_candidate_feeds_initial_phi): Likewise.
11496 (convert_mult_to_fma): Added deferring logic, split actual
11497 transformation to convert_mult_to_fma_1.
11498 (math_opts_dom_walker): New type.
11499 (math_opts_dom_walker::after_dom_children): New method, body moved
11500 here from pass_optimize_widening_mul::execute, added deferring logic
11502 (pass_optimize_widening_mul::execute): Moved most of code to
11503 math_opts_dom_walker::after_dom_children.
11504 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11505 * config/i386/i386.c (ix86_option_override_internal): Added
11506 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11508 2018-01-12 Richard Biener <rguenther@suse.de>
11511 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11512 inline instance vars.
11514 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11517 * config/rx/rx.c (rx_is_restricted_memory_address):
11518 Handle SUBREG case.
11520 2018-01-12 Richard Biener <rguenther@suse.de>
11522 PR tree-optimization/80846
11523 * target.def (split_reduction): New target hook.
11524 * targhooks.c (default_split_reduction): New function.
11525 * targhooks.h (default_split_reduction): Declare.
11526 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11527 target requests first reduce vectors by combining low and high
11529 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11530 (get_vectype_for_scalar_type_and_size): Export.
11531 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11532 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11533 * doc/tm.texi: Regenerate.
11534 * config/i386/i386.c (ix86_split_reduction): Implement
11535 TARGET_VECTORIZE_SPLIT_REDUCTION.
11537 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11540 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11541 in PIC mode except for TARGET_VXWORKS_RTP.
11542 * config/sparc/sparc.c: Include cfgrtl.h.
11543 (TARGET_INIT_PIC_REG): Define.
11544 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11545 (sparc_pic_register_p): New predicate.
11546 (sparc_legitimate_address_p): Use it.
11547 (sparc_legitimize_pic_address): Likewise.
11548 (sparc_delegitimize_address): Likewise.
11549 (sparc_mode_dependent_address_p): Likewise.
11550 (gen_load_pcrel_sym): Remove 4th parameter.
11551 (load_got_register): Adjust call to above. Remove obsolete stuff.
11552 (sparc_expand_prologue): Do not call load_got_register here.
11553 (sparc_flat_expand_prologue): Likewise.
11554 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11555 (sparc_use_pseudo_pic_reg): New function.
11556 (sparc_init_pic_reg): Likewise.
11557 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11558 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11560 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11562 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11563 Add item for branch_cost.
11565 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11567 PR rtl-optimization/83565
11568 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11569 not extend the result to a larger mode for rotate operations.
11570 (num_sign_bit_copies1): Likewise.
11572 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11575 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11577 Use values-Xc.o for -pedantic.
11578 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11580 2018-01-12 Martin Liska <mliska@suse.cz>
11583 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11585 (possible_polymorphic_call_targets): Use it.
11586 (ipa_devirt): Likewise.
11588 2018-01-12 Martin Liska <mliska@suse.cz>
11590 * profile-count.h (enum profile_quality): Use 0 as invalid
11591 enum value of profile_quality.
11593 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11595 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11596 -mext-string options.
11598 2018-01-12 Richard Biener <rguenther@suse.de>
11600 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11601 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11602 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11604 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11606 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11608 * configure.ac (--with-long-double-format): Add support for the
11609 configuration option to change the default long double format on
11611 * config.gcc (powerpc*-linux*-*): Likewise.
11612 * configure: Regenerate.
11613 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11614 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11615 used without modification.
11617 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11619 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11620 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11621 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11622 MISC_BUILTIN_SPEC_BARRIER.
11623 (rs6000_init_builtins): Likewise.
11624 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11626 (speculation_barrier): New define_insn.
11627 * doc/extend.texi: Document __builtin_speculation_barrier.
11629 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11632 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11633 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11634 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11636 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11637 integral modes instead of "ss" and "sd".
11638 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11639 vectors with 32-bit and 64-bit elements.
11640 (vecdupssescalarmodesuffix): New mode attribute.
11641 (vec_dup<mode>): Use it.
11643 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11646 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11647 frame if argument is passed on stack.
11649 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11652 * ree.c (combine_reaching_defs): Optimize also
11653 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11654 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11656 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11658 PR middle-end/83189
11659 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11661 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11663 PR middle-end/83718
11664 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11665 after they are computed.
11667 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11669 PR tree-optimization/83695
11670 * gimple-loop-linterchange.cc
11671 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11672 reset cached scev information after interchange.
11673 (pass_linterchange::execute): Remove call to scev_reset_htab.
11675 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11677 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11678 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11679 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11680 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11681 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11682 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11683 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11684 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11685 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11686 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11687 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11688 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11689 (V_lane_reg): Likewise.
11690 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11692 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11693 (vfmal_lane_low<mode>_intrinsic,
11694 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11695 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11696 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11697 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11698 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11699 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11701 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11703 * config/arm/arm-cpus.in (fp16fml): New feature.
11704 (ALL_SIMD): Add fp16fml.
11705 (armv8.2-a): Add fp16fml as an option.
11706 (armv8.3-a): Likewise.
11707 (armv8.4-a): Add fp16fml as part of fp16.
11708 * config/arm/arm.h (TARGET_FP16FML): Define.
11709 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11711 * config/arm/arm-modes.def (V2HF): Define.
11712 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11713 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11714 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11715 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11716 vfmsl_low, vfmsl_high): New set of builtins.
11717 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11718 (vfml_op): New code attribute.
11719 (VFMLHALVES): New int iterator.
11720 (VFML, VFMLSEL): New mode attributes.
11721 (V_reg): Define mapping for V2HF.
11722 (V_hi, V_lo): New mode attributes.
11723 (VF_constraint): Likewise.
11724 (vfml_half, vfml_half_selector): New int attributes.
11725 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11727 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11728 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11730 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11731 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11732 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11733 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11735 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11736 Document new effective target and option set.
11738 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11740 * config/arm/arm-cpus.in (armv8_4): New feature.
11741 (ARMv8_4a): New fgroup.
11742 (armv8.4-a): New arch.
11743 * config/arm/arm-tables.opt: Regenerate.
11744 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11745 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11746 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11747 Add matching rules for -march=armv8.4-a and extensions.
11748 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11750 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11753 * config/rx/rx.md (BW): New mode attribute.
11754 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11756 2018-01-11 Richard Biener <rguenther@suse.de>
11758 PR tree-optimization/83435
11759 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11760 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11761 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11763 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11764 Alan Hayward <alan.hayward@arm.com>
11765 David Sherwood <david.sherwood@arm.com>
11767 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11769 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11770 (aarch64_print_address_internal): Use it to check for a zero offset.
11772 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11773 Alan Hayward <alan.hayward@arm.com>
11774 David Sherwood <david.sherwood@arm.com>
11776 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11777 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11778 Return a poly_int64 rather than a HOST_WIDE_INT.
11779 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11780 rather than a HOST_WIDE_INT.
11781 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11782 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11783 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11784 final_offset from HOST_WIDE_INT to poly_int64.
11785 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11786 to_constant when getting the number of units in an Advanced SIMD
11788 (aarch64_builtin_vectorized_function): Check for a constant number
11790 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11792 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11793 attribute instead of GET_MODE_NUNITS.
11794 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11795 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11796 GET_MODE_SIZE for fixed-size registers.
11797 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11798 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11799 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11800 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11801 (aarch64_print_operand, aarch64_print_address_internal)
11802 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11803 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11804 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11805 Handle polynomial GET_MODE_SIZE.
11806 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11807 wider than SImode without modification.
11808 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11809 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11810 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11811 passing and returning SVE modes.
11812 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11813 rather than GEN_INT.
11814 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11815 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11816 (aarch64_allocate_and_probe_stack_space): Likewise.
11817 (aarch64_layout_frame): Cope with polynomial offsets.
11818 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11819 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11820 polynomial offsets.
11821 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11822 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11823 poly_int64 rather than a HOST_WIDE_INT.
11824 (aarch64_get_separate_components, aarch64_process_components)
11825 (aarch64_expand_prologue, aarch64_expand_epilogue)
11826 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11827 (aarch64_anchor_offset): New function, split out from...
11828 (aarch64_legitimize_address): ...here.
11829 (aarch64_builtin_vectorization_cost): Handle polynomial
11830 TYPE_VECTOR_SUBPARTS.
11831 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11833 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11834 number of elements from the PARALLEL rather than the mode.
11835 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11836 rather than GET_MODE_BITSIZE.
11837 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11838 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11839 (aarch64_expand_vec_perm_const_1): Handle polynomial
11840 d->perm.length () and d->perm elements.
11841 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11842 Apply to_constant to d->perm elements.
11843 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11844 polynomial CONST_VECTOR_NUNITS.
11845 (aarch64_move_pointer): Take amount as a poly_int64 rather
11847 (aarch64_progress_pointer): Avoid temporary variable.
11848 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11849 the mode attribute instead of GET_MODE.
11851 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11852 Alan Hayward <alan.hayward@arm.com>
11853 David Sherwood <david.sherwood@arm.com>
11855 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11856 x exists before using it.
11857 (aarch64_add_constant_internal): Rename to...
11858 (aarch64_add_offset_1): ...this. Replace regnum with separate
11859 src and dest rtxes. Handle the case in which they're different,
11860 including when the offset is zero. Replace scratchreg with an rtx.
11861 Use 2 additions if there is no spare register into which we can
11862 move a 16-bit constant.
11863 (aarch64_add_constant): Delete.
11864 (aarch64_add_offset): Replace reg with separate src and dest
11865 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11866 Use aarch64_add_offset_1.
11867 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11868 an rtx rather than an int. Take the delta as a poly_int64
11869 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11870 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11871 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11872 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11873 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11874 and aarch64_add_sp.
11875 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11876 aarch64_add_constant.
11878 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11880 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11881 Use scalar_float_mode.
11883 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11885 * config/aarch64/aarch64-simd.md
11886 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11887 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11888 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11889 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11890 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11891 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11892 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11893 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11894 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11895 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11897 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11900 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11901 targ_options->x_arm_arch_string is non NULL.
11903 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11905 * config/aarch64/aarch64.h
11906 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11908 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11911 * expmed.c (emit_store_flag_force): Swap if const op0
11912 and change VOIDmode to mode of op0.
11914 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11916 PR rtl-optimization/83761
11917 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11918 than bytes to mode_for_size.
11920 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11922 PR middle-end/83189
11923 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11924 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11927 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11929 PR middle-end/83575
11930 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11931 when in layout mode.
11932 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11933 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11936 2018-01-10 Michael Collison <michael.collison@arm.com>
11938 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11939 * config/aarch64/aarch64-option-extension.def: Add
11940 AARCH64_OPT_EXTENSION of 'fp16fml'.
11941 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11942 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11943 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11944 * config/aarch64/constraints.md (Ui7): New constraint.
