1 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3 * doc/sourcebuild.texi (vect_double_cond_arith): Include
4 multiplication and division.
5 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
6 (cond_udiv@var{m}, cond_umod@var{m}): Document.
7 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
8 (cond_udiv_optab, cond_umod_optab): New optabs.
9 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
10 (IFN_COND_RDIV): New internal functions.
11 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
12 TRUNC_MOD_EXPR and RDIV_EXPR.
13 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
14 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
16 (SVE_INT_BINARY): Include mult.
17 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
18 (optab, sve_int_op): Handle mult.
19 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
21 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
22 for SVE_INT_BINARY_SD.
24 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
26 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
27 (optab, sve_int_op): Handle div and udiv.
28 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
29 for SVE_INT_BINARY_SD.
30 (*<optab><mode>3): New insn for the same.
32 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
34 * tree-vect-patterns.c: Include predict.h.
35 (vect_recog_divmod_pattern): Restrict check for division support
36 to when optimizing for size.
38 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
40 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
41 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
42 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
43 (gimple_match_op::set_op): Likewise.
44 (gimple_resimplify4): Declare.
45 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
46 (expr::gen_transform): Likewise.
47 (decision_tree::gen): Generate a simplification routine for 4 operands.
48 * gimple-match-head.c (gimple_simplify): Add an overload for
49 4 operands. In the top-level function, handle up to 4 call
50 arguments and call gimple_resimplify4.
51 (gimple_resimplify4): New function.
52 (build_call_internal): Pass a fourth operand.
53 (maybe_push_to_seq): Likewise.
54 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
55 Fold VEC_COND_EXPRs of an operation and a default value into
56 an IFN_COND_* function if possible.
57 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
59 (SVE_COND_FP_BINARY): Include them.
60 (optab, sve_fp_op): Handle them.
61 (SVE_INT_BINARY_REV): New code iterator.
62 (SVE_COND_FP_BINARY_REV): New int iterator.
63 (commutative): New int attribute.
64 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
66 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
68 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
69 (*cond_<optab><mode>): New patterns for reversed operands.
71 2018-05-25 Richard Biener <rguenther@suse.de>
73 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
74 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
75 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
76 (STMT_VINFO_GROUPED_ACCESS): Adjust.
77 * tree-vect-data-refs.c (everywhere): Adjust users.
78 * tree-vect-loop.c (everywhere): Likewise.
79 * tree-vect-slp.c (everywhere): Likewise.
80 * tree-vect-stmts.c (everywhere): Likewise.
81 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
83 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
85 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
87 (gcc_cv_as_section_exclude): ... this.
88 Try Solaris as #exclude syntax.
89 * configure: Regenerate.
90 * config.in: Regenerate.
91 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
93 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
94 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
96 * varasm.c (default_elf_asm_named_section): Don't check if
97 HAVE_GAS_SECTION_EXCLUDE is defined.
99 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
101 * doc/md.texi: Update the documentation of the cond_* optabs
102 to mention the new final operand. Fix GET_MODE_NUNITS call.
103 Describe the scalar case too.
104 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
105 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
107 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
108 (get_conditional_internal_fn): Update comment.
109 * tree-vect-loop.c (vectorizable_reduction): Pass the original
110 accumulator value as a final argument to conditional functions.
111 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
112 a define_expand and add an "else" operand. Assert for now that
113 the else operand is equal to operand 2. Use SVE_INT_BINARY and
114 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
115 (*cond_<optab><mode>): New patterns.
116 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
117 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
118 (UNSPEC_COND_EOR): Delete.
119 (optab): Remove associated mappings.
120 (SVE_INT_BINARY): New code iterator.
121 (sve_int_op): Remove int attribute and add "minus" to the code
123 (SVE_COND_INT_OP): Delete.
124 (SVE_COND_FP_OP): Rename to...
125 (SVE_COND_FP_BINARY): ...this.
127 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
129 * optabs.c (can_reuse_operands_p): New function.
130 (maybe_legitimize_operands): Try to reuse the results for
133 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
135 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
136 Add {q} suffix to insn mnemonic.
138 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
140 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
141 (msp430_warn_func_return): New.
143 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
145 * fold-const.c (tree_nonzero_bits): New function.
146 * fold-const.h (tree_nonzero_bits): Likewise.
147 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
148 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
150 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
154 * varasm.c (assemble_alias): Check ifunc_resolver only on
157 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
160 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
161 when memory input operand is handled.
163 2018-05-24 Luis Machado <luis.machado@linaro.org>
165 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
167 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
169 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
171 * match.pd: Delay FMA folds until after vectorization.
173 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
176 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
177 address check not strict.
179 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
181 * gimple-match.h (gimple_match_op): New class.
182 (mprts_hook): Replace parameters with a gimple_match_op *.
183 (maybe_build_generic_op): Likewise.
184 (gimple_simplified_result_is_gimple_val): Replace parameters with
185 a const gimple_match_op *.
186 (gimple_simplify): Replace code_helper * and tree * parameters with
187 a gimple_match_op * parameter.
188 (gimple_resimplify1): Replace code_helper *, tree and tree *
189 parameters with a gimple_match_op * parameter.
190 (gimple_resimplify2): Likewise.
191 (gimple_resimplify3): Likewise.
192 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
193 parameters with a gimple_match_op * parameter.
194 * gimple-match-head.c (gimple_simplify): Change prototypes of
195 auto-generated functions to take a gimple_match_op * instead of
196 separate code_helper * and tree * parameters. Make the same
197 change in the top-level overload and update calls to the
198 gimple_resimplify routines. Update calls to the auto-generated
199 functions and to maybe_push_res_to_seq in the publicly-facing
200 operation-specific gimple_simplify overloads.
201 (gimple_match_op::MAX_NUM_OPS): Define.
202 (gimple_resimplify1): Replace rcode and ops with a single res_op
203 parameter. Update call to gimple_simplify.
204 (gimple_resimplify2): Likewise.
205 (gimple_resimplify3): Likewise.
206 (mprts_hook): Replace parameters with a gimple_match_op *.
207 (maybe_build_generic_op): Likewise.
208 (build_call_internal): Replace type, nargs and ops with
210 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
211 with a single gimple_match_op *. Update calls to mprts_hook,
212 build_call_internal and gimple_simplified_result_is_gimple_val.
213 Factor out code that is common to the tree_code and combined_fn cases.
214 * genmatch.c (expr::gen_transform): Replace tem_code and
215 tem_ops with a gimple_match_op called tem_op. Update calls
216 to the gimple_resimplify functions and maybe_push_res_to_seq.
217 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
218 res_ops. Update call to the gimple_resimplify functions.
219 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
220 (decision_tree::gen): Make the functions take a gimple_match_op *
221 called res_op instead of separate res_code and res_ops parameters.
222 Update call accordingly.
223 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
224 and ops with a single res_op parameter. Update calls to
225 maybe_build_generic_op and maybe_push_res_to_seq.
226 (fold_stmt_1): Update calls to gimple_simplify and
227 replace_stmt_with_simplification.
228 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
229 and gimple_simplified_result_is_gimple_val.
230 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
232 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
233 with a gimple_match_op *.
234 (vn_nary_build_or_lookup): Likewise. Update call to
235 vn_nary_build_or_lookup_1.
236 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
237 gimple_match_op *. Update calls to the gimple_resimplify routines
238 and to gimple_simplified_result_is_gimple_val.
239 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
240 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
241 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
242 (visit_nary_op): Likewise.
243 (visit_reference_op_load): Likewise.
245 2018-05-23 Luis Machado <luis.machado@linaro.org>
247 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
248 modifier for printing the step amount.
250 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
253 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
256 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
258 * doc/sourcebuild.texi (Endianness): New subsubsection.
260 2018-05-23 Luis Machado <luis.machado@linaro.org>
262 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
263 <prefetch_dynamic_strides>: New const bool field.
264 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
265 prefetch_dynamic_strides.
266 (exynosm1_prefetch_tune): Likewise.
267 (thunderxt88_prefetch_tune): Likewise.
268 (thunderx_prefetch_tune): Likewise.
269 (thunderx2t99_prefetch_tune): Likewise.
270 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
272 (aarch64_override_options_internal): Update to set
273 PARAM_PREFETCH_DYNAMIC_STRIDES.
274 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
275 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
276 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
277 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
278 prefetch-dynamic-strides setting.
280 2018-05-23 Luis Machado <luis.machado@linaro.org>
282 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
283 <minimum_stride>: New const int field.
284 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
285 minimum_stride field defaulting to -1.
286 (exynosm1_prefetch_tune): Likewise.
287 (thunderxt88_prefetch_tune): Likewise.
288 (thunderx_prefetch_tune): Likewise.
289 (thunderx2t99_prefetch_tune): Likewise.
290 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
291 <default_opt_level>: Set to 3.
292 (aarch64_override_options_internal): Update to set
293 PARAM_PREFETCH_MINIMUM_STRIDE.
294 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
295 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
296 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
297 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
298 stride is constant and is below the minimum stride threshold.
300 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
302 * config/arm/arm-cpus.in (mode26): Delete.
303 (armv4): Delete mode26 reference.
304 * config/arm/arm.c (arm_configure_build_target): Delete use of
307 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
309 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
311 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
312 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
314 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
315 (floatunsdidf2): Ditto.
317 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
319 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
320 (fixuns_trunc<mode>si2_avx512f): Ditto.
321 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
322 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
323 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
325 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
327 PR rtl-optimization/79985
328 * df-scan.c (df_insn_refs_collect): Remove special case for
329 global registers and asm statements.
331 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
333 * extend.texi (Global Register Variables): Rewrite the bullet list.
334 Note that the register is available for allocation. Note that access
335 via inline asm must use constraints. Add note about async-signal
336 handlers. Remove paragraph about automagic register selection.
338 2018-05-23 Richard Biener <rguenther@suse.de>
340 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
341 of fixed offset from memset VN.
343 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
345 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
347 (alloc_cand_and_find_basis): Initialize first_interp field.
348 (slsr_process_mul): Modify first_interp field.
349 (slsr_process_add): Likewise.
350 (slsr_process_cast): Modify first_interp field for each new
352 (slsr_process_copy): Likewise.
353 (dump_candidate): Dump first_interp field.
354 (replace_mult_candidate): Process all interpretations, not just
356 (replace_rhs_if_not_dup): Likewise.
357 (replace_one_candidate): Likewise.
359 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
361 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
363 (aarch64_needs_frame_chain): New function.
364 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
366 2018-05-23 Sudakshina Das <sudi.das@arm.com>
369 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
370 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
371 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
372 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
373 as true for strict-align.
374 (aarch64_can_inline_p): Perform checks even when callee has no
375 attributes to check for strict alignment.
376 * doc/extend.texi (AArch64 Function Attributes): Document
378 * doc/invoke.texi: (AArch64 Options): Likewise.
380 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
382 PR tree-optimization/85853
383 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
384 the handling of the root of the node to...
385 (vect_slp_analyze_node_operations_1): ...this new function,
386 and run the whole thing with the child nodes' def types
387 set according to their SLP node's def type.
389 2018-05-23 Richard Biener <rguenther@suse.de>
392 * tree-data-ref.c (create_runtime_alias_checks): Defer
393 and ignore overflow warnings.
395 2018-05-23 Yury Gribov <tetra2005@gmail.com>
397 PR tree-optimization/85822
398 * tree-vrp.c (is_masked_range_test): Fix handling of negative
401 2018-05-23 Richard Biener <rguenther@suse.de>
403 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
404 memset constants via native_interpret_expr.