11945 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11946 (VFMLA_SEL_W): Ditto.
11949 (VFMLA16_LOW): New int iterator.
11950 (VFMLA16_HIGH): Ditto.
11951 (UNSPEC_FMLAL): New unspec.
11952 (UNSPEC_FMLSL): Ditto.
11953 (UNSPEC_FMLAL2): Ditto.
11954 (UNSPEC_FMLSL2): Ditto.
11955 (f16mac): New code attribute.
11956 * config/aarch64/aarch64-simd-builtins.def
11957 (aarch64_fmlal_lowv2sf): Ditto.
11958 (aarch64_fmlsl_lowv2sf): Ditto.
11959 (aarch64_fmlalq_lowv4sf): Ditto.
11960 (aarch64_fmlslq_lowv4sf): Ditto.
11961 (aarch64_fmlal_highv2sf): Ditto.
11962 (aarch64_fmlsl_highv2sf): Ditto.
11963 (aarch64_fmlalq_highv4sf): Ditto.
11964 (aarch64_fmlslq_highv4sf): Ditto.
11965 (aarch64_fmlal_lane_lowv2sf): Ditto.
11966 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11967 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11968 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11969 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11970 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11971 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11972 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11973 (aarch64_fmlal_lane_highv2sf): Ditto.
11974 (aarch64_fmlsl_lane_highv2sf): Ditto.
11975 (aarch64_fmlal_laneq_highv2sf): Ditto.
11976 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11977 (aarch64_fmlalq_lane_highv4sf): Ditto.
11978 (aarch64_fmlsl_lane_highv4sf): Ditto.
11979 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11980 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11981 * config/aarch64/aarch64-simd.md:
11982 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11983 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11984 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11985 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11986 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11987 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11988 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11989 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11990 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11991 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11992 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11993 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11994 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11995 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11996 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11997 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11998 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11999 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12000 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12001 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12002 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12003 (vfmlsl_low_u32): Ditto.
12004 (vfmlalq_low_u32): Ditto.
12005 (vfmlslq_low_u32): Ditto.
12006 (vfmlal_high_u32): Ditto.
12007 (vfmlsl_high_u32): Ditto.
12008 (vfmlalq_high_u32): Ditto.
12009 (vfmlslq_high_u32): Ditto.
12010 (vfmlal_lane_low_u32): Ditto.
12011 (vfmlsl_lane_low_u32): Ditto.
12012 (vfmlal_laneq_low_u32): Ditto.
12013 (vfmlsl_laneq_low_u32): Ditto.
12014 (vfmlalq_lane_low_u32): Ditto.
12015 (vfmlslq_lane_low_u32): Ditto.
12016 (vfmlalq_laneq_low_u32): Ditto.
12017 (vfmlslq_laneq_low_u32): Ditto.
12018 (vfmlal_lane_high_u32): Ditto.
12019 (vfmlsl_lane_high_u32): Ditto.
12020 (vfmlal_laneq_high_u32): Ditto.
12021 (vfmlsl_laneq_high_u32): Ditto.
12022 (vfmlalq_lane_high_u32): Ditto.
12023 (vfmlslq_lane_high_u32): Ditto.
12024 (vfmlalq_laneq_high_u32): Ditto.
12025 (vfmlslq_laneq_high_u32): Ditto.
12026 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12027 (AARCH64_FL_FOR_ARCH8_4): New.
12028 (AARCH64_ISA_F16FML): New ISA flag.
12029 (TARGET_F16FML): New feature flag for fp16fml.
12030 (doc/invoke.texi): Document new fp16fml option.
12032 2018-01-10 Michael Collison <michael.collison@arm.com>
12034 * config/aarch64/aarch64-builtins.c:
12035 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12036 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12037 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12038 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12039 (AARCH64_ISA_SHA3): New ISA flag.
12040 (TARGET_SHA3): New feature flag for sha3.
12041 * config/aarch64/iterators.md (sha512_op): New int attribute.
12042 (CRYPTO_SHA512): New int iterator.
12043 (UNSPEC_SHA512H): New unspec.
12044 (UNSPEC_SHA512H2): Ditto.
12045 (UNSPEC_SHA512SU0): Ditto.
12046 (UNSPEC_SHA512SU1): Ditto.
12047 * config/aarch64/aarch64-simd-builtins.def
12048 (aarch64_crypto_sha512hqv2di): New builtin.
12049 (aarch64_crypto_sha512h2qv2di): Ditto.
12050 (aarch64_crypto_sha512su0qv2di): Ditto.
12051 (aarch64_crypto_sha512su1qv2di): Ditto.
12052 (aarch64_eor3qv8hi): Ditto.
12053 (aarch64_rax1qv2di): Ditto.
12054 (aarch64_xarqv2di): Ditto.
12055 (aarch64_bcaxqv8hi): Ditto.
12056 * config/aarch64/aarch64-simd.md:
12057 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12058 (aarch64_crypto_sha512su0qv2di): Ditto.
12059 (aarch64_crypto_sha512su1qv2di): Ditto.
12060 (aarch64_eor3qv8hi): Ditto.
12061 (aarch64_rax1qv2di): Ditto.
12062 (aarch64_xarqv2di): Ditto.
12063 (aarch64_bcaxqv8hi): Ditto.
12064 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12065 (vsha512h2q_u64): Ditto.
12066 (vsha512su0q_u64): Ditto.
12067 (vsha512su1q_u64): Ditto.
12068 (veor3q_u16): Ditto.
12069 (vrax1q_u64): Ditto.
12070 (vxarq_u64): Ditto.
12071 (vbcaxq_u16): Ditto.
12072 * config/arm/types.md (crypto_sha512): New type attribute.
12073 (crypto_sha3): Ditto.
12074 (doc/invoke.texi): Document new sha3 option.
12076 2018-01-10 Michael Collison <michael.collison@arm.com>
12078 * config/aarch64/aarch64-builtins.c:
12079 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12080 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12081 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12082 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12083 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12084 (AARCH64_ISA_SM4): New ISA flag.
12085 (TARGET_SM4): New feature flag for sm4.
12086 * config/aarch64/aarch64-simd-builtins.def
12087 (aarch64_sm3ss1qv4si): Ditto.
12088 (aarch64_sm3tt1aq4si): Ditto.
12089 (aarch64_sm3tt1bq4si): Ditto.
12090 (aarch64_sm3tt2aq4si): Ditto.
12091 (aarch64_sm3tt2bq4si): Ditto.
12092 (aarch64_sm3partw1qv4si): Ditto.
12093 (aarch64_sm3partw2qv4si): Ditto.
12094 (aarch64_sm4eqv4si): Ditto.
12095 (aarch64_sm4ekeyqv4si): Ditto.
12096 * config/aarch64/aarch64-simd.md:
12097 (aarch64_sm3ss1qv4si): Ditto.
12098 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12099 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12100 (aarch64_sm4eqv4si): Ditto.
12101 (aarch64_sm4ekeyqv4si): Ditto.
12102 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12103 (sm3part_op): Ditto.
12104 (CRYPTO_SM3TT): Ditto.
12105 (CRYPTO_SM3PART): Ditto.
12106 (UNSPEC_SM3SS1): New unspec.
12107 (UNSPEC_SM3TT1A): Ditto.
12108 (UNSPEC_SM3TT1B): Ditto.
12109 (UNSPEC_SM3TT2A): Ditto.
12110 (UNSPEC_SM3TT2B): Ditto.
12111 (UNSPEC_SM3PARTW1): Ditto.
12112 (UNSPEC_SM3PARTW2): Ditto.
12113 (UNSPEC_SM4E): Ditto.
12114 (UNSPEC_SM4EKEY): Ditto.
12115 * config/aarch64/constraints.md (Ui2): New constraint.
12116 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12117 * config/arm/types.md (crypto_sm3): New type attribute.
12118 (crypto_sm4): Ditto.
12119 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12120 (vsm3tt1aq_u32): Ditto.
12121 (vsm3tt1bq_u32): Ditto.
12122 (vsm3tt2aq_u32): Ditto.
12123 (vsm3tt2bq_u32): Ditto.
12124 (vsm3partw1q_u32): Ditto.
12125 (vsm3partw2q_u32): Ditto.
12126 (vsm4eq_u32): Ditto.
12127 (vsm4ekeyq_u32): Ditto.
12128 (doc/invoke.texi): Document new sm4 option.
12130 2018-01-10 Michael Collison <michael.collison@arm.com>
12132 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12133 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12134 (AARCH64_FL_FOR_ARCH8_4): New.
12135 (AARCH64_FL_V8_4): New flag.
12136 (doc/invoke.texi): Document new armv8.4-a option.
12138 2018-01-10 Michael Collison <michael.collison@arm.com>
12140 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12141 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12142 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12143 * config/aarch64/aarch64-option-extension.def: Add
12144 AARCH64_OPT_EXTENSION of 'sha2'.
12145 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12146 (crypto): Disable sha2 and aes if crypto disabled.
12147 (crypto): Enable aes and sha2 if enabled.
12148 (simd): Disable sha2 and aes if simd disabled.
12149 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12151 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12152 (TARGET_SHA2): New feature flag for sha2.
12153 (TARGET_AES): New feature flag for aes.
12154 * config/aarch64/aarch64-simd.md:
12155 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12156 conditional on TARGET_AES.
12157 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12158 (aarch64_crypto_sha1hsi): Make pattern conditional
12160 (aarch64_crypto_sha1hv4si): Ditto.
12161 (aarch64_be_crypto_sha1hv4si): Ditto.
12162 (aarch64_crypto_sha1su1v4si): Ditto.
12163 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12164 (aarch64_crypto_sha1su0v4si): Ditto.
12165 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12166 (aarch64_crypto_sha256su0v4si): Ditto.
12167 (aarch64_crypto_sha256su1v4si): Ditto.
12168 (doc/invoke.texi): Document new aes and sha2 options.
12170 2018-01-10 Martin Sebor <msebor@redhat.com>
12172 PR tree-optimization/83781
12173 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12176 2018-01-11 Martin Sebor <msebor@gmail.com>
12177 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12179 PR tree-optimization/83501
12180 PR tree-optimization/81703
12182 * tree-ssa-strlen.c (get_string_cst): Rename...
12183 (get_string_len): ...to this. Handle global constants.
12184 (handle_char_store): Adjust.
12186 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12187 Jim Wilson <jimw@sifive.com>
12189 * config/riscv/riscv-protos.h (riscv_output_return): New.
12190 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12191 (riscv_attribute_table, riscv_output_return),
12192 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12193 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12194 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12196 (riscv_expand_prologue): Add early return for naked function.
12197 (riscv_expand_epilogue): Likewise.
12198 (riscv_function_ok_for_sibcall): Return false for naked function.
12199 (riscv_set_current_function): New.
12200 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12201 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12202 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12203 * doc/extend.texi (RISC-V Function Attributes): New.