406 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
409 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
411 (cgraph_node::create_alias): Likewise.
412 (cgraph_node::get_availability): Check ifunc_resolver instead
413 of looking up ifunc attribute.
414 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
415 * varasm.c (do_assemble_alias): Likewise.
416 (assemble_alias): Likewise.
417 (default_binds_local_p_3): Likewise.
418 * cgraph.h (cgraph_node): Add ifunc_resolver.
419 (cgraph_node::only_called_directly_or_aliased_p): Return false
421 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
423 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
424 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
425 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
426 instead of looking up ifunc attribute.
428 2018-05-22 Luis Machado <luis.machado@linaro.org>
430 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
432 2018-05-22 Martin Sebor <msebor@redhat.com>
435 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
436 only when expasion succeeds.
437 (expand_builtin_strcmp): Same.
438 (expand_builtin_strncmp): Same.
440 2018-05-22 Martin Sebor <msebor@redhat.com>
442 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
444 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
445 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
447 * config/aarch64/aarch64-ldpstp.md: Replace uses of
448 aarch64_mem_pair_operand with memory_operand and delete operand swapping
450 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
451 Add check for legitimate_address.
452 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
453 (aarch64_swap_ldrstr_operands): New.
454 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
457 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
458 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
460 * config/aarch64/aarch64.md: New patterns to generate stp
462 (store_pair_sw, store_pair_dw): New patterns to generate stp for
463 single words and double words.
464 (load_pair_sw, load_pair_dw): Likewise.
465 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
467 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
469 * config/aarch64/aarch64-ldpstp.md: Modify peephole
470 for different mode ldpstp and add peephole for merged zero stores.
472 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
474 (aarch64_gen_store_pair): Rename calls to match new patterns.
475 (aarch64_gen_load_pair): Rename calls to match new patterns.
476 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
477 (load_pair<DREG:mode><DREG2:mode>): ... This.
478 (store_pair<mode>): Rename to...
479 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
480 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
482 (V_INT_EQUIV): Handle SImode.
483 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
486 2018-05-22 Martin Sebor <msebor@redhat.com>
489 * calls.c (maybe_warn_nonstring_arg): Use string length to set
490 or ajust the presumed bound on an operation to avoid unnecessary
493 2018-05-22 Martin Sebor <msebor@redhat.com>
495 PR tree-optimization/85826
496 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
497 assuming that a DECL necesarily has a constant size.
499 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
502 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
504 2018-05-22 Richard Biener <rguenther@suse.de>
506 PR tree-optimization/85834
507 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
508 non-constant and non-zero memset arguments.
510 2018-05-22 Martin Liska <mliska@suse.cz>
513 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
515 2018-05-22 Richard Biener <rguenther@suse.de>
517 PR tree-optimization/85863
518 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
519 comparisons when vectype is specified.
520 (vectorizable_condition): Do not specify vectype for
521 vect_is_simple_cond when SLP vectorizing.
523 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
526 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
527 define __ibm128 as long double.
528 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
529 as a distinct type when IEEE 128-bit support is enabled.
530 (init_float128_ieee): Fix up conversions between IFmode and IEEE
531 128-bit types to use the correct functions.
532 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
533 convert between 128-bit floating point types that have different
534 modes but the same representation, instead of using gen_lowpart to
536 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
538 (IFKF_reg): New attributes to give the register constraints for
540 (extend<mode>tf2_internal): New insns to mark an explicit
541 conversion between 128-bit floating point types that have a
542 different mode but share the same representation.
544 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
546 PR tree-optimization/85814
547 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
548 a null return from get_strinfo when unsharing the next
549 strinfo in the chain.
551 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
554 * varasm.c (weak_finish): Clean up weak_decls.
556 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
558 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
559 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
560 UNSPEC_UADALP values.
561 * config/aarch64/iterators.md (ABAL): New int iterator.
564 (sur): Add mappings for the above.
565 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
567 (aarch64_<sur>abal<mode>_4): Likewise.
568 (aarch64_<sur>adalp<mode>_3): Likewise.
569 (<sur>sadv16qi): New define_expand.
571 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
573 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
574 (*movdf_internal): Ditto.
575 (*rcpsf2_sse): Ditto.
576 (*rsqrtsf2_sse): Ditto.
577 (*sqrt<mode>2_sse): Ditto.
579 2018-05-21 Tamar Christina <tamar.christina@arm.com>
581 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
583 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
584 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
585 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
586 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
588 * config/aarch64/arm_neon.h: Likewise.
589 * config/aarch64/iterators.md (VQ_I): New.
591 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
593 * config.gcc: Add arc/t-multilib-linux to tmake_file for
595 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
598 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
600 * config/nds32/constraints.md (S): New constraint.
601 * config/nds32/nds32.md (call_internal): Use constraint S.
602 (call_value_internal): Likewise.
603 (sibcall_internal): Likewise.
604 (sibcall_value_internal): Likewise.
606 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
607 Chung-Ju Wu <jasonwucj@gmail.com>
609 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
612 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
613 Chung-Ju Wu <jasonwucj@gmail.com>
615 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
616 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
617 (nds32_rtx_costs_impl): Simplify.
618 (nds32_address_cost_impl): Simplify.
619 (nds32_init_rtx_costs): New function.
620 (nds32_rtx_costs_speed_prefer): Likewise.
621 (nds32_rtx_costs_size_prefer): Likewise.
622 (nds32_address_cost_speed_prefer): Likewise.
623 (nds32_address_cost_speed_fwprop): Likewise.
624 (nds32_address_cost_size_prefer): Likewise.
625 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
626 * config/nds32/nds32.c (nds32_option_override): Use
627 nds32_init_rtx_costs function.
629 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
631 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
632 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
633 (TARGET_PIPELINE_N8): Likewise.
634 (TARGET_PIPELINE_N10): Likewise.
635 (TARGET_PIPELINE_N13): Likewise.
636 (TARGET_PIPELINE_GRAYWOLF): Likewise.
638 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
640 * config/nds32/nds32-fpu.md: Update copyright year.
642 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
644 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
646 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
648 * config/nds32/nds32.c
649 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
650 * config/nds32/nds32.opt (minline-asm-r15): New option.
652 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
654 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
656 * config/nds32/nds32.md (abssi2): New pattern.
658 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
660 * config/i386/i386.md (rex64namesuffix): New mode attribute.
661 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
662 Merge insn pattern from sse_cvtsi2ss<round_name> and
663 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
664 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
665 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
666 using SWI48 mode iterator.
667 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
668 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
669 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
670 pattern from sse_cvttss2si<round_saeonly_name>
671 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
672 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
673 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
674 using SWI48 mode iterator.
675 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
676 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
677 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
678 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
679 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
680 using SWI48 mode iterator.
681 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
682 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
683 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
684 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
685 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
687 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
688 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
689 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
690 pattern from sse_cvttsd2si<round_saeonly_name>
691 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
693 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
695 * config/nds32/nds32-md-auxiliary.c
696 (nds32_valid_smw_lwm_base_p): Refine.
697 (nds32_output_smw_single_word): Refine.
698 (nds32_output_smw_double_word): New.
699 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
701 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
703 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
704 (nds32_output_stack_pop): Refine.
705 (nds32_expand_unaligned_load): Refine.
706 (nds32_expand_unaligned_store): Refine.
708 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
709 Chung-Ju Wu <jasonwucj@gmail.com>
711 * config/nds32/constants.md: Add TP_REGNUM constant.
712 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
713 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
715 * config/nds32/nds32-doubleword.md: Consider flag_pic.
716 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
717 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
718 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
719 and PIC code generation.
720 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
722 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
724 * config/nds32/nds32.md: Support TLS and PIC.
725 * config/nds32/nds32.c: Support TLS and PIC.
726 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
727 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
730 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
732 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
735 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
736 Chung-Ju Wu <jasonwucj@gmail.com>
738 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
739 * config/nds32/nds32-md-auxiliary.c
740 (symbolic_reference_mentioned_p): New.
741 (nds32_legitimize_ict_address): New.
742 (nds32_expand_ict_move): New.
743 (nds32_indirect_call_referenced_p): New.
744 (nds32_symbol_binds_local_p): Delete.
745 (nds32_long_call_p): Modify.
746 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
747 * config/nds32/nds32-protos.h
748 (symbolic_reference_mentioned_p): Declare.
749 (nds32_legitimize_ict_address): Declare.
750 (nds32_expand_ict_move): Declare.
751 (nds32_indirect_call_referenced_p): Declare.
752 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
753 (nds32_relax_group): Use nds32_ict_const_p as condition.
754 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
755 (nds32_asm_file_start): Output ict_model directive in asm code.
756 (nds32_legitimate_address_p): Consider indirect call.
757 (nds32_print_operand): Consider indirect call.
758 (nds32_print_operand_address): Consider indirect call.
759 (nds32_insert_attributes): Handle "indirect_call" attribute.
760 (TARGET_LEGITIMATE_ADDRESS_P): Define.
761 (TARGET_LEGITIMATE_CONSTANT_P): Define.
762 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
763 (TARGET_DELEGITIMIZE_ADDRESS): Define.
764 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
765 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
766 (TARGET_ICT_MODEL_SMALL): Define.
767 (TARGET_ICT_MODEL_LARGE): Define.
768 * config/nds32/nds32.md (movsi): Consider ict model.
769 (call, call_value): Consider ict model.
770 (sibcall, sibcall_value): Consider ict model.
771 * config/nds32/nds32.opt (mict-model): New option.
772 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
775 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
776 Monk Chiang <sh.chiang04@gmail.com>
777 Jim Wilson <jimw@sifive.com>
779 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
780 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
781 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
782 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
783 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
784 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
785 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
786 compute save_libcall_adjustment properly.
787 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
788 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
789 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
790 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
791 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
792 (ABI_SPEC): Handle mabi=ilp32e.
793 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
795 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
796 <-march>: Add rv32e as an example.
798 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
801 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
802 (intra_create_variable_infos): Handle C++ constructors.
804 2018-05-18 Martin Liska <mliska@suse.cz>
806 * passes.def: Remove a redundant pass.
808 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
811 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
813 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
815 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
817 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
819 (armv2, armv2a, armv3, armv3m): Delete architectures.
820 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
821 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
822 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
824 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
825 (*mulsidi3adddi): Likewise.
826 (mulsidi3): Likewise.
827 (*mulsidi3_nov6): Likewise.
828 (umulsidi3): Likewise.
829 (umulsidi3_nov6): Likewise.
830 (umaddsidi4): Likewise.
831 (*umulsidi3adddi): Likewise.
832 (smulsi3_highpart): Likewise.
833 (*smulsi3_highpart_nov6): Likewise.
834 (umulsi3_highpart): Likewise.
835 (*umulsi3_highpart_nov6): Likewise.
836 * config/arm/arm.h (arm_arch3m): Delete.
837 * config/arm/arm.c (arm_arch3m): Delete.
838 (arm_option_override_internal): Update armv3-related comment.
839 (arm_configure_build_target): Delete use of isa_bit_mode32.
840 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
841 (arm_rtx_costs_internal): Delete check of arm_arch3m.
842 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
845 * config/arm/arm-protos.h (arm_arch3m): Delete.
846 * config/arm/arm-tables.opt: Regenerate.
847 * config/arm/arm-tune.md: Likewise.
848 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
849 deleted architectures.
851 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
853 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
854 (armv5t, armv5te): New features.