12205 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12207 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12208 check for 128-bit long double before checking TCmode.
12209 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12210 128-bit long doubles before checking TFmode or TCmode.
12211 (FLOAT128_IBM_P): Likewise.
12213 2018-01-10 Martin Sebor <msebor@redhat.com>
12215 PR tree-optimization/83671
12216 * builtins.c (c_strlen): Unconditionally return zero for the empty
12218 Use -Warray-bounds for warnings.
12219 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12220 for non-constant array indices with COMPONENT_REF, arrays of
12221 arrays, and pointers to arrays.
12222 (gimple_fold_builtin_strlen): Determine and set length range for
12223 non-constant character arrays.
12225 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12227 PR middle-end/81897
12228 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12231 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12233 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12235 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12238 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12239 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12240 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12241 indexed_or_indirect_operand predicate.
12242 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12243 (*vsx_le_perm_load_v8hi): Likewise.
12244 (*vsx_le_perm_load_v16qi): Likewise.
12245 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12246 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12247 (*vsx_le_perm_store_v8hi): Likewise.
12248 (*vsx_le_perm_store_v16qi): Likewise.
12249 (eight unnamed splitters): Likewise.
12251 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12253 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12254 * config/rs6000/emmintrin.h: Likewise.
12255 * config/rs6000/mmintrin.h: Likewise.
12256 * config/rs6000/xmmintrin.h: Likewise.
12258 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12261 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12263 * tree.c (tree_nop_conversion): Return true for location wrapper
12265 (maybe_wrap_with_location): New function.
12266 (selftest::check_strip_nops): New function.
12267 (selftest::test_location_wrappers): New function.
12268 (selftest::tree_c_tests): Call it.
12269 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12270 (maybe_wrap_with_location): New decl.
12271 (EXPR_LOCATION_WRAPPER_P): New macro.
12272 (location_wrapper_p): New inline function.
12273 (tree_strip_any_location_wrapper): New inline function.
12275 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12278 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12279 stack_realign_offset for the largest alignment of stack slot
12281 (ix86_find_max_used_stack_alignment): New function.
12282 (ix86_finalize_stack_frame_flags): Use it. Set
12283 max_used_stack_alignment if we don't realign stack.
12284 * config/i386/i386.h (machine_function): Add
12285 max_used_stack_alignment.
12287 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12289 * config/arm/arm.opt (-mbranch-cost): New option.
12290 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12293 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12296 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12297 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12299 2018-01-10 Richard Biener <rguenther@suse.de>
12302 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12303 early out so it also covers the case where we have a non-NULL
12306 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12308 PR tree-optimization/83753
12309 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12310 for non-strided grouped accesses if the number of elements is 1.
12312 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12315 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12316 * i386.h (TARGET_USE_GATHER): Define.
12317 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12319 2018-01-10 Martin Liska <mliska@suse.cz>
12322 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12323 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12325 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12326 CLEANUP_NO_PARTITIONING is not set.
12328 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12330 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12331 for vectors, as a partial revert of r254296.
12332 * rtl.h (const_vec_p): Delete.
12333 (const_vec_duplicate_p): Don't test for vector CONSTs.
12334 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12335 * expmed.c (make_tree): Likewise.
12338 * common.md (E, F): Use CONSTANT_P instead of checking for
12340 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12341 checking for CONST_VECTOR.
12343 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12345 PR middle-end/83575
12346 * predict.c (force_edge_cold): Handle in more sane way edges
12347 with no prediction.
12349 2018-01-09 Carl Love <cel@us.ibm.com>
12351 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12353 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12354 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12355 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12356 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12357 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12358 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12359 * config/rs6000/rs6000-protos.h: Add extern defition for
12360 rs6000_generate_float2_double_code.
12361 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12363 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12364 (float2_v2df): Add define_expand.
12366 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12369 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12370 op_mode in the force_to_mode call.
12372 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12374 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12375 instead of checking each element individually.
12376 (aarch64_evpc_uzp): Likewise.
12377 (aarch64_evpc_zip): Likewise.
12378 (aarch64_evpc_ext): Likewise.
12379 (aarch64_evpc_rev): Likewise.
12380 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12381 instead of checking each element individually. Return true without
12383 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12384 whether all selected elements come from the same input, instead of
12385 checking each element individually. Remove calls to gen_rtx_REG,
12386 start_sequence and end_sequence and instead assert that no rtl is
12389 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12391 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12392 order of HIGH and CONST checks.
12394 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12396 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12397 if the destination isn't an SSA_NAME.
12399 2018-01-09 Richard Biener <rguenther@suse.de>
12401 PR tree-optimization/83668
12402 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12404 (canonicalize_loop_form): ... here, renamed from ...
12405 (canonicalize_loop_closed_ssa_form): ... this and amended to
12406 swap successor edges for loop exit blocks to make us use
12407 the RPO order we need for initial schedule generation.
12409 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12411 PR tree-optimization/64811
12412 * match.pd: When optimizing comparisons with Inf, avoid
12413 introducing or losing exceptions from comparisons with NaN.
12415 2018-01-09 Martin Liska <mliska@suse.cz>
12418 * asan.c (shadow_mem_size): Add gcc_assert.
12420 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12422 Don't save registers in main().
12425 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12426 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12427 * config/avr/avr.c (avr_set_current_function): Don't error if
12428 naked, OS_task or OS_main are specified at the same time.
12429 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12431 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12433 * common/config/avr/avr-common.c (avr_option_optimization_table):
12434 Switch on -mmain-is-OS_task for optimizing compilations.
12436 2018-01-09 Richard Biener <rguenther@suse.de>
12438 PR tree-optimization/83572
12439 * graphite.c: Include cfganal.h.
12440 (graphite_transform_loops): Connect infinite loops to exit
12441 and remove fake edges at the end.
12443 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12445 * ipa-inline.c (edge_badness): Revert accidental checkin.
12447 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12450 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12451 symbols; not inline clones.
12453 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12456 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12457 hard registers. Formatting fixes.
12459 PR preprocessor/83722
12460 * gcc.c (try_generate_repro): Pass
12461 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12462 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12465 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12466 Kito Cheng <kito.cheng@gmail.com>
12468 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12469 (riscv_leaf_function_p): Delete.
12470 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12472 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12474 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12476 (do_ifelse): New function.
12477 (do_isel): New function.
12478 (do_sub3): New function.
12479 (do_add3): New function.
12480 (do_load_mask_compare): New function.
12481 (do_overlap_load_compare): New function.
12482 (expand_compare_loop): New function.
12483 (expand_block_compare): Call expand_compare_loop() when appropriate.
12484 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12485 option description.
12486 (-mblock-compare-inline-loop-limit): New option.
12488 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12491 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12492 Reverse order of second and third operands in first alternative.
12493 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12494 of first and second elements in UNSPEC_VPERMR vector.
12495 (altivec_expand_vec_perm_le): Likewise.
12497 2018-01-08 Jeff Law <law@redhat.com>
12499 PR rtl-optimizatin/81308
12500 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12501 (process_switch): If group_case_labels makes a change, then set
12503 (pass_convert_switch::execute): If a switch is converted, then
12504 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12506 PR rtl-optimization/81308
12507 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12510 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12512 PR target/83663 - Revert r255946
12513 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12514 generation for cases where splatting a value is not useful.
12515 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12516 across a vec_duplicate and a paradoxical subreg forming a vector
12517 mode to a vec_concat.
12519 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12521 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12522 -march=armv8.3-a variants.
12523 * config/arm/t-multilib: Likewise.
12524 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12526 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12528 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12530 (cceq_ior_compare_complement): Give it a name so I can use it, and
12531 change boolean_or_operator predicate to boolean_operator so it can
12532 be used to generate a crand.
12533 (eqne): New code iterator.
12534 (bd/bd_neg): New code_attrs.
12535 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12536 a single define_insn.
12537 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12538 decrement (bdnzt/bdnzf/bdzt/bdzf).
12539 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12540 with the new names of the branch decrement patterns, and added the
12541 names of the branch decrement conditional patterns.
12543 2018-01-08 Richard Biener <rguenther@suse.de>
12545 PR tree-optimization/83563
12546 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12549 2018-01-08 Richard Biener <rguenther@suse.de>
12551 PR middle-end/83713
12552 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12554 2018-01-08 Richard Biener <rguenther@suse.de>
12556 PR tree-optimization/83685
12557 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12558 references to abnormals.
12560 2018-01-08 Richard Biener <rguenther@suse.de>
12563 * dwarf2out.c (output_indirect_strings): Handle empty
12564 skeleton_debug_str_hash.
12565 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12567 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12569 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12570 (emit_store_direct): Likewise.
12571 (arc_trampoline_adjust_address): Likewise.
12572 (arc_asm_trampoline_template): New function.
12573 (arc_initialize_trampoline): Use asm_trampoline_template.
12574 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12575 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12576 * config/arc/arc.md (flush_icache): Delete pattern.
12578 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12580 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12581 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12584 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12587 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12588 by not USED_FOR_TARGET.
12589 (make_pass_resolve_sw_modes): Likewise.
12591 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12593 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12596 2018-01-08 Richard Biener <rguenther@suse.de>
12598 PR middle-end/83580
12599 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12601 2018-01-08 Richard Biener <rguenther@suse.de>
12603 PR middle-end/83517
12604 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12606 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12608 PR middle-end/81897
12609 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12610 basic blocks with a small number of successors.
12611 (convert_control_dep_chain_into_preds): Improve handling of
12613 (dump_predicates): Split apart into...
12614 (dump_pred_chain): ...here...
12615 (dump_pred_info): ...and here.
12616 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12617 (can_chain_union_be_invalidated_p): Improve check for invalidation
12619 (uninit_uses_cannot_happen): Avoid unnecessary if
12620 convert_control_dep_chain_into_preds yielded nothing.
12622 2018-01-06 Martin Sebor <msebor@redhat.com>
12624 PR tree-optimization/83640
12625 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12626 subtracting negative offset from size.
12627 (builtin_access::overlap): Adjust offset bounds of the access to fall
12628 within the size of the object if possible.
12630 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12632 PR rtl-optimization/83699
12633 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12634 extract_bit_field_as_subreg to cases in which the extracted
12635 value is also a vector.
12637 * lra-constraints.c (process_alt_operands): Test for the equivalence
12638 substitutions when detecting a possible reload cycle.
12640 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12643 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12644 by default if flag_selective_schedling{,2}. Formatting fixes.
12646 PR rtl-optimization/83682
12647 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12648 if it has non-VECTOR_MODE element mode.
12649 (vec_duplicate_p): Likewise.
12651 PR middle-end/83694
12652 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12653 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12655 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12658 * config/i386/i386-builtin.def
12659 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12660 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12661 Require also OPTION_MASK_ISA_AVX512F in addition to
12662 OPTION_MASK_ISA_GFNI.
12663 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12664 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12665 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12666 to OPTION_MASK_ISA_GFNI.