855 (ARMv5, ARMv5e): Delete fgroups.
856 (ARMv5t, ARMv5te): Adjust for above changes.
858 (armv5, armv5e): Delete arches.
859 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
861 (*call_reg_arm): Likewise.
862 (*call_value_reg_armv5): Likewise.
863 (*call_value_reg_arm): Likewise.
864 (*call_symbol): Likewise.
865 (*call_value_symbol): Likewise.
866 (*sibcall_insn): Likewise.
867 (*sibcall_value_insn): Likewise.
869 (prefetch): Likewise.
870 (define_split and define_peephole2 dependent on arm_arch5):
872 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
874 (TARGET_ARM_QBIT): Likewise.
875 (TARGET_DSP_MULTIPLY): Likewise.
876 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
877 (arm_arch5, arm_arch5e): Delete.
878 (arm_arch5t, arm_arch5te): Declare.
879 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
880 (arm_arch5t): Declare.
881 (arm_option_reconfigure_globals): Update for the above.
882 (arm_options_perform_arch_sanity_checks): Update comment, replace
883 use of arm_arch5 with arm_arch5t.
884 (use_return_insn): Likewise.
885 (arm_emit_call_insn): Likewise.
886 (output_return_instruction): Likewise.
887 (arm_final_prescan_insn): Likewise.
888 (arm_coproc_builtin_available): Likewise.
889 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
890 arm_arch5e with arm_arch5t and arm_arch5te.
891 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
892 (arm_arch5t, arm_arch5te): Declare.
893 * config/arm/arm-tables.opt: Regenerate.
894 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
895 * config/arm/t-multilib: Likewise.
896 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
897 instead of arm_arch5.
898 (*call_reg_thumb1): Likewise.
899 (*call_value_reg_thumb1_v5): Likewise.
900 (*call_value_reg_thumb1): Likewise.
901 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
903 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
905 2018-05-18 Martin Liska <mliska@suse.cz>
907 PR gcov-profile/84846
908 * doc/gcov.texi: Document -t option of gcov tool.
910 2018-05-18 Martin Liska <mliska@suse.cz>
912 PR gcov-profile/84846
913 * gcov.c (print_usage): Add new -t option.
914 (process_args): Handle the option.
915 (generate_results): Use stdout as output when requested by
918 2018-05-18 Martin Liska <mliska@suse.cz>
920 PR gcov-profile/84846
921 * coverage.c (coverage_init): Write PWD to .gcno file.
922 * doc/gcov.texi: Document how working directory is printed.
923 * gcov-dump.c (dump_gcov_file): Print PWD.
924 * gcov.c (output_intermediate_file): Likewise.
925 (read_graph_file): Read PWD string.
926 (output_lines): Print PWD.
928 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
931 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
932 for retval and return false if all args to phi are zero.
934 2018-05-18 Richard Biener <rguenther@suse.de>
936 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
938 (evrp_dom_walker::before_dom_children): Call it.
940 2018-05-18 Richard Biener <rguenther@suse.de>
942 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
943 results when processing array refs with variable index.
945 2018-05-18 Toon Moene <toon@moene.org>
947 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
948 directly after that of -floop-interchange. Indicate that both
949 options are enabled by default when specifying -O3.
951 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
953 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
954 iterator. Delete separate integer-mode vec_set<mode> expander.
955 (aarch64_simd_vec_setv2di): Delete.
956 (vec_setv2di): Delete.
957 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
958 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
959 the "w, r" alternative.
961 2018-05-18 Martin Liska <mliska@suse.cz>
963 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
964 * tree-pass.h (make_pass_lower_switch_O0): New function.
965 * tree-switch-conversion.c (node_has_low_bound): Remove.
966 (node_has_high_bound): Likewise.
967 (node_is_bounded): Likewise.
968 (class pass_lower_switch): Make it a template type and create
970 (pass_lower_switch::execute): Add template argument.
971 (make_pass_lower_switch): New function.
972 (make_pass_lower_switch_O0): New function.
973 (do_jump_if_equal): Remove.
974 (emit_case_nodes): Simplify to just handle all 3 cases and leave
975 all the hard work to tree optimization passes.
977 2018-05-18 Martin Liska <mliska@suse.cz>
979 * dbgcnt.c (limit_low): Renamed from limit.
980 (limit_high): New variable.
981 (dbg_cnt_is_enabled): Check for upper limit.
982 (dbg_cnt): Adjust dumping.
983 (dbg_cnt_set_limit_by_index): Add new argument for high
985 (dbg_cnt_set_limit_by_name): Likewise.
986 (dbg_cnt_process_single_pair): Parse new format.
987 (dbg_cnt_process_opt): Use strtok.
988 (dbg_cnt_list_all_counters): Remove 'value' and add
990 * doc/invoke.texi: Document changes.
992 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
994 * doc/sourcebuild.texi (scalar_all_fma): Document.
995 * tree.def (FMA_EXPR): Delete.
996 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
997 * internal-fn.c (ternary_direct): New macro.
998 (expand_ternary_optab_fn): Likewise.
999 (direct_ternary_optab_supported_p): Likewise.
1000 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1001 * builtins.c (fold_builtin_fma): Delete.
1002 (fold_builtin_3): Don't call it.
1003 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1004 * expr.c (expand_expr_real_2): Likewise.
1005 * fold-const.c (operand_equal_p): Likewise.
1006 (fold_ternary_loc): Likewise.
1007 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1008 * gimple.c (DEFTREECODE): Likewise.
1009 * gimplify.c (gimplify_expr): Likewise.
1010 * optabs-tree.c (optab_for_tree_code): Likewise.
1011 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1012 * tree-eh.c (operation_could_trap_p): Likewise.
1013 (stmt_could_throw_1_p): Likewise.
1014 * tree-inline.c (estimate_operator_cost): Likewise.
1015 * tree-pretty-print.c (dump_generic_node): Likewise.
1016 (op_code_prio): Likewise.
1017 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1018 * tree-ssa-operands.c (get_expr_operands): Likewise.
1019 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1020 * fold-const-call.h (fold_fma): Delete.
1021 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1022 CFN_FNMA and CFN_FNMS.
1024 * genmatch.c (combined_fn): New enum.
1025 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1026 (commutative_op): New function.
1027 (commutate): Use it. Handle more than 2 operands.
1028 (dt_operand::gen_gimple_expr): Use commutative_op.
1029 (parser::parse_expr): Allow :c to be used with non-binary
1030 operators if the commutative operand is known.
1031 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1032 CFN_FMS, CFN_FNMA and CFN_FNMS.
1033 (backprop::process_assign_use): Remove FMA_EXPR handling.
1034 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1035 (gen_hsa_fma): New function.
1036 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1037 IFN_FNMA and IFN_FNMS.
1038 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1039 * gimple-fold.h (follow_all_ssa_edges): Declare.
1040 * gimple-fold.c (follow_all_ssa_edges): New function.
1041 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1042 gimple_build interface and use follow_all_ssa_edges to fold the result.
1043 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1044 instead of checking for optabs directly.
1045 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1046 rather than FMA_EXPRs.
1047 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1048 call to IFN_FMA instead of an FMA_EXPR.
1050 2018-05-17 Jim Wilson <jimw@sifive.com>
1052 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1053 sign extended promoted subreg, and the range does not have the sign bit
1054 set, then do a sign extend.
1056 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1057 test, check for sign extended subreg and/or constant operands, and
1058 do a sign extend in that case.
1060 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1062 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1063 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1065 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1066 Change logics_shift_reg to logics_shift_imm.
1067 (thunderx2t99_fp_loadpair_basic): Delete.
1068 (thunderx2t99_fp_storepair_basic): Delete.
1069 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1070 (thunderx2t99_asimd_polynomial): Delete.
1071 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1072 and neon_fp_mul_d_scalar_q.
1073 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1074 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1075 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1076 (thunderx2t99_asimd_lut): Add missing tbl types.
1077 (thunderx2t99_asimd_ext): Delete.
1078 (thunderx2t99_asimd_load1_1_mult): Delete.
1079 (thunderx2t99_asimd_load1_2_mult): Delete.
1080 (thunderx2t99_asimd_load1_ldp): New.
1081 (thunderx2t99_asimd_load1): New.
1082 (thunderx2t99_asimd_load2): Add missing *load2* types.
1083 (thunderx2t99_asimd_load3): New.
1084 (thunderx2t99_asimd_load4): New.
1085 (thunderx2t99_asimd_store1_1_mult): Delete.
1086 (thunderx2t99_asimd_store1_2_mult): Delete.
1087 (thunderx2t99_asimd_store2_mult): Delete.
1088 (thunderx2t99_asimd_store2_onelane): Delete.
1089 (thunderx2t99_asimd_store_stp): New.
1090 (thunderx2t99_asimd_store1): New.
1091 (thunderx2t99_asimd_store2): New.
1092 (thunderx2t99_asimd_store3): New.
1093 (thunderx2t99_asimd_store4): New.
1095 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1097 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1098 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1100 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1101 Segher Boessenkool <segher@kernel.crashing.org>
1104 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1107 2018-05-17 Richard Biener <rguenther@suse.de>
1109 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1110 for pruning loop and prune defs feeding only already visited PHIs.
1112 2018-05-17 Richard Biener <rguenther@suse.de>
1114 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1116 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1117 Richard Biener <rguenther@suse.de>
1119 PR tree-optimization/85793
1120 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1121 for VMAT_ELEMENTWISE.
1123 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1125 * internal-fn.h (lookup_internal_fn): Declare
1126 * internal-fn.c (lookup_internal_fn): New function.
1127 * gimple.c (gimple_build_call_from_tree): Handle calls to
1129 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1130 internal function names.
1131 * tree-pretty-print.c (dump_generic_node): Likewise.
1132 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1134 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1136 * gimple-fold.h (gimple_build): Make the function forms take
1137 combined_fn rather than built_in_function.
1138 (gimple_simplify): Likewise.
1139 * gimple-match-head.c (gimple_simplify): Likewise.
1140 * gimple-fold.c (gimple_build): Likewise.
1141 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1142 rather than gimple_build_call_internal.
1143 (get_initial_defs_for_reduction): Likewise.
1144 (vect_create_epilog_for_reduction): Likewise.
1145 (vectorizable_live_operation): Likewise.
1147 2018-05-17 Martin Liska <mliska@suse.cz>
1149 * gimple-ssa-sprintf.c (format_directive): Do not use
1150 space in between 'G_' and '('.
1152 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1155 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1156 even if the mask is not all ones.
1159 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1161 (ix86_gimple_fold_builtin): Likewise.
1164 * config/i386/i386.c: Include tree-vector-builder.h.
1165 (ix86_vector_shift_count): New function.
1166 (ix86_fold_builtin): Fold shift builtins by scalar count.
1167 (ix86_gimple_fold_builtin): Likewise.
1169 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1170 _mm512_setzero): New intrinsics.
1172 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1173 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1175 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1176 code generation for cases where splatting a value is not useful.
1177 * simplify-rtx.c (simplify_ternary_operation): Simplify
1178 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1179 mode to a vec_concat.
1181 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1183 * config.gcc: Support "goldmont-plus".
1184 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1186 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1187 PROCESSOR_GOLDMONT_PLUS.
1188 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1189 (processor_target_table): Add "goldmont-plus".
1190 (PTA_GOLDMONT_PLUS): Define.
1191 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1192 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1193 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1194 (fold_builtin_cpu): Add "goldmont-plus".
1195 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1196 (ix86_option_override_internal): Add "goldmont-plus".