12667 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12668 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12669 OPTION_MASK_ISA_AVX512BW.
12670 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12671 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12672 addition to OPTION_MASK_ISA_GFNI.
12673 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12674 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12675 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12676 to OPTION_MASK_ISA_GFNI.
12677 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12678 a requirement for all ISAs rather than any of them with a few
12680 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12682 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12683 bitmasks to be enabled with 3 exceptions, instead of requiring any
12684 enabled ISA with lots of exceptions.
12685 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12686 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12687 Change avx512bw in isa attribute to avx512f.
12688 * config/i386/sgxintrin.h: Add license boilerplate.
12689 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12690 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12691 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12692 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12694 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12695 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12696 temporarily sse2 rather than sse if not enabled already.
12699 * config/i386/sse.md (VI248_VLBW): Rename to ...
12700 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12701 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12702 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12703 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12704 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12705 mode iterator instead of VI248_VLBW.
12707 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12709 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12710 (record_modified): Skip clobbers; add debug output.
12711 (param_change_prob): Use sreal frequencies.
12713 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12715 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12716 punt for user-aligned variables.
12718 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12720 * tree-chrec.c (chrec_contains_symbols): Return true for
12723 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12726 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12727 of (x|y) == x for BICS pattern.
12729 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12731 PR tree-optimization/83605
12732 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12733 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12736 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12738 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12739 * config/epiphany/rtems.h: New file.
12741 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12742 Uros Bizjak <ubizjak@gmail.com>
12745 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12746 QIreg_operand instead of register_operand predicate.
12747 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12748 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12749 comments instead of -fmitigate[-_]rop.
12751 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12754 * cgraphunit.c (symbol_table::compile): Switch to text_section
12755 before calling assembly_start debug hook.
12756 * run-rtl-passes.c (run_rtl_passes): Likewise.
12759 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12761 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12762 range_int_cst_p rather than !symbolic_range_p before calling
12763 extract_range_from_multiplicative_op_1.
12765 2018-01-04 Jeff Law <law@redhat.com>
12767 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12768 redundant test in assertion.
12770 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12772 * doc/rtl.texi: Document machine_mode wrapper classes.
12774 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12776 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12777 using tree_to_uhwi.
12779 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12781 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12782 the VEC_PERM_EXPR fold to fail.
12784 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12787 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12788 to switched_sections.
12790 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12793 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12794 test for d.testing.
12796 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12799 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12800 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12802 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12805 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12806 is BLKmode and bitpos not zero or mode change is needed.
12808 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12811 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12814 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12817 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12818 instead of MULT rtx. Update all corresponding splitters.
12819 (*saddl_se): Ditto.
12820 (*ssub<modesuffix>): Ditto.
12821 (*ssubl_se): Ditto.
12822 (*cmp_sadd_di): Update split patterns.
12823 (*cmp_sadd_si): Ditto.
12824 (*cmp_sadd_sidi): Ditto.
12825 (*cmp_ssub_di): Ditto.
12826 (*cmp_ssub_si): Ditto.
12827 (*cmp_ssub_sidi): Ditto.
12828 * config/alpha/predicates.md (const23_operand): New predicate.
12829 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12830 Look for ASHIFT, not MULT inner operand.
12831 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12833 2018-01-04 Martin Liska <mliska@suse.cz>
12835 PR gcov-profile/83669
12836 * gcov.c (output_intermediate_file): Add version to intermediate
12838 * doc/gcov.texi: Document new field 'version' in intermediate
12839 file format. Fix location of '-k' option of gcov command.
12841 2018-01-04 Martin Liska <mliska@suse.cz>
12844 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12846 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12848 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12850 2018-01-03 Martin Sebor <msebor@redhat.com>
12852 PR tree-optimization/83655
12853 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12854 checking calls with invalid arguments.
12856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12858 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12859 (vectorizable_mask_load_store): Delete.
12860 (vectorizable_call): Return false for masked loads and stores.
12861 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12862 instead of gimple_assign_rhs1.
12863 (vectorizable_load): Handle IFN_MASK_LOAD.
12864 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12868 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12870 (vectorizable_mask_load_store): ...here.
12871 (vectorizable_load): ...and here.
12873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12875 * tree-vect-stmts.c (vect_build_all_ones_mask)
12876 (vect_build_zero_merge_argument): New functions, split out from...
12877 (vectorizable_load): ...here.
12879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12881 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12883 (vectorizable_mask_load_store): ...here.
12884 (vectorizable_store): ...and here.
12886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12888 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12890 (vectorizable_mask_load_store): ...here.
12892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12894 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12895 (vect_model_store_cost): Take a vec_load_store_type instead of a
12897 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12898 (vect_model_store_cost): Take a vec_load_store_type instead of a
12900 (vectorizable_mask_load_store): Update accordingly.
12901 (vectorizable_store): Likewise.
12902 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12906 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12907 IFN_MASK_LOAD calls here rather than...
12908 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12911 Alan Hayward <alan.hayward@arm.com>
12912 David Sherwood <david.sherwood@arm.com>
12914 * expmed.c (extract_bit_field_1): For vector extracts,
12915 fall back to extract_bit_field_as_subreg if vec_extract
12918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12919 Alan Hayward <alan.hayward@arm.com>
12920 David Sherwood <david.sherwood@arm.com>
12922 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12923 they are variable or constant sized.
12924 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12925 slots for constant-sized data.
12927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12928 Alan Hayward <alan.hayward@arm.com>
12929 David Sherwood <david.sherwood@arm.com>
12931 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12932 handling COND_EXPRs with boolean comparisons, try to find a better
12933 basis for the mask type than the boolean itself.
12935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12937 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12938 is calculated and how it can be overridden.
12939 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12940 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12942 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12946 Alan Hayward <alan.hayward@arm.com>
12947 David Sherwood <david.sherwood@arm.com>
12949 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12950 Remove the mode argument.
12951 (aarch64_simd_valid_immediate): Remove the mode and inverse
12953 * config/aarch64/iterators.md (bitsize): New iterator.
12954 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12955 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12956 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12957 aarch64_simd_valid_immediate.
12958 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12959 (aarch64_reg_or_bic_imm): Likewise.
12960 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12961 with an insn_type enum and msl with a modifier_type enum.
12962 Replace element_width with a scalar_mode. Change the shift
12963 to unsigned int. Add constructors for scalar_float_mode and
12964 scalar_int_mode elements.
12965 (aarch64_vect_float_const_representable_p): Delete.
12966 (aarch64_can_const_movi_rtx_p)
12967 (aarch64_simd_scalar_immediate_valid_for_move)
12968 (aarch64_simd_make_constant): Update call to
12969 aarch64_simd_valid_immediate.
12970 (aarch64_advsimd_valid_immediate_hs): New function.
12971 (aarch64_advsimd_valid_immediate): Likewise.
12972 (aarch64_simd_valid_immediate): Remove mode and inverse
12973 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12974 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12975 and aarch64_float_const_representable_p on the result.
12976 (aarch64_output_simd_mov_immediate): Remove mode argument.
12977 Update call to aarch64_simd_valid_immediate and use of
12978 simd_immediate_info.
12979 (aarch64_output_scalar_simd_mov_immediate): Update call
12982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12983 Alan Hayward <alan.hayward@arm.com>
12984 David Sherwood <david.sherwood@arm.com>
12986 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12987 (mode_nunits): Likewise CONST_MODE_NUNITS.
12988 * machmode.def (ADJUST_NUNITS): Document.
12989 * genmodes.c (mode_data::need_nunits_adj): New field.
12990 (blank_mode): Update accordingly.
12991 (adj_nunits): New variable.
12992 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12994 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12995 listed in adj_nunits.
12996 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12997 listed in adj_nunits. Don't emit case statements for such modes.
12998 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12999 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13000 nothing if adj_nunits is nonnull.
13001 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13002 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13003 (emit_mode_fbit): Update use of print_maybe_const_decl.
13004 (emit_move_size): Likewise. Treat the array as non-const
13006 (emit_mode_adjustments): Handle adj_nunits.
13008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13010 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13011 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13012 (VECTOR_MODES): Use it.
13013 (make_vector_modes): Take the prefix as an argument.
13015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13016 Alan Hayward <alan.hayward@arm.com>
13017 David Sherwood <david.sherwood@arm.com>
13019 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13020 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13021 for MODE_VECTOR_BOOL.
13022 * machmode.def (VECTOR_BOOL_MODE): Document.
13023 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13024 (make_vector_bool_mode): New function.
13025 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13027 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13028 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13030 * stor-layout.c (int_mode_for_mode): Likewise.
13031 * tree.c (build_vector_type_for_mode): Likewise.
13032 * varasm.c (output_constant_pool_2): Likewise.
13033 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13034 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13035 for MODE_VECTOR_BOOL.
13036 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13037 of mode class checks.
13038 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13039 instead of a list of mode class checks.
13040 (expand_vector_scalar_condition): Likewise.
13041 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13044 Alan Hayward <alan.hayward@arm.com>
13045 David Sherwood <david.sherwood@arm.com>
13047 * machmode.h (mode_size): Change from unsigned short to
13049 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13050 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13051 or if measurement_type is not polynomial.
13052 (fixed_size_mode::includes_p): Check for constant-sized modes.
13053 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13054 return a poly_uint16 rather than an unsigned short.
13055 (emit_mode_size): Change the type of mode_size from unsigned short
13056 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13057 (emit_mode_adjustments): Cope with polynomial vector sizes.
13058 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13060 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13062 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13063 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13064 * caller-save.c (setup_save_areas): Likewise.
13065 (replace_reg_with_saved_mem): Likewise.
13066 * calls.c (emit_library_call_value_1): Likewise.
13067 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13068 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13069 (gen_lowpart_for_combine): Likewise.
13070 * convert.c (convert_to_integer_1): Likewise.
13071 * cse.c (equiv_constant, cse_insn): Likewise.
13072 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13073 (cselib_subst_to_values): Likewise.
13074 * dce.c (word_dce_process_block): Likewise.
13075 * df-problems.c (df_word_lr_mark_ref): Likewise.
13076 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13077 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13078 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13079 (rtl_for_decl_location): Likewise.
13080 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13081 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13082 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13083 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13084 (expand_expr_real_1): Likewise.
13085 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13086 (pad_below): Likewise.
13087 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13088 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13089 * ira.c (get_subreg_tracking_sizes): Likewise.
13090 * ira-build.c (ira_create_allocno_objects): Likewise.
13091 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13092 (ira_sort_regnos_for_alter_reg): Likewise.
13093 * ira-costs.c (record_operand_costs): Likewise.
13094 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13095 (resolve_simple_move): Likewise.
13096 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13097 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13098 (lra_constraints): Likewise.
13099 (CONST_POOL_OK_P): Reject variable-sized modes.
13100 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13101 (add_pseudo_to_slot, lra_spill): Likewise.
13102 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13103 * optabs-query.c (get_best_extraction_insn): Likewise.
13104 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13105 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13106 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13107 * recog.c (offsettable_address_addr_space_p): Likewise.
13108 * regcprop.c (maybe_mode_change): Likewise.
13109 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13110 * regrename.c (build_def_use): Likewise.
13111 * regstat.c (dump_reg_info): Likewise.
13112 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13113 (find_reloads, find_reloads_subreg_address): Likewise.
13114 * reload1.c (eliminate_regs_1): Likewise.
13115 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13116 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13117 (simplify_binary_operation_1, simplify_subreg): Likewise.
13118 * targhooks.c (default_function_arg_padding): Likewise.
13119 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13120 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13121 (verify_gimple_assign_ternary): Likewise.
13122 * tree-inline.c (estimate_move_cost): Likewise.
13123 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13124 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13125 (get_address_cost_ainc): Likewise.
13126 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13127 (vect_supportable_dr_alignment): Likewise.
13128 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13129 (vectorizable_reduction): Likewise.
13130 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13131 (vectorizable_operation, vectorizable_load): Likewise.
13132 * tree.c (build_same_sized_truth_vector_type): Likewise.
13133 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13134 * var-tracking.c (emit_note_insn_var_location): Likewise.
13135 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13136 (ADDR_VEC_ALIGN): Likewise.
13138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13139 Alan Hayward <alan.hayward@arm.com>
13140 David Sherwood <david.sherwood@arm.com>
13142 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13144 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13145 or if measurement_type is polynomial.
13146 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13147 * combine.c (make_extraction): Likewise.
13148 * dse.c (find_shift_sequence): Likewise.
13149 * dwarf2out.c (mem_loc_descriptor): Likewise.
13150 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13151 (extract_bit_field, extract_low_bits): Likewise.
13152 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13153 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13154 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13155 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13156 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13157 * reload.c (find_reloads): Likewise.
13158 * reload1.c (alter_reg): Likewise.
13159 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13160 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13161 * tree-if-conv.c (predicate_mem_writes): Likewise.
13162 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13163 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13164 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13165 * valtrack.c (dead_debug_insert_temp): Likewise.
13166 * varasm.c (mergeable_constant_section): Likewise.
13167 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13170 Alan Hayward <alan.hayward@arm.com>
13171 David Sherwood <david.sherwood@arm.com>
13173 * expr.c (expand_assignment): Cope with polynomial mode sizes
13174 when assigning to a CONCAT.
13176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13177 Alan Hayward <alan.hayward@arm.com>
13178 David Sherwood <david.sherwood@arm.com>
13180 * machmode.h (mode_precision): Change from unsigned short to
13182 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13184 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13185 or if measurement_type is not polynomial.
13186 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13187 in which the mode is already known to be a scalar_int_mode.
13188 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13189 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13191 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13192 for GET_MODE_PRECISION.
13193 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13194 for GET_MODE_PRECISION.
13195 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13197 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13198 (expand_field_assignment, make_extraction): Likewise.
13199 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13200 (get_last_value): Likewise.
13201 * convert.c (convert_to_integer_1): Likewise.
13202 * cse.c (cse_insn): Likewise.
13203 * expr.c (expand_expr_real_1): Likewise.
13204 * lra-constraints.c (simplify_operand_subreg): Likewise.
13205 * optabs-query.c (can_atomic_load_p): Likewise.
13206 * optabs.c (expand_atomic_load): Likewise.
13207 (expand_atomic_store): Likewise.
13208 * ree.c (combine_reaching_defs): Likewise.
13209 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13210 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13211 * tree.h (type_has_mode_precision_p): Likewise.
13212 * ubsan.c (instrument_si_overflow): Likewise.
13214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13215 Alan Hayward <alan.hayward@arm.com>
13216 David Sherwood <david.sherwood@arm.com>
13218 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13219 polynomial numbers of units.
13220 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13221 (valid_vector_subparts_p): New function.
13222 (build_vector_type): Remove temporary shim and take the number
13223 of units as a poly_uint64 rather than an int.
13224 (build_opaque_vector_type): Take the number of units as a
13225 poly_uint64 rather than an int.
13226 * tree.c (build_vector_from_ctor): Handle polynomial
13227 TYPE_VECTOR_SUBPARTS.
13228 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13229 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13230 (build_vector_from_val): If the number of units is variable,
13231 use build_vec_duplicate_cst for constant operands and
13232 VEC_DUPLICATE_EXPR otherwise.
13233 (make_vector_type): Remove temporary is_constant ().
13234 (build_vector_type, build_opaque_vector_type): Take the number of
13235 units as a poly_uint64 rather than an int.
13236 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13238 * cfgexpand.c (expand_debug_expr): Likewise.
13239 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13240 (store_constructor, expand_expr_real_1): Likewise.
13241 (const_scalar_mask_from_tree): Likewise.
13242 * fold-const-call.c (fold_const_reduction): Likewise.
13243 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13244 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13245 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13246 (fold_relational_const): Likewise.
13247 (native_interpret_vector): Likewise. Change the size from an
13248 int to an unsigned int.
13249 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13250 TYPE_VECTOR_SUBPARTS.
13251 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13252 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13253 duplicating a non-constant operand into a variable-length vector.
13254 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13255 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13256 * ipa-icf.c (sem_variable::equals): Likewise.
13257 * match.pd: Likewise.
13258 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13259 * print-tree.c (print_node): Likewise.
13260 * stor-layout.c (layout_type): Likewise.
13261 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13262 * tree-cfg.c (verify_gimple_comparison): Likewise.
13263 (verify_gimple_assign_binary): Likewise.
13264 (verify_gimple_assign_ternary): Likewise.
13265 (verify_gimple_assign_single): Likewise.
13266 * tree-pretty-print.c (dump_generic_node): Likewise.
13267 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13268 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13269 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13270 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13271 (vect_shift_permute_load_chain): Likewise.
13272 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13273 (expand_vector_condition, optimize_vector_constructor): Likewise.
13274 (lower_vec_perm, get_compute_type): Likewise.
13275 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13276 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13277 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13278 (vect_recog_mask_conversion_pattern): Likewise.
13279 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13280 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13281 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13282 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13283 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13284 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13285 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13286 (supportable_widening_operation): Likewise.
13287 (supportable_narrowing_operation): Likewise.
13288 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13290 * varasm.c (output_constant): Likewise.
13292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13293 Alan Hayward <alan.hayward@arm.com>
13294 David Sherwood <david.sherwood@arm.com>
13296 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13297 so that both the length == 3 and length != 3 cases set up their
13298 own permute vectors. Add comments explaining why we know the
13299 number of elements is constant.
13300 (vect_permute_load_chain): Likewise.
13302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13303 Alan Hayward <alan.hayward@arm.com>
13304 David Sherwood <david.sherwood@arm.com>
13306 * machmode.h (mode_nunits): Change from unsigned char to
13308 (ONLY_FIXED_SIZE_MODES): New macro.
13309 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13310 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13311 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13313 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13314 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13315 or if measurement_type is not polynomial.
13316 * genmodes.c (ZERO_COEFFS): New macro.
13317 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13319 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13320 Use ZERO_COEFFS when emitting initializers.
13321 * data-streamer.h (bp_pack_poly_value): New function.
13322 (bp_unpack_poly_value): Likewise.
13323 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13324 for GET_MODE_NUNITS.
13325 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13326 for GET_MODE_NUNITS.
13327 * tree.c (make_vector_type): Remove temporary shim and make
13328 the real function take the number of units as a poly_uint64
13329 rather than an int.
13330 (build_vector_type_for_mode): Handle polynomial nunits.
13331 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13332 * emit-rtl.c (const_vec_series_p_1): Likewise.
13333 (gen_rtx_CONST_VECTOR): Likewise.
13334 * fold-const.c (test_vec_duplicate_folding): Likewise.
13335 * genrecog.c (validate_pattern): Likewise.
13336 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13337 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13338 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13339 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13340 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13341 * rtlanal.c (subreg_get_info): Likewise.
13342 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13343 (vect_grouped_load_supported): Likewise.
13344 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13345 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13346 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13347 (simplify_const_unary_operation, simplify_binary_operation_1)
13348 (simplify_const_binary_operation, simplify_ternary_operation)
13349 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13350 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13351 instead of CONST_VECTOR_NUNITS.
13352 * varasm.c (output_constant_pool_2): Likewise.
13353 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13354 explicit-encoded elements in the XVEC for variable-length vectors.
13356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13358 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13361 Alan Hayward <alan.hayward@arm.com>
13362 David Sherwood <david.sherwood@arm.com>
13364 * coretypes.h (fixed_size_mode): Declare.
13365 (fixed_size_mode_pod): New typedef.
13366 * builtins.h (target_builtins::x_apply_args_mode)
13367 (target_builtins::x_apply_result_mode): Change type to
13368 fixed_size_mode_pod.
13369 * builtins.c (apply_args_size, apply_result_size, result_vector)
13370 (expand_builtin_apply_args_1, expand_builtin_apply)
13371 (expand_builtin_return): Update accordingly.
13373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13375 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13376 * cselib.c (cselib_hash_rtx): Likewise.
13377 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13378 CONST_VECTOR encoding.
13380 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13381 Jeff Law <law@redhat.com>
13384 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13385 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13386 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13387 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13390 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13391 explicitly probe *sp in a noreturn function if there were any callee
13392 register saves or frame pointer is needed.
13394 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13397 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13398 BLKmode for ternary, binary or unary expressions.
13401 * var-tracking.c (delete_vta_debug_insn): New inline function.
13402 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13403 insns from get_insns () to NULL instead of each bb separately.
13404 Use delete_vta_debug_insn. No longer static.
13405 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13406 delete_vta_debug_insns callers.
13407 * rtl.h (delete_vta_debug_insns): Declare.
13408 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13409 instead of variable_tracking_main.
13411 2018-01-03 Martin Sebor <msebor@redhat.com>
13413 PR tree-optimization/83603
13414 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13415 arguments past the endof the argument list in functions declared
13416 without a prototype.
13417 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13418 Avoid checking when arguments are null.
13420 2018-01-03 Martin Sebor <msebor@redhat.com>
13423 * doc/extend.texi (attribute const): Fix a typo.
13424 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13425 issuing -Wsuggest-attribute for void functions.
13427 2018-01-03 Martin Sebor <msebor@redhat.com>
13429 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13430 offset_int::from instead of wide_int::to_shwi.
13431 (maybe_diag_overlap): Remove assertion.
13432 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13433 * gimple-ssa-sprintf.c (format_directive): Same.
13434 (parse_directive): Same.
13435 (sprintf_dom_walker::compute_format_length): Same.
13436 (try_substitute_return_value): Same.