1197 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1198 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1199 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1200 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1202 2018-05-17 Richard Biener <rguenther@suse.de>
1204 PR tree-optimization/85757
1205 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1206 remove defs that only feed that PHI from further processing.
1208 2018-05-16 Jim Wilson <jimw@sifive.com>
1210 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1212 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1214 2018-05-16 Mark Wielaard <mark@klomp.org>
1216 * dwarf2out.c (count_index_strings): New function.
1217 (output_indirect_strings): Call count_index_strings and generate
1218 header for dwarf_version >= 5.
1220 2018-05-16 Mark Wielaard <mark@klomp.org>
1222 * dwarf2out.c (dwarf_FORM): New function.
1223 (set_indirect_string): Use dwarf_FORM.
1224 (reset_indirect_string): Likewise.
1225 (size_of_die): Likewise.
1226 (value_format): Likewise.
1227 (output_die): Likewise.
1228 (add_skeleton_AT_string): Likewise.
1229 (output_macinfo_op): Likewise.
1230 (index_string): Likewise.
1231 (output_index_string_offset): Likewise.
1232 (output_index_string): Likewise.
1233 (count_index_strings): Likewise.
1235 2018-05-16 Carl Love <cel@us.ibm.com>
1237 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1238 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1240 2018-05-16 Martin Jambor <mjambor@suse.cz>
1242 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1243 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1245 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1247 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1248 (fnma<mode>4): Likewise.
1249 (fms<mode>4): Likewise.
1250 (fnms<mode>4): Likewise.
1251 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1252 (aarch64_fnma<mode>4): Likewise.
1253 (aarch64_fms<mode>4): Likewise.
1254 (aarch64_fnms<mode>4): Likewise.
1255 (aarch64_fnmadd<mode>4): Likewise.
1257 2018-05-16 Jason Merrill <jason@redhat.com>
1259 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1261 2018-05-16 Richard Biener <rguenther@suse.de>
1263 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1264 (dump_stmt_cost): Declare.
1265 (add_stmt_cost): Dump cost we add.
1266 (add_stmt_costs): New function.
1267 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1269 (vect_analyze_stmt): Adjust prototype.
1270 (vectorizable_condition): Likewise.
1271 (vectorizable_live_operation): Likewise.
1272 (vectorizable_reduction): Likewise.
1273 (vectorizable_induction): Likewise.
1274 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1275 cost vector to pass to vectorizable_ and record afterwards.
1276 (vect_model_reduction_cost): Take cost vector argument and adjust.
1277 (vect_model_induction_cost): Likewise.
1278 (vectorizable_reduction): Likewise.
1279 (vectorizable_induction): Likewise.
1280 (vectorizable_live_operation): Likewise.
1281 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1282 SLP_TREE_NUMBER_OF_VEC_STMTS.
1283 (vect_analyze_slp_cost_1): Remove.
1284 (vect_analyze_slp_cost): Likewise.
1285 (vect_slp_analyze_node_operations): Take visited args and
1286 a target cost vector. Avoid processing already visited stmt sets.
1287 (vect_slp_analyze_operations): Use a local cost vector to gather
1288 costs and register those of non-discarded instances.
1289 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1290 (vect_schedule_slp_instance): Remove copying of
1291 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1293 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1294 adding cost. Record cost entry location.
1295 (vect_prologue_cost_for_slp_op): Function to compute cost of
1296 a constant or invariant generated for SLP vect in the prologue,
1297 split out from vect_analyze_slp_cost_1.
1298 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1299 (vect_model_promotion_demotion_cost): Likewise.
1300 (vect_model_store_cost): Likewise, make static.
1301 (vect_model_load_cost): Likewise.
1302 (vectorizable_bswap): Add cost vector arg and adjust.
1303 (vectorizable_call): Likewise.
1304 (vectorizable_simd_clone_call): Likewise.
1305 (vectorizable_conversion): Likewise.
1306 (vectorizable_assignment): Likewise.
1307 (vectorizable_shift): Likewise.
1308 (vectorizable_operation): Likewise.
1309 (vectorizable_store): Likewise.
1310 (vectorizable_load): Likewise.
1311 (vectorizable_condition): Likewise.
1312 (vectorizable_comparison): Likewise.
1313 (can_vectorize_live_stmts): Likewise.
1314 (vect_analyze_stmt): Likewise.
1315 (vect_transform_stmt): Adjust calls to vectorizable_*.
1316 * tree-vectorizer.c: Include gimple-pretty-print.h.
1317 (dump_stmt_cost): New function.
1319 2018-05-16 Richard Biener <rguenther@suse.de>
1321 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1322 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1323 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1324 (check_name): New callback.
1325 (dse_classify_store): Track cycles via a visited bitmap of PHI
1326 defs and simplify handling of in-loop and across loop dead stores
1327 and properly fail for loop-variant refs. Handle byte-tracking with
1328 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1331 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1333 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1334 (vect_get_mask_type_for_stmt): Likewise.
1335 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1337 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1338 to determine the statement's vector type and the vector type that
1339 should be used for calculating nunits. Deal with cases in which
1340 the type has to be deferred.
1341 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1342 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1343 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1344 (vect_determine_vf_for_stmt): New functions, split out from...
1345 (vect_determine_vectorization_factor): ...here.
1346 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1347 (vect_get_mask_type_for_stmt): New functions, split out from
1348 vect_determine_vectorization_factor.
1350 2018-05-16 Richard Biener <rguenther@suse.de>
1352 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1353 verify the [VEC_]COND_EXPR embedded comparison.
1355 2018-05-15 Martin Sebor <msebor@redhat.com>
1357 PR tree-optimization/85753
1358 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1359 RECORD_TYPE in addition to ARRAY_TYPE.
1361 2018-05-15 Martin Sebor <msebor@redhat.com>
1364 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1366 2018-05-15 Richard Biener <rguenther@suse.de>
1368 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1369 add by_clobber_p one. Change algorithm to collect all defs
1370 representing uses we need to walk and try reducing them to
1371 a single one before failing.
1372 (dse_dom_walker::dse_optimize_stmt): Adjust.
1374 2018-05-13 Mark Wielaard <mark@klomp.org>
1376 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1377 (size_of_loc_descr): Likewise.
1378 (output_loc_operands): Likewise.
1379 (output_loc_operands_raw): Likewise.
1380 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1381 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1382 (hash_loc_operands): Likewise.
1383 (compare_loc_operands): Likewise.
1385 2018-05-14 Mark Wielaard <mark@klomp.org>
1387 * dwarf2out.c (count_index_addrs): New function.
1388 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1390 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1392 PR tree-optimization/83648
1393 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1394 return value as malloc candidate.
1396 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1399 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1400 as true in call to suggest_attribute.
1402 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1404 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1405 -mreadonly-in-sdata.
1407 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1409 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1411 (aarch64_crypto_aesd_fused): Likewise.
1413 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1415 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1416 (movsi_aarch64): Likewise.
1417 (load_pairsi): Likewise.
1418 (load_pairdi): Likewise.
1419 (store_pairsi): Likewise.
1420 (store_pairdi): Likewise.
1421 (load_pairsf): Likewise.
1422 (load_pairdf): Likewise.
1423 (store_pairsf): Likewise.
1424 (store_pairdf): Likewise.
1425 (zero_extend): Likewise.
1426 (trunc): Swap alternatives.
1427 (fcvt_target): Add '?' to prefer w over r.
1429 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1432 * config/i386/i386.md: Disallow non-commutative arithmetics in
1433 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1434 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1435 in the peephole2 before it.
1437 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1439 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1440 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1441 (ix86_handle_option): Handle -mcldemote.
1442 * config.gcc: New header.
1443 * config/i386/cldemoteintrin.h: New file.
1444 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1445 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1447 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1448 OPTION_MASK_ISA_CLDEMOTE.
1449 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1450 (ix86_valid_target_attribute_inner_p): Ditto.
1451 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1452 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1453 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1454 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1455 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1457 * config/i386/i386.opt: Add -mcldemote.
1458 * config/i386/x86intrin.h: New header.
1459 * doc/invoke.texi: Add -mcldemote.
1461 2018-05-14 Richard Biener <rguenther@suse.de>
1463 * doc/match-and-simplify.texi: Adjust :s documentation.
1465 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1467 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1468 intended memcpy size.
1469 (REORDER_45): Likewise.
1471 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1473 * sort.cc: New file.
1474 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1475 * vec.c (qsort_chk): Use gcc_qsort.
1476 * Makefile.in (OBJS-libcommon): Add sort.o.
1477 (build/sort.o): New target. Use it...
1478 (BUILD_RTL): ... here, and...
1479 (build/gencfn-macros): ... here, and...
1480 (build/genmatch): ... here.
1482 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1483 Chung-Ju Wu <jasonwucj@gmail.com>
1485 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1486 * config/nds32/nds32-graywolf.md: New file.
1487 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1488 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1490 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1491 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1492 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1493 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1494 * config/nds32/pipelines.md: Include n15 settings.
1496 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1497 Chung-Ju Wu <jasonwucj@gmail.com>
1499 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1500 * config/nds32/nds32-n13.md: New file.
1501 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1502 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1504 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1505 * config/nds32/nds32.md (pipeline_model): Add n13.
1506 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1507 * config/nds32/pipelines.md: Include n13 settings.
1509 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1510 Chung-Ju Wu <jasonwucj@gmail.com>
1512 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1513 * config/nds32/nds32-n10.md: New file.
1514 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1515 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1517 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1518 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1519 * config/nds32/nds32.md (pipeline_model): Add n10.
1520 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1521 * config/nds32/pipelines.md: Include n10 settings.
1523 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1524 Kito Cheng <kito.cheng@gmail.com>
1525 Chung-Ju Wu <jasonwucj@gmail.com>
1527 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1528 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1529 Add enum values for DSP extension instructions.
1530 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1532 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1533 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1535 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1536 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1537 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1538 * config/nds32/nds32-intrinsic.md: Likewise.
1539 * config/nds32/nds32_intrinsic.h: Likewise.
1540 * config/nds32/nds32-md-auxiliary.c: Likewise.
1541 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1542 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1543 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1544 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1545 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1546 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1547 TYPE_DMAC in switch statement.
1548 * config/nds32/nds32.c: New checking and implementation for DSP
1549 extension instructions.
1550 * config/nds32/nds32.h: Likewise.
1551 * config/nds32/nds32.md: Likewise.
1552 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1553 * config/nds32/predicates.md: Implement new predicates for DSP
1556 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1558 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1559 Reformat alternatives and attributes so it is easier to identify
1560 which constraints/attributes go with which instruction.
1561 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1562 (mov<mode>_softfloat32, FMOVE64): Likewise.
1563 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1564 (mov<mode>_softfloat64, FMOVE64): Likewise.
1566 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1568 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1570 (Basic PowerPC Built-in Functions): The new name of the
1571 subsection previously known as "PowerPC Built-in Functions".
1572 (Basic PowerPC Built-in Functions Available on all Configurations):
1574 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1575 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1576 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1577 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1579 2018-05-11 Martin Jambor <mjambor@suse.cz>
1582 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1585 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1588 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1590 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1592 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1593 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1594 (ix86_handle_option): Handle -mwaitpkg.
1595 * config.gcc: New header.
1596 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1597 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1598 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1600 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1601 OPTION_MASK_ISA_WAITPKG.
1602 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1603 (ix86_option_override_internal): Add PTA_WAITPKG.
1604 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1605 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1606 IX86_BUILTIN_TPAUSE.