13438 2018-01-03 Jeff Law <law@redhat.com>
13440 PR middle-end/83654
13441 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13442 non-constant residual for zero at runtime and avoid probing in
13443 that case. Reorganize code for trailing problem to mirror handling
13446 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13448 PR tree-optimization/83501
13449 * tree-ssa-strlen.c (get_string_cst): New.
13450 (handle_char_store): Call get_string_cst.
13452 2018-01-03 Martin Liska <mliska@suse.cz>
13454 PR tree-optimization/83593
13455 * tree-ssa-strlen.c: Include tree-cfg.h.
13456 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13457 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13458 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13460 (strlen_dom_walker::before_dom_children): Call
13461 gimple_purge_dead_eh_edges. Dump tranformation with details
13463 (strlen_dom_walker::before_dom_children): Update call by adding
13464 new argument cleanup_eh.
13465 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13467 2018-01-03 Martin Liska <mliska@suse.cz>
13470 * cif-code.def (VARIADIC_THUNK): New enum value.
13471 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13474 2018-01-03 Jan Beulich <jbeulich@suse.com>
13476 * sse.md (mov<mode>_internal): Tighten condition for when to use
13477 vmovdqu<ssescalarsize> for TI and OI modes.
13479 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13481 Update copyright years.
13483 2018-01-03 Martin Liska <mliska@suse.cz>
13486 * ipa-visibility.c (function_and_variable_visibility): Skip
13487 functions with noipa attribure.
13489 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13491 * gcc.c (process_command): Update copyright notice dates.
13492 * gcov-dump.c (print_version): Ditto.
13493 * gcov.c (print_version): Ditto.
13494 * gcov-tool.c (print_version): Ditto.
13495 * gengtype.c (create_file): Ditto.
13496 * doc/cpp.texi: Bump @copying's copyright year.
13497 * doc/cppinternals.texi: Ditto.
13498 * doc/gcc.texi: Ditto.
13499 * doc/gccint.texi: Ditto.
13500 * doc/gcov.texi: Ditto.
13501 * doc/install.texi: Ditto.
13502 * doc/invoke.texi: Ditto.
13504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13506 * vector-builder.h (vector_builder::m_full_nelts): Change from
13507 unsigned int to poly_uint64.
13508 (vector_builder::full_nelts): Update prototype accordingly.
13509 (vector_builder::new_vector): Likewise.
13510 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13511 (vector_builder::operator ==): Likewise.
13512 (vector_builder::finalize): Likewise.
13513 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13514 Take the number of elements as a poly_uint64 rather than an
13516 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13517 from unsigned int to poly_uint64.
13518 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13519 (vec_perm_indices::new_vector): Likewise.
13520 (vec_perm_indices::length): Likewise.
13521 (vec_perm_indices::nelts_per_input): Likewise.
13522 (vec_perm_indices::input_nelts): Likewise.
13523 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13524 number of elements per input as a poly_uint64 rather than an
13525 unsigned int. Use the original encoding for variable-length
13526 vectors, rather than clamping each individual element.
13527 For the second and subsequent elements in each pattern,
13528 clamp the step and base before clamping their sum.
13529 (vec_perm_indices::series_p): Handle polynomial element counts.
13530 (vec_perm_indices::all_in_range_p): Likewise.
13531 (vec_perm_indices_to_tree): Likewise.
13532 (vec_perm_indices_to_rtx): Likewise.
13533 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13534 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13535 (tree_vector_builder::new_binary_operation): Handle polynomial
13536 element counts. Return false if we need to know the number
13537 of elements at compile time.
13538 * fold-const.c (fold_vec_perm): Punt if the number of elements
13539 isn't known at compile time.
13541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13543 * vec-perm-indices.h (vec_perm_builder): Change element type
13544 from HOST_WIDE_INT to poly_int64.
13545 (vec_perm_indices::element_type): Update accordingly.
13546 (vec_perm_indices::clamp): Handle polynomial element_types.
13547 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13548 (vec_perm_indices::all_in_range_p): Likewise.
13549 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13551 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13552 polynomial vec_perm_indices element types.
13553 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13554 * fold-const.c (fold_vec_perm): Likewise.
13555 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13556 * tree-vect-generic.c (lower_vec_perm): Likewise.
13557 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13558 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13559 element type to HOST_WIDE_INT.
13561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13562 Alan Hayward <alan.hayward@arm.com>
13563 David Sherwood <david.sherwood@arm.com>
13565 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13566 rather than an int. Use plus_constant.
13567 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13568 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13571 Alan Hayward <alan.hayward@arm.com>
13572 David Sherwood <david.sherwood@arm.com>
13574 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13575 a HOST_WIDE_INT to a poly_int64.
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 * calls.c (load_register_parameters): Cope with polynomial
13582 mode sizes. Require a constant size for BLKmode parameters
13583 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13584 forces a parameter to be padded at the lsb end in order to
13585 fill a complete number of words, require the parameter size
13586 to be ordered wrt UNITS_PER_WORD.
13588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13589 Alan Hayward <alan.hayward@arm.com>
13590 David Sherwood <david.sherwood@arm.com>
13592 * reload1.c (spill_stack_slot_width): Change element type
13593 from unsigned int to poly_uint64_pod.
13594 (alter_reg): Treat mode sizes as polynomial.
13596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13597 Alan Hayward <alan.hayward@arm.com>
13598 David Sherwood <david.sherwood@arm.com>
13600 * reload.c (complex_word_subreg_p): New function.
13601 (reload_inner_reg_of_subreg, push_reload): Use it.
13603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13604 Alan Hayward <alan.hayward@arm.com>
13605 David Sherwood <david.sherwood@arm.com>
13607 * lra-constraints.c (process_alt_operands): Reject matched
13608 operands whose sizes aren't ordered.
13609 (match_reload): Refer to this check here.
13611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13612 Alan Hayward <alan.hayward@arm.com>
13613 David Sherwood <david.sherwood@arm.com>
13615 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13616 that the mode size is in the set {1, 2, 4, 8, 16}.
13618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13619 Alan Hayward <alan.hayward@arm.com>
13620 David Sherwood <david.sherwood@arm.com>
13622 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13623 Use plus_constant instead of gen_rtx_PLUS.
13625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13626 Alan Hayward <alan.hayward@arm.com>
13627 David Sherwood <david.sherwood@arm.com>
13629 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13630 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13631 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13632 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13633 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13634 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13635 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13636 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13637 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13638 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13640 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13641 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13642 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13643 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13644 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13645 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13646 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13647 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13648 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13649 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13651 * expr.c (emit_move_resolve_push): Treat the input and result
13652 of PUSH_ROUNDING as a poly_int64.
13653 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13654 (emit_push_insn): Likewise.
13655 * lra-eliminations.c (mark_not_eliminable): Likewise.
13656 * recog.c (push_operand): Likewise.
13657 * reload1.c (elimination_effects): Likewise.
13658 * rtlanal.c (nonzero_bits1): Likewise.
13659 * calls.c (store_one_arg): Likewise. Require the padding to be
13660 known at compile time.
13662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13663 Alan Hayward <alan.hayward@arm.com>
13664 David Sherwood <david.sherwood@arm.com>
13666 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13667 Use plus_constant instead of gen_rtx_PLUS.
13669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13670 Alan Hayward <alan.hayward@arm.com>
13671 David Sherwood <david.sherwood@arm.com>
13673 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13674 rather than an int.
13676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13677 Alan Hayward <alan.hayward@arm.com>
13678 David Sherwood <david.sherwood@arm.com>
13680 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13681 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13682 via stack temporaries. Treat the mode size as polynomial too.
13684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13685 Alan Hayward <alan.hayward@arm.com>
13686 David Sherwood <david.sherwood@arm.com>
13688 * expr.c (expand_expr_real_2): When handling conversions involving
13689 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13690 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13691 as a poly_uint64 too.
13693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13694 Alan Hayward <alan.hayward@arm.com>
13695 David Sherwood <david.sherwood@arm.com>
13697 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
13704 REGMODE_NATURAL_SIZE.
13705 * expmed.c (store_bit_field_1): Likewise.
13706 * expr.c (store_constructor): Likewise.
13707 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13708 and polynomial REGMODE_NATURAL_SIZE.
13709 (gen_lowpart_common): Likewise.
13710 * reginfo.c (record_subregs_of_mode): Likewise.
13711 * rtlanal.c (read_modify_subreg_p): Likewise.
13713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13714 Alan Hayward <alan.hayward@arm.com>
13715 David Sherwood <david.sherwood@arm.com>
13717 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13718 numbers of elements.
13720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13721 Alan Hayward <alan.hayward@arm.com>
13722 David Sherwood <david.sherwood@arm.com>
13724 * match.pd: Cope with polynomial numbers of vector elements.
13726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13727 Alan Hayward <alan.hayward@arm.com>
13728 David Sherwood <david.sherwood@arm.com>
13730 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13731 in a POINTER_PLUS_EXPR.
13733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13734 Alan Hayward <alan.hayward@arm.com>
13735 David Sherwood <david.sherwood@arm.com>
13737 * omp-simd-clone.c (simd_clone_subparts): New function.
13738 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13739 (ipa_simd_modify_function_body): Likewise.
13741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13742 Alan Hayward <alan.hayward@arm.com>
13743 David Sherwood <david.sherwood@arm.com>
13745 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13746 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13747 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13748 (expand_vector_condition, vector_element): Likewise.
13749 (subparts_gt): New function.
13750 (get_compute_type): Use subparts_gt.
13751 (count_type_subparts): Delete.
13752 (expand_vector_operations_1): Use subparts_gt instead of
13753 count_type_subparts.
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-data-refs.c (vect_no_alias_p): Replace with...
13760 (vect_compile_time_alias): ...this new function. Do the calculation
13761 on poly_ints rather than trees.
13762 (vect_prune_runtime_alias_test_list): Update call accordingly.
13764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13765 Alan Hayward <alan.hayward@arm.com>
13766 David Sherwood <david.sherwood@arm.com>
13768 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13770 (vect_schedule_slp_instance): Likewise.
13772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13773 Alan Hayward <alan.hayward@arm.com>
13774 David Sherwood <david.sherwood@arm.com>
13776 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13777 constant and extern definitions for variable-length vectors.
13778 (vect_get_constant_vectors): Note that the number of units
13779 is known to be constant.
13781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13782 Alan Hayward <alan.hayward@arm.com>
13783 David Sherwood <david.sherwood@arm.com>
13785 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13786 of units as polynomial. Choose between WIDE and NARROW based
13789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13790 Alan Hayward <alan.hayward@arm.com>
13791 David Sherwood <david.sherwood@arm.com>
13793 * tree-vect-stmts.c (simd_clone_subparts): New function.