1607 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1608 __builtin_ia32_umwait and __builtin_ia32_tpause.
1609 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1610 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1611 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1612 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1613 UNSPECV_TPAUSE): New.
1614 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1615 * config/i386/i386.opt: Add -mwaitpkg.
1616 * config/i386/waitpkgintrin.h: New file.
1617 * config/i386/x86intrin.h: New header.
1618 * doc/invoke.texi: Add -mwaitpkg.
1620 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1623 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1625 (cortex-m0): Use armv6s-m isa.
1626 (cortex-m0plus): Likewise.
1627 (cortex-m1): Likewise.
1628 (cortex-m0.small-multiply): Likewise.
1629 (cortex-m0plus.small-multiply): Likewise.
1630 (cortex-m1.small-multiply): Likewise.
1632 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1633 Jakub Jelinek <jakub@redhat.com>
1635 PR tree-optimization/85692
1636 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1637 source permute as well.
1639 2018-05-11 Martin Liska <mliska@suse.cz>
1642 * doc/extend.texi: Document LLVM style format for no_sanitize
1645 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1647 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1648 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1649 (mode_supports_vsx_dform_quad): Likewise.
1650 (mode_supports_vmx_dform): Move these functions to be next to the
1651 other mode_supports functions.
1652 (mode_supports_dq_form): Likewise.
1653 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1654 mode_supports_dq_form.
1655 (reg_offset_addressing_ok_p): Likewise.
1656 (offsettable_ok_by_alignment): Likewise.
1657 (rs6000_legitimate_offset_address_p): Likewise.
1658 (legitimate_lo_sum_address_p): Likewise.
1659 (rs6000_legitimize_address): Likewise.
1660 (rs6000_legitimize_reload_address): Likewise.
1661 (rs6000_secondary_reload_inner): Likewise.
1662 (rs6000_preferred_reload_class): Likewise.
1663 (rs6000_output_move_128bit): Likewise.
1665 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1667 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1668 Generate SImode target register for null target.
1669 <case IX86_BUILTIN_XGETBV>: Ditto.
1670 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1671 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1673 2018-05-10 Carl Love <cel@us.ibm.com>
1675 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1676 dcbtt and dcbtstt if operands[2] is 0.
1678 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1681 * config/i386/sse.md (usadv64qi): New expander.
1683 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1685 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1686 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1687 -maltivec=be support.
1688 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1689 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1690 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1691 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1692 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1693 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1694 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1695 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1696 altivec_vsumsws): Adjust.
1697 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1698 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1699 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1701 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1702 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1703 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1704 (altivec_lve<VI_char>x): Delete expand.
1705 (*altivec_lve<VI_char>x_internal): Rename to...
1706 (altivec_lve<VI_char>x): ... this.
1707 (altivec_lvxl_<mode>): Delete expand.
1708 (*altivec_lvxl_<mode>_internal): Rename to ...
1709 (altivec_lvxl_<mode>): ... this.
1710 (altivec_stvxl_<mode>): Delete expand.
1711 (*altivec_stvxl_<mode>_internal): Rename to ...
1712 (altivec_stvxl_<mode>): ... this.
1713 (altivec_stve<VI_char>x): Delete expand.
1714 (*altivec_stve<VI_char>x_internal): Rename to ...
1715 (altivec_stve<VI_char>x): ... this.
1716 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1717 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1718 reduc_plus_scal_<mode>): Adjust.
1719 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1721 (rs6000_cpu_cpp_builtins): Adjust.
1722 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1723 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1724 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1725 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1726 -maltivec=be support.
1727 (rs6000_split_vec_extract_var): Adjust.
1728 (rs6000_split_v4si_init): Adjust.
1729 (swap_selector_for_mode): Delete.
1730 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1731 altivec_expand_stvex_be): Delete.
1732 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1733 -maltivec=be support.
1734 (rs6000_gimple_fold_builtin): Ditto.
1735 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1737 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1738 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1739 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1740 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1741 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1742 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1743 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1744 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1745 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1746 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1747 anonymous split): Adjust.
1748 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1749 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1751 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1753 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1754 when --with-gxx-include-dir is also specified.
1755 * configure: Regenerate.
1757 2018-05-09 Jim Wilson <jimw@sifive.com>
1760 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1761 * config/riscv/t-withmultilib: New.
1762 * config/riscv/withmultilib.h: New.
1763 * doc/install.texi: Document RISC-V --with-multilib-list support.
1765 2018-05-09 Richard Biener <rguenther@suse.de>
1767 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1769 (vect_bb_vectorization_profitable_p): Adjust. Compute
1770 actual scalar cost using the cost vector and the add_stmt_cost
1773 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1775 PR rtl-optimization/85645
1776 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1777 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1779 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1781 PR rtl-optimization/85645
1782 * shrink-wrap.c (spread_components): Return a boolean saying if
1783 anything was changed.
1784 (try_shrink_wrapping_separate): Iterate spread_components until
1785 nothing changes anymore.
1787 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1789 PR rtl-optimization/85645
1790 * regrename.c (build_def_use): Also kill the chains that include the
1791 destination of a REG_CFA_REGISTER note.
1793 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1795 PR rtl-optimization/85645
1796 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1797 insn that has a REG_CFA_REGISTER note.
1799 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1801 * cfgexpand.c (expand_clobber): New function.
1802 (expand_gimple_stmt_1): Use it.
1803 * tree-vect-stmts.c (vect_clobber_variable): New function,
1805 (vectorizable_simd_clone_call): ...here.
1806 (vectorizable_store): Emit a clobber either side of an
1807 IFN_STORE_LANES sequence.
1808 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1810 2018-05-09 Tom de Vries <tom@codesourcery.com>
1813 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1814 (define_insn "trap_if_false"): Add exit after trap.
1816 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1818 PR rtl-optimization/85638
1819 * bb-reorder.c: Include common/common-target.h.
1820 (create_forwarder_block): New function extracted from...
1821 (fix_up_crossing_landing_pad): ...here. Rename into...
1822 (dw2_fix_up_crossing_landing_pad): ...this.
1823 (sjlj_fix_up_crossing_landing_pad): New function.
1824 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1825 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1826 from both partitions and exit the loop after one iteration.
1828 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1831 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1833 (Basic PowerPC Built-in Functions): The new name of the
1834 subsection previously known as "PowerPC Built-in Functions".
1835 (Basic PowerPC Built-in Functions Available on all Configurations):
1837 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1839 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1841 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1843 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1846 2018-05-08 Jim Wilson <jimw@sifive.com>
1848 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1849 (LD_EMUL_SUFFIX): New.
1850 (LINK_SPEC): Use it.
1852 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1854 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1856 (Basic PowerPC Built-in Functions): The new name of the
1857 subsection previously known as "PowerPC Built-in Functions".
1858 (Basic PowerPC Built-in Functions Available on all Configurations):
1860 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1862 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1864 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1866 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1869 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1872 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1873 after cmpelim optimization.
1875 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1877 * config.gcc: Support "goldmont".
1878 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1879 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1881 * config/i386/i386.c (m_GOLDMONT): Define.
1882 (processor_target_table): Add "goldmont".
1883 (PTA_GOLDMONT): Define.
1884 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1885 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1886 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1887 (fold_builtin_cpu): Add "goldmont".
1888 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1889 (ix86_option_override_internal): Add "goldmont".
1890 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1891 (processor_type): Add PROCESSOR_GOLDMONT.
1892 * config/i386/i386.md: Add CPU "glm".
1893 * config/i386/glm.md: New file.
1894 * config/i386/x86-tune.def: Add m_GOLDMONT.
1895 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1897 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1900 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1902 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1903 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1904 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1907 * config/i386/i386.c (ix86_fold_builtin): Handle
1908 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1911 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1912 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1914 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1917 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1918 (check_arch): Likewise.
1919 (check_fpu): Return the result rather than printing it.
1920 (end arch): Fix operator precedence.
1921 (end cpu): Likewise.
1922 (END): Print the result from check_fpu.
1924 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1925 Alan Hayward <alan.hayward@arm.com>
1926 David Sherwood <david.sherwood@arm.com>
1928 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1929 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1930 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1931 (*fcmuo<mode>_and): New patterns.
1933 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1935 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1936 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1937 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1938 (cmp_op, sve_imm_con): New code attributes.
1939 (SVE_COND_INT_CMP, imm_con): Delete.
1940 (cmp_op): Remove above unspecs from int attribute.
1941 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1943 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1944 comparison-specific unspecs.
1945 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1946 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1947 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1948 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1949 (*vec_fcm<cmp_op><mode>): Rename to...
1950 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1951 (*vec_fcmuo<mode>): Rename to...
1952 (*fcmuo<mode>): ...this and adjust likewise.
1953 (*pred_fcm<cmp_op><mode>): New pattern.
1954 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1955 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1957 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1959 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1960 (aarch64_emit_sve_predicated_cond): New function.
1961 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1962 (aarch64_emit_unspec_cond_or): Replace with...
1963 (aarch64_emit_sve_or_conds): ...this new function. Use
1964 aarch64_emit_sve_ptrue_op for the individual comparisons and
1965 aarch64_emit_binop to OR them together.
1966 (aarch64_emit_inverted_unspec_cond): Replace with...
1967 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1968 aarch64_emit_sve_ptrue_op for the comparison and
1969 aarch64_emit_unop to invert the result.
1970 (aarch64_expand_sve_vec_cmp_float): Update after the above
1971 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1973 2018-05-07 Nathan Sidwell <nathan@acm.org>
1975 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1976 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1977 (Backwards Compatibility): Likewise.
1979 2018-05-07 Luis Machado <luis.machado@linaro.org>
1983 2018-05-07 Luis Machado <luis.machado@linaro.org>
1985 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1986 <prefetch_dynamic_strides>: New const bool field.
1987 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1988 prefetch_dynamic_strides.
1989 (exynosm1_prefetch_tune): Likewise.
1990 (thunderxt88_prefetch_tune): Likewise.
1991 (thunderx_prefetch_tune): Likewise.
1992 (thunderx2t99_prefetch_tune): Likewise.
1993 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1994 (aarch64_override_options_internal): Update to set
1995 PARAM_PREFETCH_DYNAMIC_STRIDES.
1996 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1997 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1998 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1999 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2000 prefetch-dynamic-strides setting.
2002 2018-05-07 Luis Machado <luis.machado@linaro.org>
2004 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2005 <minimum_stride>: New const int field.
2006 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2007 minimum_stride field.
2008 (exynosm1_prefetch_tune): Likewise.
2009 (thunderxt88_prefetch_tune): Likewise.
2010 (thunderx_prefetch_tune): Likewise.
2011 (thunderx2t99_prefetch_tune): Likewise.
2012 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2013 (aarch64_override_options_internal): Update to set
2014 PARAM_PREFETCH_MINIMUM_STRIDE.
2015 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2016 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2017 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2018 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2019 stride is constant and is below the minimum stride threshold.
2021 2018-05-07 Luis Machado <luis.machado@linaro.org>
2023 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2026 2018-05-07 Luis Machado <luis.machado@linaro.org>
2028 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2029 <prefetch_dynamic_strides>: New const bool field.
2030 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2031 prefetch_dynamic_strides.
2032 (exynosm1_prefetch_tune): Likewise.
2033 (thunderxt88_prefetch_tune): Likewise.
2034 (thunderx_prefetch_tune): Likewise.
2035 (thunderx2t99_prefetch_tune): Likewise.