13794 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13797 Alan Hayward <alan.hayward@arm.com>
13798 David Sherwood <david.sherwood@arm.com>
13800 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13801 vectors as polynomial. Use build_index_vector for
13802 IFN_GOMP_SIMD_LANE.
13804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13805 Alan Hayward <alan.hayward@arm.com>
13806 David Sherwood <david.sherwood@arm.com>
13808 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13809 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13810 for variable-length vectors.
13811 (vectorizable_mask_load_store): Treat the number of units as
13812 polynomial, asserting that it is constant if the condition has
13813 already been enforced.
13814 (vectorizable_store, vectorizable_load): Likewise.
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-loop.c (vectorizable_live_operation): Treat the number
13821 of units as polynomial. Punt if we can't tell at compile time
13822 which vector contains the final result.
13824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13825 Alan Hayward <alan.hayward@arm.com>
13826 David Sherwood <david.sherwood@arm.com>
13828 * tree-vect-loop.c (vectorizable_induction): Treat the number
13829 of units as polynomial. Punt on SLP inductions. Use an integer
13830 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13831 cast of such a series for variable-length floating-point
13834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13835 Alan Hayward <alan.hayward@arm.com>
13836 David Sherwood <david.sherwood@arm.com>
13838 * tree.h (build_index_vector): Declare.
13839 * tree.c (build_index_vector): New function.
13840 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13841 of units as polynomial, forcibly converting it to a constant if
13842 vectorizable_reduction has already enforced the condition.
13843 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13844 to create a {1,2,3,...} vector.
13845 (vectorizable_reduction): Treat the number of units as polynomial.
13846 Choose vectype_in based on the largest scalar element size rather
13847 than the smallest number of units. Enforce the restrictions
13850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13851 Alan Hayward <alan.hayward@arm.com>
13852 David Sherwood <david.sherwood@arm.com>
13854 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13855 number of units as polynomial.
13857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13858 Alan Hayward <alan.hayward@arm.com>
13859 David Sherwood <david.sherwood@arm.com>
13861 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13862 * target.def (autovectorize_vector_sizes): Return the vector sizes
13863 by pointer, using vector_sizes rather than a bitmask.
13864 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13865 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13866 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13868 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13869 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13870 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13871 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13872 * omp-general.c (omp_max_vf): Likewise.
13873 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13874 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13875 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13876 * tree-vect-slp.c (vect_slp_bb): Likewise.
13877 * doc/tm.texi: Regenerate.
13878 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13880 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13881 the vector size as a poly_uint64 rather than an unsigned int.
13882 (current_vector_size): Change from an unsigned int to a poly_uint64.
13883 (get_vectype_for_scalar_type): Update accordingly.
13884 * tree.h (build_truth_vector_type): Take the size and number of
13885 units as a poly_uint64 rather than an unsigned int.
13886 (build_vector_type): Add a temporary overload that takes
13887 the number of units as a poly_uint64 rather than an unsigned int.
13888 * tree.c (make_vector_type): Likewise.
13889 (build_truth_vector_type): Take the number of units as a poly_uint64
13890 rather than an unsigned int.
13892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13893 Alan Hayward <alan.hayward@arm.com>
13894 David Sherwood <david.sherwood@arm.com>
13896 * target.def (get_mask_mode): Take the number of units and length
13897 as poly_uint64s rather than unsigned ints.
13898 * targhooks.h (default_get_mask_mode): Update accordingly.
13899 * targhooks.c (default_get_mask_mode): Likewise.
13900 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13901 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13908 * omp-general.c (omp_max_vf): Likewise.
13909 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13910 (expand_omp_simd): Handle polynomial safelen.
13911 * omp-low.c (omplow_simd_context): Add a default constructor.
13912 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13913 (lower_rec_simd_input_clauses): Update accordingly.
13914 (lower_rec_input_clauses): Likewise.
13916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13917 Alan Hayward <alan.hayward@arm.com>
13918 David Sherwood <david.sherwood@arm.com>
13920 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13921 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13922 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13923 (vect_analyze_slp_cost): Likewise.
13924 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13925 (vect_model_load_cost): Likewise.
13927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13928 Alan Hayward <alan.hayward@arm.com>
13929 David Sherwood <david.sherwood@arm.com>
13931 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13932 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13933 from an unsigned int * to a poly_uint64_pod *.
13934 (calculate_unrolling_factor): New function.
13935 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13938 Alan Hayward <alan.hayward@arm.com>
13939 David Sherwood <david.sherwood@arm.com>
13941 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13942 from an unsigned int to a poly_uint64.
13943 (_loop_vec_info::slp_unrolling_factor): Likewise.
13944 (_loop_vec_info::vectorization_factor): Change from an int
13946 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13947 (vect_get_num_vectors): New function.
13948 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13949 (vect_get_num_copies): Use vect_get_num_vectors.
13950 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13951 to an unsigned int *.
13952 (vect_analyze_data_refs): Change min_vf from an int * to a
13954 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13955 than an unsigned HOST_WIDE_INT.
13956 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13957 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13958 to an unsigned int *.
13959 (vect_analyze_data_ref_dependences): Likewise.
13960 (vect_compute_data_ref_alignment): Handle polynomial vf.
13961 (vect_enhance_data_refs_alignment): Likewise.
13962 (vect_prune_runtime_alias_test_list): Likewise.
13963 (vect_shift_permute_load_chain): Likewise.
13964 (vect_supportable_dr_alignment): Likewise.
13965 (dependence_distance_ge_vf): Take the vectorization factor as a
13966 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13967 (vect_analyze_data_refs): Change min_vf from an int * to a
13969 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13970 vfm1 as a poly_uint64 rather than an int. Make the same change
13971 for the returned bound_scalar.
13972 (vect_gen_vector_loop_niters): Handle polynomial vf.
13973 (vect_do_peeling): Likewise. Update call to
13974 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13975 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13977 * tree-vect-loop.c (vect_determine_vectorization_factor)
13978 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13979 (vect_get_known_peeling_cost): Likewise.
13980 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13981 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13982 (vect_transform_loop): Likewise. Use the lowest possible VF when
13983 updating the upper bounds of the loop.
13984 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13985 rather than an int.
13986 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13987 polynomial unroll factors.
13988 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13989 (vect_make_slp_decision): Likewise.
13990 (vect_supported_load_permutation_p): Likewise, and polynomial
13992 (vect_analyze_slp_cost): Handle polynomial vf.
13993 (vect_slp_analyze_node_operations): Likewise.
13994 (vect_slp_analyze_bb_1): Likewise.
13995 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13996 than an unsigned HOST_WIDE_INT.
13997 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13998 (vectorizable_load): Handle polynomial vf.
13999 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14001 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14004 Alan Hayward <alan.hayward@arm.com>
14005 David Sherwood <david.sherwood@arm.com>
14007 * match.pd: Handle bit operations involving three constants
14008 and try to fold one pair.
14010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14012 * tree-vect-loop-manip.c: Include gimple-fold.h.
14013 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14014 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14015 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14016 Add a path that uses a step of VF instead of 1, but disable it
14018 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14019 and niters_no_overflow parameters. Update calls to
14020 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14021 Create a new SSA name if the latter choses to use a ste other
14022 than zero, and return it via niters_vector_mult_vf_var.
14023 * tree-vect-loop.c (vect_transform_loop): Update calls to
14024 vect_do_peeling, vect_gen_vector_loop_niters and
14025 slpeel_make_loop_iterate_ntimes.
14026 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14027 (vect_gen_vector_loop_niters): Update declarations after above changes.
14029 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14031 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14032 128-bit round to integer instructions.
14033 (ceil<mode>2): Likewise.
14034 (btrunc<mode>2): Likewise.
14035 (round<mode>2): Likewise.
14037 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14039 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14040 unaligned VSX load/store on P8/P9.
14041 (expand_block_clear): Allow the use of unaligned VSX
14042 load/store on P8/P9.
14044 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14046 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14048 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14049 swap associated with both a load and a store.
14051 2018-01-02 Andrew Waterman <andrew@sifive.com>
14053 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14054 * config/riscv/riscv.md (clear_cache): Use it.
14056 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14058 * web.c: Remove out-of-date comment.
14060 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14062 * expr.c (fixup_args_size_notes): Check that any existing
14063 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14064 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14065 (emit_single_push_insn): ...here.
14067 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14069 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14070 (const_vector_encoded_nelts): New function.
14071 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14072 (const_vector_int_elt, const_vector_elt): Declare.
14073 * emit-rtl.c (const_vector_int_elt_1): New function.
14074 (const_vector_elt): Likewise.
14075 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14076 of CONST_VECTOR_ELT.
14078 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14080 * expr.c: Include rtx-vector-builder.h.
14081 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14082 directly on the tree encoding.
14083 (const_vector_from_tree): Likewise.
14084 * optabs.c: Include rtx-vector-builder.h.
14085 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14086 sequence of "u" values.
14087 * vec-perm-indices.c: Include rtx-vector-builder.h.
14088 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14089 directly on the vec_perm_indices encoding.
14091 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14093 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14094 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14095 * rtx-vector-builder.h: New file.
14096 * rtx-vector-builder.c: Likewise.
14097 * rtl.h (rtx_def::u2): Add a const_vector field.
14098 (CONST_VECTOR_NPATTERNS): New macro.
14099 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14100 (CONST_VECTOR_DUPLICATE_P): Likewise.
14101 (CONST_VECTOR_STEPPED_P): Likewise.
14102 (CONST_VECTOR_ENCODED_ELT): Likewise.
14103 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14104 (unwrap_const_vec_duplicate): Likewise.
14105 (const_vec_series_p): Check for a non-duplicated vector encoding.
14106 Say that the function only returns true for integer vectors.
14107 * emit-rtl.c: Include rtx-vector-builder.h.
14108 (gen_const_vec_duplicate_1): Delete.
14109 (gen_const_vector): Call gen_const_vec_duplicate instead of
14110 gen_const_vec_duplicate_1.
14111 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14112 (gen_const_vec_duplicate): Use rtx_vector_builder.
14113 (gen_const_vec_series): Likewise.
14114 (gen_rtx_CONST_VECTOR): Likewise.
14115 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14116 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14117 Build a new vector rather than modifying a CONST_VECTOR in-place.
14118 (handle_special_swappables): Update call accordingly.
14119 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14120 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14121 Build a new vector rather than modifying a CONST_VECTOR in-place.
14122 (handle_special_swappables): Update call accordingly.
14124 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14126 * simplify-rtx.c (simplify_const_binary_operation): Use
14127 CONST_VECTOR_ELT instead of XVECEXP.
14129 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14131 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14132 the selector elements to be different from the data elements
14133 if the selector is a VECTOR_CST.
14134 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14135 ssizetype for the selector.
14137 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14139 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14140 before testing each element individually.
14141 * tree-vect-generic.c (lower_vec_perm): Likewise.
14143 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14145 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14146 * selftest-run-tests.c (selftest::run_tests): Call it.
14147 * vector-builder.h (vector_builder::operator ==): New function.
14148 (vector_builder::operator !=): Likewise.
14149 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14150 (vec_perm_indices::all_from_input_p): New function.
14151 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14152 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14153 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14154 instead of reading the VECTOR_CST directly. Detect whether both
14155 vector inputs are the same before constructing the vec_perm_indices,
14156 and update the number of inputs argument accordingly. Use the
14157 utility functions added above. Only construct sel2 if we need to.
14159 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14161 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14162 the broadcast of the low byte.
14163 (expand_mult_highpart): Use an explicit encoding for the permutes.
14164 * optabs-query.c (can_mult_highpart_p): Likewise.
14165 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14166 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14167 (vectorizable_bswap): Likewise.
14168 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14169 explicit encoding for the power-of-2 permutes.
14170 (vect_permute_store_chain): Likewise.
14171 (vect_grouped_load_supported): Likewise.
14172 (vect_permute_load_chain): Likewise.
14174 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14176 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14177 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14178 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14179 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14180 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14181 (vect_gen_perm_mask_any): Likewise.
14183 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14185 * int-vector-builder.h: New file.
14186 * vec-perm-indices.h: Include int-vector-builder.h.
14187 (vec_perm_indices): Redefine as an int_vector_builder.
14188 (auto_vec_perm_indices): Delete.
14189 (vec_perm_builder): Redefine as a stand-alone class.
14190 (vec_perm_indices::vec_perm_indices): New function.
14191 (vec_perm_indices::clamp): Likewise.
14192 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14193 (vec_perm_indices::new_vector): New function.
14194 (vec_perm_indices::new_expanded_vector): Update for new
14195 vec_perm_indices class.
14196 (vec_perm_indices::rotate_inputs): New function.
14197 (vec_perm_indices::all_in_range_p): Operate directly on the
14198 encoded form, without computing elided elements.
14199 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14200 encoding. Update for new vec_perm_indices class.
14201 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14202 the given vec_perm_builder.
14203 (expand_vec_perm_var): Update vec_perm_builder constructor.
14204 (expand_mult_highpart): Use vec_perm_builder instead of
14205 auto_vec_perm_indices.
14206 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14207 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14208 or double series encoding as appropriate.
14209 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14210 vec_perm_indices instead of auto_vec_perm_indices.
14211 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14212 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14213 (vect_permute_store_chain): Likewise.
14214 (vect_grouped_load_supported): Likewise.
14215 (vect_permute_load_chain): Likewise.
14216 (vect_shift_permute_load_chain): Likewise.
14217 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14218 (vect_transform_slp_perm_load): Likewise.
14219 (vect_schedule_slp_instance): Likewise.
14220 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14221 (vectorizable_mask_load_store): Likewise.
14222 (vectorizable_bswap): Likewise.
14223 (vectorizable_store): Likewise.
14224 (vectorizable_load): Likewise.
14225 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14226 vec_perm_indices instead of auto_vec_perm_indices. Use
14227 tree_to_vec_perm_builder to read the vector from a tree.
14228 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14229 vec_perm_builder instead of a vec_perm_indices.
14230 (have_whole_vector_shift): Use vec_perm_builder and
14231 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14232 truncation to calc_vec_perm_mask_for_shift.
14233 (vect_create_epilog_for_reduction): Likewise.
14234 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14235 from auto_vec_perm_indices to vec_perm_indices.
14236 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14237 instead of changing individual elements.
14238 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14239 the vector in d.perm.
14240 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14241 from auto_vec_perm_indices to vec_perm_indices.
14242 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14243 instead of changing individual elements.
14244 (arm_vectorize_vec_perm_const): Use new_vector to install
14245 the vector in d.perm.
14246 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14247 Update vec_perm_builder constructor.
14248 (rs6000_expand_interleave): Likewise.
14249 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14250 (rs6000_expand_interleave): Likewise.
14252 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14254 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14255 to qimode could truncate the indices.
14256 * optabs.c (expand_vec_perm_var): Likewise.
14258 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14260 * Makefile.in (OBJS): Add vec-perm-indices.o.
14261 * vec-perm-indices.h: New file.
14262 * vec-perm-indices.c: Likewise.
14263 * target.h (vec_perm_indices): Replace with a forward class
14265 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14266 * optabs.h: Include vec-perm-indices.h.
14267 (expand_vec_perm): Delete.
14268 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14269 (expand_vec_perm_const): Declare.
14270 * target.def (vec_perm_const_ok): Replace with...
14271 (vec_perm_const): ...this new hook.
14272 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14273 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14274 * doc/tm.texi: Regenerate.
14275 * optabs.def (vec_perm_const): Delete.
14276 * doc/md.texi (vec_perm_const): Likewise.
14277 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14278 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14279 expand_vec_perm for constant permutation vectors. Assert that
14280 the mode of variable permutation vectors is the integer equivalent
14281 of the mode that is being permuted.
14282 * optabs-query.h (selector_fits_mode_p): Declare.
14283 * optabs-query.c: Include vec-perm-indices.h.
14284 (selector_fits_mode_p): New function.
14285 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14286 is defined, instead of checking whether the vec_perm_const_optab
14287 exists. Use targetm.vectorize.vec_perm_const instead of
14288 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14289 fit in the vector mode before using a variable permute.
14290 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14291 vec_perm_indices instead of an rtx.
14292 (expand_vec_perm): Replace with...
14293 (expand_vec_perm_const): ...this new function. Take the selector
14294 as a vec_perm_indices rather than an rtx. Also take the mode of
14295 the selector. Update call to shift_amt_for_vec_perm_mask.
14296 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14297 Use vec_perm_indices::new_expanded_vector to expand the original
14298 selector into bytes. Check whether the indices fit in the vector
14299 mode before using a variable permute.
14300 (expand_vec_perm_var): Make global.
14301 (expand_mult_highpart): Use expand_vec_perm_const.
14302 * fold-const.c: Includes vec-perm-indices.h.
14303 * tree-ssa-forwprop.c: Likewise.
14304 * tree-vect-data-refs.c: Likewise.
14305 * tree-vect-generic.c: Likewise.
14306 * tree-vect-loop.c: Likewise.
14307 * tree-vect-slp.c: Likewise.
14308 * tree-vect-stmts.c: Likewise.
14309 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14311 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14312 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14313 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14314 (aarch64_vectorize_vec_perm_const): ...this new function.
14315 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14316 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14317 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14318 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14319 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14320 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14321 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14323 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14324 check for NEON modes.
14325 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14326 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14327 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14328 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14330 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14331 the old VEC_PERM_CONST conditions.
14332 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14333 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14334 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14335 (ia64_vectorize_vec_perm_const_ok): Merge into...
14336 (ia64_vectorize_vec_perm_const): ...this new function.
14337 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14338 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14339 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14340 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14341 * config/mips/mips.c (mips_expand_vec_perm_const)
14342 (mips_vectorize_vec_perm_const_ok): Merge into...
14343 (mips_vectorize_vec_perm_const): ...this new function.
14344 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14345 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14346 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14347 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14348 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14349 (rs6000_expand_vec_perm_const): Delete.
14350 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14352 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14353 (altivec_expand_vec_perm_const_le): Take each operand individually.
14354 Operate on constant selectors rather than rtxes.
14355 (altivec_expand_vec_perm_const): Likewise. Update call to
14356 altivec_expand_vec_perm_const_le.
14357 (rs6000_expand_vec_perm_const): Delete.
14358 (rs6000_vectorize_vec_perm_const_ok): Delete.
14359 (rs6000_vectorize_vec_perm_const): New function.
14360 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14361 an element count and rtx array.
14362 (rs6000_expand_extract_even): Update call accordingly.
14363 (rs6000_expand_interleave): Likewise.
14364 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14365 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14366 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14367 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14368 (rs6000_expand_vec_perm_const): Delete.
14369 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14370 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14371 (altivec_expand_vec_perm_const_le): Take each operand individually.
14372 Operate on constant selectors rather than rtxes.
14373 (altivec_expand_vec_perm_const): Likewise. Update call to
14374 altivec_expand_vec_perm_const_le.
14375 (rs6000_expand_vec_perm_const): Delete.
14376 (rs6000_vectorize_vec_perm_const_ok): Delete.
14377 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14378 reference to the SPE evmerge intructions.
14379 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14380 an element count and rtx array.
14381 (rs6000_expand_extract_even): Update call accordingly.
14382 (rs6000_expand_interleave): Likewise.
14383 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14384 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14386 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14388 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14390 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14391 vector mode and that that mode matches the mode of the data
14393 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14394 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14395 directly using expand_vec_perm_1 when forcing selectors into
14397 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14399 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14401 * optabs-query.h (can_vec_perm_p): Delete.
14402 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14403 * optabs-query.c (can_vec_perm_p): Split into...
14404 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14405 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14406 particular selector is valid.
14407 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14408 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14409 (vect_grouped_load_supported): Likewise.
14410 (vect_shift_permute_load_chain): Likewise.
14411 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14412 (vect_transform_slp_perm_load): Likewise.
14413 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14414 (vectorizable_bswap): Likewise.
14415 (vect_gen_perm_mask_checked): Likewise.
14416 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14417 implementations of variable permutation vectors into account
14418 when deciding which selector to use.
14419 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14420 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14421 with a false third argument.
14422 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14423 to test whether the constant selector is valid and can_vec_perm_var_p
14424 to test whether a variable selector is valid.
14426 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14428 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14429 * optabs-query.c (can_vec_perm_p): Likewise.
14430 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14431 instead of vec_perm_indices.
14432 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14433 (vect_gen_perm_mask_checked): Likewise,
14434 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14435 (vect_gen_perm_mask_checked): Likewise,
14437 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14439 * optabs-query.h (qimode_for_vec_perm): Declare.
14440 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14441 (qimode_for_vec_perm): ...this new function.
14442 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14444 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14446 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14447 does not have a conditional at the top.
14449 2018-01-02 Richard Biener <rguenther@suse.de>
14451 * ipa-inline.c (big_speedup_p): Fix expression.
14453 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14456 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14459 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14463 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14464 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14465 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14466 cond_taken_branch_cost 3->4.
14468 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14470 PR tree-optimization/83581
14471 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14472 TODO_cleanup_cfg if any changes have been made.
14474 PR middle-end/83608
14475 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14476 convert_modes if target mode has the right side, but different mode
14479 PR middle-end/83609
14480 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14481 last argument when extracting from CONCAT. If either from_real or
14482 from_imag is NULL, use expansion through memory. If result is not
14483 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14484 the parts directly to inner mode, if even that fails, use expansion
14487 PR middle-end/83623
14488 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14489 check for bswap in mode rather than HImode and use that in expand_unop
14492 Copyright (C) 2018 Free Software Foundation, Inc.
14494 Copying and distribution of this file, with or without modification,
14495 are permitted in any medium without royalty provided the copyright
14496 notice and this notice are preserved.