2036 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2037 (aarch64_override_options_internal): Update to set
2038 PARAM_PREFETCH_DYNAMIC_STRIDES.
2039 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2040 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2041 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2042 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2043 prefetch-dynamic-strides setting.
2045 2018-05-07 Luis Machado <luis.machado@linaro.org>
2047 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2048 <minimum_stride>: New const int field.
2049 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2050 minimum_stride field.
2051 (exynosm1_prefetch_tune): Likewise.
2052 (thunderxt88_prefetch_tune): Likewise.
2053 (thunderx_prefetch_tune): Likewise.
2054 (thunderx2t99_prefetch_tune): Likewise.
2055 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2056 (aarch64_override_options_internal): Update to set
2057 PARAM_PREFETCH_MINIMUM_STRIDE.
2058 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2059 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2060 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2061 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2062 stride is constant and is below the minimum stride threshold.
2064 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2067 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2068 the type is addressable. Don't force op into register if it has
2071 2018-05-05 Roland McGrath <mcgrathr@google.com>
2074 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2075 any section for which we don't know a specific type it should have,
2076 regardless of name. Previously this was done only for the exact
2077 names ".init_array", ".fini_array", and ".preinit_array".
2078 (default_elf_asm_named_section): Add comment about
2079 relationship with default_section_type_flags and SECTION_NOTYPE.
2080 (get_section): Don't consider it a type conflict if one side has
2081 SECTION_NOTYPE and the other doesn't, as long as neither has the
2082 SECTION_BSS et al used in the default_section_type_flags logic.
2084 2018-05-05 Tom de Vries <tom@codesourcery.com>
2087 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2088 (workaround_barsyncs): New function.
2089 (nvptx_reorg): Use workaround_barsyncs.
2090 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2091 (define_expand "nvptx_membar_cta"): New define_expand.
2092 (define_insn "*nvptx_membar_cta"): New insn.
2094 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2096 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2097 To improve optimization opportunities.
2098 * builtin-types.def: The new needed builtin types for the above.
2100 2018-05-04 Richard Biener <rguenther@suse.de>
2102 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2103 * gimple-ssa-store-merging.c
2104 (imm_store_chain_info::output_merged_store): Remove redundant create,
2105 release split_store vector contents on failure.
2106 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2107 scalar stmt vector on cache hit.
2109 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2111 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2113 * config.gcc (powerpc-xilinx-eabi*): Remove.
2114 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2116 (fusion_addis_mem_combo_load): Ditto.
2117 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2119 (rs6000_cpu_cpp_builtins): Ditto.
2120 * config/rs6000/rs6000-linux.c
2121 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2122 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2123 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2125 (rs6000_setup_reg_addr_masks): Ditto.
2126 (rs6000_init_hard_regno_mode_ok): Ditto.
2127 (rs6000_option_override_internal): Ditto.
2128 (legitimate_lo_sum_address_p): Ditto.
2129 (rs6000_legitimize_address): Ditto.
2130 (rs6000_legitimize_reload_address): Ditto.
2131 (rs6000_legitimate_address_p): Ditto.
2132 (abi_v4_pass_in_fpr): Ditto.
2133 (setup_incoming_varargs): Ditto.
2134 (rs6000_gimplify_va_arg): Ditto.
2135 (rs6000_split_multireg_move): Ditto.
2136 (rs6000_savres_strategy): Ditto.
2137 (rs6000_emit_prologue_components): Ditto.
2138 (rs6000_emit_epilogue_components): Ditto.
2139 (rs6000_emit_prologue): Ditto.
2140 (rs6000_emit_epilogue): Ditto.
2141 (rs6000_elf_file_end): Ditto.
2142 (rs6000_function_value): Ditto.
2143 (rs6000_libcall_value): Ditto.
2144 * config/rs6000/rs6000.h: Ditto.
2145 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2146 (TARGET_MINMAX): ... this. New.
2147 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2148 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2149 (*movsi_internal1_single): Delete.
2150 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2151 mfpu=, mxilinx-fpu): Delete.
2152 * config/rs6000/singlefp.h: Delete.
2153 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2154 * config/rs6000/t-rs6000: Ditto.
2155 * config/rs6000/t-xilinx: Delete.
2156 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2157 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2158 (VStype_simple): Delete.
2159 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2160 * config/rs6000/xfpu.h: Delete.
2161 * config/rs6000/xfpu.md: Delete.
2162 * config/rs6000/xilinx.h: Delete.
2163 * config/rs6000/xilinx.opt: Delete.
2164 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2165 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2167 2018-05-04 Tom de Vries <tom@codesourcery.com>
2170 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2173 2018-05-04 Richard Biener <rguenther@suse.de>
2176 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2177 (expand_complex_div_wide): Likewise.
2178 (expand_complex_operations_1): Likewise.
2179 (expand_complex_libcall): Preserve EH info of the original stmt.
2180 (tree_lower_complex): Handle removed blocks.
2181 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2182 on complex multiplication and division libcall builtins.
2184 2018-05-04 Richard Biener <rguenther@suse.de>
2187 * fold-const.c (negate_expr_p): Restrict negation of operand
2188 zero of a division to when we know that can happen without
2190 (fold_negate_expr_1): Likewise.
2192 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2195 * real.h (real_nextafter): Declare.
2196 * real.c (real_nextafter): New function.
2197 * fold-const-call.c (fold_const_nextafter): New function.
2198 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2199 CASE_CFN_NEXTTOWARD.
2200 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2201 even when arg1_mode is different from arg0_mode.
2203 2018-05-03 Nathan Sidwell <nathan@acm.org>
2205 * doc/extend.texi (Deprecated Features): Remove
2207 (Backwards Compatibility): Likewise.
2208 * doc/invoke.texi (C++ Language Options): Likewise.
2209 (C++ Dialect Options): Likewise.
2211 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2214 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2215 _mm512_mask_mullox_epi64): New intrinsics.
2217 2018-05-03 Tom de Vries <tom@codesourcery.com>
2220 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2221 dump files): Add offload-tree.
2223 2018-05-03 Richard Biener <rguenther@suse.de>
2225 PR tree-optimization/85615
2226 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2227 to loops not nested in BBs loop father to avoid creating multi-entry
2230 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2232 PR tree-optimization/70291
2233 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2234 arguments. Change return type to tree. Emit libcall as a new
2235 statement rather than replacing existing one when inplace_p is true.
2236 (expand_complex_multiplication_components): New function.
2237 (expand_complex_multiplication): Expand floating-point complex
2238 multiplication using the above.
2239 (expand_complex_division): Rename inner_type parameter to type.
2240 Update expand_complex_libcall call-site.
2241 (expand_complex_operations_1): Update expand_complex_multiplication
2242 and expand_complex_division call-sites.
2244 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2247 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2248 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2249 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2250 the highest significant bit of the shift count mask is clear. In
2251 check whether and[sq]i3 is needed verify that all significant bits
2252 of the shift count other than the highest are set.
2254 2018-05-02 Tom de Vries <tom@codesourcery.com>
2257 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2258 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2259 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2260 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2261 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2262 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2263 __builtin_goacc_parlevel_size.
2265 2018-05-02 Richard Biener <rguenther@suse.de>
2267 PR tree-optimization/85597
2268 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2269 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2272 2018-05-02 Tom de Vries <tom@codesourcery.com>
2275 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2276 dump files): Add ltrans-tree.
2278 2018-05-02 Tom de Vries <tom@codesourcery.com>
2281 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2282 dump files): Add wpa-ipa.
2284 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2286 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2287 powerpc*-*-linux*paired* target.
2288 * config/rs6000/750cl.h: Delete.
2289 * config/rs6000/paired.h: Delete.
2290 * config/rs6000/paired.md: Delete.
2291 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2293 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2294 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2295 comment. Remove paired float support.
2296 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2297 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2299 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2300 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2302 * config/rs6000/rs6000.c: Remove paired float support.
2303 (paired_expand_vector_init, paired_expand_vector_move,
2304 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2305 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2306 paired_expand_builtin, paired_expand_predicate_builtin,
2307 paired_init_builtins): Delete.
2308 * config/rs6000/rs6000.h: Remove paired float support.
2309 * config/rs6000/rs6000.md: Remove paired float support.
2310 (move_from_CR_ov_bit): Delete.
2311 * config/rs6000/rs6000.opt (mpaired): Delete.
2312 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2313 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2315 2018-05-02 Richard Biener <rguenther@suse.de>
2318 * gimplify.c (gimplify_save_expr): When in SSA form allow
2319 SAVE_EXPRs to compute to SSA vars.
2321 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2324 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2325 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2326 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2327 clobber operands[2], instead use a new pseudo. Formatting fixes.
2329 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2331 PR tree-optimization/85586
2332 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2333 exit early for statements in the same group if the accesses are
2336 2018-05-02 Tom de Vries <tom@codesourcery.com>
2339 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2342 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2344 PR tree-optimization/85143
2345 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2347 2018-05-01 Tom de Vries <tom@codesourcery.com>
2350 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2351 not found" error message.
2353 2018-05-01 Tom de Vries <tom@codesourcery.com>
2356 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2357 * vec.c (test_ordered_remove_if): New function.
2358 (vec_c_tests): Call test_ordered_remove_if.
2359 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2360 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2361 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2362 VEC_ORDERED_REMOVE_IF.
2364 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2366 PR tree-optimization/82665
2367 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2368 pointer subtraction where arguments come from a memchr call.
2370 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2372 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2373 --push-state --as-needed and --pop-state instead of --as-needed and
2374 --no-as-needed if ld supports it.
2375 * configure: Regenerated.
2378 * doc/install.texi2html: Replace _002d with - and _002a with * in
2379 generated html files using sed.
2381 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2384 * gcc-rich-location.c (blank_line_before_p): New function.
2385 (use_new_line): New function.
2386 (gcc_rich_location::add_fixit_insert_formatted): New function.
2387 * gcc-rich-location.h
2388 (gcc_rich_location::add_fixit_insert_formatted): New function.
2390 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2392 * selftest.c (assert_streq): Rename "expected" and "actual" to
2393 "val1" and "val2". Extend NULL-handling to cover both inputs
2394 symmetrically, while still requiring both to be non-NULL for a pass.
2395 * selftest.h (assert_streq): Rename "expected" and "actual" to
2397 (ASSERT_EQ): Likewise.
2398 (ASSERT_EQ_AT): Likewise.
2399 (ASSERT_KNOWN_EQ): Likewise.
2400 (ASSERT_KNOWN_EQ_AT): Likewise.
2401 (ASSERT_NE): Likewise.
2402 (ASSERT_MAYBE_NE): Likewise.
2403 (ASSERT_MAYBE_NE_AT): Likewise.
2404 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2405 the assertion to pass.
2406 (ASSERT_STREQ_AT): Likewise.
2408 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2410 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2411 interaction with -pie.
2413 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2415 * selftest.h: Fix alphabetization of per-source-file selftest
2418 2018-04-30 Jason Merrill <jason@redhat.com>
2420 PR c++/61982 - dead stores to destroyed objects.
2421 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2424 2018-04-30 Jason Merrill <jason@redhat.com>
2426 * tree.c (build_clobber): New.
2427 * tree.h: Declare it.
2428 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2430 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2432 * diagnostic-show-locus.c (layout::layout): Update for
2433 location_get_source_line returning a char_span.
2434 (struct char_span): Move to input.h.
2435 (struct correction): Update for fields in char_span becoming
2437 (struct source_line): Update for location_get_source_line
2438 returning a char_span.
2439 (layout::print_line): Likewise.
2440 * edit-context.c (edited_file::print_content): Likewise.
2441 (edited_file::print_diff_hunk): Likewise.
2442 (edited_file::print_run_of_changed_lines): Likewise.
2443 (edited_file::get_num_lines): Likewise.
2444 (edited_line::edited_line): Likewise.
2445 * final.c (asm_show_source): Likewise.
2446 * input.c (location_get_source_line): Convert return type
2447 from const char * to char_span, losing the final "line_len"
2449 (dump_location_info): Update for the above.
2450 (get_substring_ranges_for_loc): Likewise. Use a char_span
2451 when handling the literal within the line.
2452 (test_reading_source_line): Update for location_get_source_line
2453 returning a char_span.
2454 * input.h (class char_span): Move here from
2455 diagnostic-show-locus.c, converting from a struct to a class.
2456 Make data members private.
2457 (char_span::operator bool): New.
2458 (char_span::length): New.
2459 (char_span::get_buffer): New.
2460 (char_span::operator[]): New.
2461 (char_span::subspan): Make const.
2462 (char_span::xstrdup): New.
2463 (location_get_source_line): Convert return type from const char *
2464 to char_span, losing the final "line_size" param.
2466 2018-04-30 Jan Hubicka <jh@suse.cz>
2468 * lto-wrapper.c (ltrans_priorities): New static var.
2469 (cmp_priority): New.
2470 (run_gcc): Read priorities and if doing parallel build order
2471 the Makefile by them.
2473 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2475 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2477 2018-04-30 Richard Biener <rguenther@suse.de>
2479 * tree-cfg.c (verify_address): Remove base argument, add
2480 flag whether to check TREE_ADDRESSABLE and do that.
2481 (verify_expr): Remove.
2482 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2483 (verify_gimple_assign_single): Likewise.
2484 (verify_gimple_switch): Likewise.
2485 (verify_expr_location_1): Dereference tp once. Add (disabled)
2486 piece from verify_expr.
2487 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2489 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2491 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2493 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2495 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2496 (small_data_pattern): Likewise.
2497 (arc_rewrite_small_data): Likewise.
2498 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2499 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2500 (get_symbol_alignment): New function.
2501 (legitimate_small_data_address_p): Likewise.
2502 (legitimate_scaled_address): Update, call
2503 legitimate_small_data_address_p.
2504 (output_sdata): New static variable.
2505 (arc_print_operand): Update how we handle small data operands.
2506 (arc_print_operand_address): Likewise.
2507 (arc_legitimate_address_p): Update, use
2508 legitimate_small_data_address_p.
2509 (arc_rewrite_small_data_p): Remove.
2510 (arc_rewrite_small_data_1): Likewise.
2511 (arc_rewrite_small_data): Likewise.
2512 (small_data_pattern): Likewise.
2513 (compact_sda_memory_operand): Update to use
2514 legitimate_small_data_address_p and get_symbol_alignment.
2515 (prepare_move_operands): Don't rewite sdata pattern.
2516 (prepare_extend_operands): Remove.
2517 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2519 (zero_extendqisi2): Likewise.
2520 (zero_extendhisi2): Likewise.
2521 (extendqihi2): Likewise.
2522 (extendqisi2): Likewise.
2523 (extendhisi2): Likewise.
2527 * config/arc/constraints.md (Usd): Change it to memory constraint.
2529 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2531 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2532 as source of std instructions.
2533 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2534 allow 6-bit constants as source for store instructions.
2535 (movdi_insn): Update instruction pattern to allow 6-bit constants
2536 as source for store instructions.
2538 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2540 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2542 2018-04-30 Nathan Sidwell <nathan@acm.org>
2543 Sandra Loosemore <sandra@codesourcery.com>
2545 * dumpfile.c (dump_open): Allow '-' for stdout.
2546 * doc/invoke.texi (Developer Options): Document dump filename
2547 determination early. Document stdin/stdout selection.
2549 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2551 Microblaze Target: PIC data text relative
2553 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2554 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2556 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2557 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2558 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2559 New addressing mode for data-text relative position indepenedent code.
2560 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2561 'ADDRESS_SYMBOLIC_TXT_REL'.
2562 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2563 (microblaze_legitimate_pic_operand): Exclude function calls from
2564 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2565 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2567 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2568 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2569 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2570 for 'address + offset'.
2571 (microblaze_expand_prologue): Add new function prologue call for
2573 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2574 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2575 table in case of TARGET_PIC_DATA_TEXT_REL.
2576 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2577 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2578 Add new macros 'UNSPEC_TEXT',
2579 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2580 + exclude function calls from 'UNSPEC_PLT' in case of data text
2582 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2583 new target hook for generating address diff vector tables in case of
2585 * doc/tm.texi : Regenerate.
2586 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2587 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2588 of addr diff vector generation.
2589 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2590 target hook definition.
2591 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2592 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2593 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2594 option for data text relative.
2596 2018-04-30 Richard Biener <rguenther@suse.de>
2598 * tree-chrec.h (evolution_function_is_constant_p): Remove
2600 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2602 2018-04-30 Richard Biener <rguenther@suse.de>
2605 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2607 2018-04-30 Richard Biener <rguenther@suse.de>
2609 PR tree-optimization/28364
2610 PR tree-optimization/85275
2611 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2612 copying first exit test.
2614 2018-04-28 Mark Wielaard <mark@klomp.org>
2616 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2618 (dwarf_AT): Handle DW_AT_addr_base.
2619 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2621 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2624 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2625 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2626 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2627 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2629 2018-04-28 Richard Biener <rguenther@suse.de>
2631 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2632 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2633 to reflect use. Only add interesting stmts.
2635 2018-04-27 Martin Jambor <mjambor@suse.cz>
2638 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2639 the jump function allows for passing through aggregate values.
2641 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2643 * input.h (in_system_header_at): Convert from macro to inline
2645 (from_macro_expansion_at): Likewise.
2646 (from_macro_definition_at): Likewise.
2648 2018-04-27 Jeff Law <law@redhat.com>
2650 * config.gcc: Mark tile* targets as deprecated/obsolete.
2652 2018-04-27 Richard Biener <rguenther@suse.de>
2654 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2657 2018-04-27 Richard Biener <rguenther@suse.de>
2659 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2661 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2663 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2664 with Yd constraint. Set "preferred_for_speed" attribute from
2665 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2667 (*movdi_internal): Ditto.
2668 (movti_interunit splitters): Remove
2669 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2670 (movdi_interunit splitters): Ditto.
2671 * config/i386/constraints.md (Ye): Remove.
2672 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2674 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2677 * config/aarch64/constraints.md (Usg): Limit to 31.
2680 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2682 PR tree-optimization/85529
2683 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2684 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2685 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2686 zero extension or masking of the MSB bit.
2687 (optimize_range_tests): Add FIRST_BB argument, pass it through
2688 to optimize_range_tests_var_bound.
2689 (maybe_optimize_range_tests, reassociate_bb): Adjust
2690 optimize_range_tests callers.
2692 2018-04-26 Richard Biener <rguenther@suse.de>
2693 Jakub Jelinek <jakub@redhat.com>
2695 * cgraph.h (symbol_table): Just declare debug method here.
2696 * symtab.c (symbol_table::debug): Define.
2698 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2700 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2702 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2704 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2705 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2706 (*movdi_internal): Substitute Yi and Yj constraint with x
2707 and Ym and Yn constraint with y constraint. Update "isa"
2708 attribute and set "preferred_for_speed" attribute from
2709 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2710 (*movsi_internal): Ditto.
2711 (*movdf_internal): Ditto.
2712 (*movsf_internal): Ditto.
2713 (*zero_extendsidi2): Ditto.
2714 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2715 (sse2_loadld): Ditto.
2716 (*vec_extract<ssevecmodelower>_0): Ditto.
2717 (*vec_extractv4si_0_zext_sse4): Ditto.
2718 (vec_concatv2di): Ditto.
2719 (*vec_dup<mode>): Ditto.
2720 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2721 * config/i386/constraints.md (Yi): Remove.
2726 2018-04-26 Nathan Sidwell <nathan@acm.org>
2728 * dumpfile.c (dump_open): New.
2729 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2730 (dump_finish): Detect stdio/stderr by value not name.
2732 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2734 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2736 2018-04-26 Tom de Vries <tom@codesourcery.com>
2739 * config/nvptx/nvptx.c (verify_neutering_jumps)
2740 (verify_neutering_labels): New function
2741 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2743 2018-04-26 Tom de Vries <tom@codesourcery.com>
2746 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2747 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2750 2018-04-26 Richard Biener <rguenther@suse.de>
2751 Tom de Vries <tom@codesourcery.com>
2754 * lto-streamer-out.c (output_function): Fixup loops if required to match
2755 discovery done in the reader.
2757 2018-04-26 Richard Biener <rguenther@suse.de>
2759 PR tree-optimization/85116
2760 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2761 have a loop exit from the single latch predecessor. Remove
2762 case of header with just condition.
2763 (ch_base::copy_headers): Exclude infinite loops from any
2765 (pass_ch::execute): Record exits.
2767 2018-04-26 Richard Biener <rguenther@suse.de>
2769 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2770 prologue cost vector and pass it to vect_get_load_cost.
2771 (vect_get_peeling_costs_all_drs): Likewise.
2772 (vect_peeling_hash_get_lowest_cost): Likewise.
2773 (vect_enhance_data_refs_alignment): Likewise.
2775 2018-04-26 Richard Biener <rguenther@suse.de>
2778 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2779 checking of integer<->pointer conversions.
2780 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2781 sign-/zero-extending pointer types.
2782 (expand_omp_for_static_chunk): Likewise.
2784 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2785 Jean Lee <xiaoyur347@gmail.com>
2787 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2788 (TARGET_ASAN_SHADOW_OFFSET): Define.
2789 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2790 true for -fsanitize=address.
2792 2018-04-25 Mark Wielaard <mark@klomp.org>
2794 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2797 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2799 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2800 than "alu", remove explicit "memory" and "imm_disp" attributes.
2801 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2804 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2805 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2806 gen_lowpart_no_emit.
2808 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2811 * config/i386/i386.c (ix86_expand_builtin): Change memory
2812 operand to XI, extend p0 to Pmode.
2813 * config/i386/i386.md: Change unspec volatile and operand
2814 1 mode to XI, change operand 0 mode to P.
2816 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2818 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2819 GET_MODE_MASK before any checking.
2820 (nds32_can_use_bset_p): Likewise.
2821 (nds32_can_use_btgl_p): Likewise.
2823 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2825 * config/nds32/nds32-doubleword.md: New define_split pattern for
2826 illegal register number.
2828 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2830 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2832 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2834 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2836 2018-04-25 Richard Biener <rguenther@suse.de>
2838 * lto-streamer.h (LTO_major_version): Bump to 8.
2840 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2842 * BASE-VER: Set to 9.0.0.
2844 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2846 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2847 in __abskf2 and __powikf2.
2849 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2852 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2853 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2854 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2855 Use the above on operand 2. Reindent.
2856 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2858 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2861 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2863 * config/i386/i386.opt (mcet): Removed.
2864 * doc/install.texi: Remove -mcet documentation.
2865 * doc/invoke.texi: Likewise.
2867 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2870 * doc/install.texi: Remove -mcet from bootstrap-cet.
2872 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2875 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2876 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2880 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2881 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2882 containing a CONST_VECTOR.
2884 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2886 * doc/install.texi: Update newlib dependency for nvptx.
2888 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2891 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2892 instead of INTVAL when shifting x left.
2894 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2896 PR tree-optimization/85478
2897 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2898 vect_grouped_store_supported for single element vectors.
2900 2018-04-24 Richard Biener <rguenther@suse.de>
2903 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2904 load cost increase to the case of non-constant step.
2906 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2909 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2910 destination if any_malformed_asm.
2912 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2915 * expr.c (store_field): In the bitfield case, if the value comes from
2916 a function call and is returned in registers by means of a PARALLEL,
2917 do not change the mode of the temporary unless BLKmode and VOIDmode.
2919 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2921 PR rtl-optimization/85423
2922 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2923 dependencies to debug insns when the previous insn is non-debug.
2925 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2927 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2928 enums into a single definition.
2929 (fls): Fix predicates and printing.
2932 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2934 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2935 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2936 and short u6 immediate.
2937 (check_if_valid_sleep_operand): Remove.
2938 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2940 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2942 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2943 flag_always_save_lp condition.
2944 * config/nds32/nds32.opt (malways-save-lp): New option.
2946 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2948 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2949 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2950 * config/nds32/nds32.h
2951 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2952 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2954 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2956 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2957 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2959 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2960 Chung-Ju Wu <jasonwucj@gmail.com>
2962 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2964 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2965 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2967 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2969 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2971 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2973 * config/nds32/nds32-protos.h (nds32_data_alignment,
2974 nds32_local_alignment): Declare.
2975 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2976 nds32_local_alignment): New functions.
2977 (TARGET_CONSTANT_ALIGNMENT): Define.
2978 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2980 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2982 * config/nds32/nds32.c
2983 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2984 (TARGET_MODES_TIEABLE_P): Likewise.
2986 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2988 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2991 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2992 Chung-Ju Wu <jasonwucj@gmail.com>
2994 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2995 for unaligned access.
2996 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2998 * config/nds32/nds32-intrinsic.md: Likewise.
2999 * config/nds32/nds32_intrinsic.h: Likewise.
3000 * config/nds32/nds32.h (nds32_builtins): Likewise.
3001 * config/nds32/nds32.opt (munaligned-access): New option.
3002 * config/nds32/nds32.c (nds32_asm_file_start): Display
3003 flag_unaligned_access status.
3005 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3007 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3008 -mno-relax is present.
3009 * config/riscv/linux.h (LINK_SPEC): Ditto.
3011 2018-04-20 Martin Sebor <msebor@redhat.com>
3014 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3016 (gimple_fold_builtin_stxcpy_chk): Same.
3017 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3019 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3022 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3023 __powikf2 when long double is IEEE 128-bit.
3025 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3027 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3028 step to make sure stack always aligned.
3030 2018-04-20 Carl Love <cel@us.ibm.com>
3033 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3034 size check for arg0.
3036 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3037 Tom de Vries <tom@codesourcery.com>
3040 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3041 Emit insns for calls too.
3042 (nvptx_find_par): Always look for worker-level predecessor insn.
3043 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3045 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3046 (nvptx_process_pars): Propagate frames for calls.
3048 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3051 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3053 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3054 (ix86_handle_option): Don't handle OPT_mibt.
3055 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3057 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3059 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3060 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3061 (ix86_target_macros): Define __CET__ with flag_cf_protection
3062 for -fcf-protection.
3063 * config/i386/i386.c (isa2_opts): Remove -mibt.
3064 * config/i386/i386.h (TARGET_IBT): Removed.
3065 (TARGET_IBT_P): Likewise.
3066 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3067 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3068 * config/i386/i386.opt (mcet): Update help message.
3071 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3072 -mcet as an alias for -mshstk.
3074 2018-04-20 Richard Biener <rguenther@suse.de>
3077 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3078 complexity by forcing a single use of the multiply operand.
3080 2018-04-20 Martin Jambor <mjambor@suse.cz>
3083 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3084 recursion dependency to only apply to non-clones.
3086 2018-04-20 Martin Jambor <mjambor@suse.cz>
3089 * ipa-cp.c (create_specialized_node): Check that clones of
3090 self-recursive edges exist during IPA-CP.
3092 2018-04-19 Toon Moene <toon@moene.org>
3094 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3097 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3099 PR tree-optimization/85467
3100 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3101 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3102 VECTOR_CST element to type.
3104 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3107 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3108 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3109 (builtin_longjmp): Likewise.
3110 (save_stack_nonlocal): New pattern.
3111 (restore_stack_nonlocal): Likewise.
3113 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3116 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3117 Replace ASM_OUTPUT_LABEL with fprintf.
3119 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3122 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3123 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3124 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3125 define __IBT__ and __SHSTK__ for -fcf-protection.
3126 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3128 (ix86_trampoline_init): Likewise.
3129 (x86_output_mi_thunk): Likewise.
3130 (ix86_notrack_prefixed_insn_p): Likewise.
3131 (ix86_option_override_internal): Don't disallow -fcf-protection.
3132 * config/i386/i386.md (rdssp<mode>): Also enable for
3134 (incssp<mode>): Likewise.
3135 (nop_endbr): Likewise.
3136 * config/i386/i386.opt (mcet): Change help message to built-in
3140 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3141 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3142 enable CET built-in functions.
3144 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3146 * common/config/i386/i386-common.c
3147 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3148 OPTION_MASK_ISA_MOVDIRI_UNSET,
3149 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3150 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3151 * config.gcc (movdirintrin.h): New header.
3152 * config/i386/cpuid.h (bit_MOVDIRI,
3153 bit_MOVDIR64B): New bits.
3154 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3156 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3157 (VOID, PVOID, PCVOID)): New function types.
3158 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3159 __builtin_ia32_directstoreu_u64,
3160 __builtin_ia32_movdir64b): New builtins.
3161 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3162 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3164 (ix86_valid_target_attribute_inner_p): Ditto.
3165 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3166 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3167 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3168 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3169 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3170 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3171 (movdiri<mode>, movdir64b_<mode>): New.
3172 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3173 * config/i386/immintrin.h: Include movdirintrin.h.
3174 * config/i386/movdirintrin.h: New file.
3175 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3177 2018-04-19 Richard Biener <rguenther@suse.de>
3180 * cfg.c (clear_bb_flags): When loop state says we have
3181 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3183 2018-04-19 Richard Biener <rguenther@suse.de>
3185 PR tree-optimization/84737
3186 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3187 copying restrict info.
3188 (vect_setup_realignment): Use it.
3189 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3190 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3191 the first DR to all generated stores.
3192 (vectorizable_load): Likewise for loads.
3194 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3196 PR tree-optimization/85446
3197 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3198 the integral and pointer types to have the same precision.
3200 * doc/install.texi: Document --disable-cet being the default and
3203 2018-04-18 Martin Liska <mliska@suse.cz>
3205 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3208 2018-04-18 Martin Liska <mliska@suse.cz>
3211 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3214 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3215 arguments if they are comparable.
3217 2018-04-18 Martin Liska <mliska@suse.cz>
3220 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3223 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3226 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3229 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3230 ENDBR after calling __morestack.
3232 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3235 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3236 by using gcc_base_ver to generate a gcc_driver_version, and use
3237 it when generating GCC_DRIVER_NAME.
3238 * configure: Regenerate.
3240 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3243 * config.gcc: Obsolete powerpc*-*-*spe*.
3245 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3248 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3249 (stabstr_D): Change type of unum from unsigned int to
3250 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3253 2018-04-17 Jim Wilson <jimw@sifive.com>
3256 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3257 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3258 Set arg_pointer_offset after using pretend_args_size.
3260 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3262 PR rtl-optimization/85431
3263 * dse.c (record_store): Ignore zero width stores.
3266 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3267 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3268 __builtin_stack_restore rather than after it.
3269 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3270 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3271 argument instead of virtual_dynamic_stack_rtx.
3273 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3275 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3277 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3278 Add note to error message to explain internal mapping of overloaded
3279 built-in function name to non-overloaded built-in function name.
3280 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3283 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3286 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3287 where the inputs overlap with the output.
3289 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3291 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3292 (=v, v) alternative and explicit "memory" attribute.
3293 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3294 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3296 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3297 "sselog1" type instead of "sselog".
3298 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3299 "sselog". Remove explicit "memory" attribute.
3300 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3301 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3303 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3304 "sselog1" type instead of "sselog". Remove explicit "memory"
3306 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3307 use "sselog1" type instead of "sselog". Remove explicit "memory"
3309 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3310 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3312 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3313 "sselog1" type instead of "sselog". Remove explicit "memory"
3315 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3316 use "sselog1" type instead of "sselog". Remove explicit "memory"
3320 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3323 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3326 2018-04-17 Martin Jambor <mjambor@suse.cz>
3329 * ipa-cp.c (create_specialized_node): Call
3330 expand_all_artificial_thunks if necessary.
3332 2018-04-17 Martin Liska <mliska@suse.cz>
3335 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3336 in message, remote space in between '_G' and '('.
3338 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3341 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3342 avx512f_vmcmp<mode>3<round_saeonly_name>,
3343 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3344 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3345 avx512f_rndscale<mode><round_saeonly_name>,
3346 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3347 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3348 Use %<iptr>2 instead of %2 for -masm=intel.
3349 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3350 avx512f_vcvttss2usi<round_saeonly_name>,
3351 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3353 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3354 avx512f_vcvttsd2usi<round_saeonly_name>,
3355 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3356 Use %q1 instead of %1 for -masm=intel.
3357 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3358 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3359 of %3 for -masm=intel.
3360 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3362 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3364 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3366 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3367 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3369 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3371 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3372 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3373 %g1 and one with %0 and %1.
3374 (avx512er_vmrcp28<mode><round_saeonly_name>,
3375 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3377 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3378 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3379 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3380 of %0 and %{%4%} for -masm=intel.
3381 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3382 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3383 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3384 order of %0 and %{%5%}%{z%} for -masm=intel.
3386 2018-04-17 Jan Hubicka <jh@suse.cz>
3389 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3391 2018-04-17 Martin Liska <mliska@suse.cz>
3394 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3395 for target_clone error message. Make default implementation
3396 clone to be a local declaration.
3397 (separate_attrs): Add new argument and check for an empty
3399 (expand_target_clones): Handle it.
3400 (ipa_target_clone): Make redirection just for target_clones
3403 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3404 Tom de Vries <tom@codesourcery.com>
3407 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3408 tiled basic blocks without omp continue statements.
3410 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3413 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3414 vec_extract expression as having side effects to make sure it gets
3417 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3420 * config/i386/i386.c (get_builtin_code_for_version): Check
3423 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3426 * gcc/config.gcc: Support "skylake".
3427 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3429 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3430 (processor_target_table): Add "skylake".
3431 (ix86_option_override_internal): Add "skylake".
3432 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3433 PROCESSOR_CANNONLAKE.
3434 (get_builtin_code_for_version): Fix priority for
3435 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3436 PROCESSOR_SKYLAKE-AVX512.
3437 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3438 (processor_type): Add PROCESSOR_SKYLAKE.
3440 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3441 Jason Merrill <jason@redhat.com>
3444 * convert.c (convert_to_integer_1): Use direct recursion for
3445 enumeral types and types with a precision less than the number
3446 of bits in their mode.
3448 2018-04-16 Julia Koval <julia.koval@intel.com>
3451 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3452 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3454 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3457 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3458 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3459 and -mno-direct-move.
3461 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3464 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3465 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3466 Ensure negative shifts result in {0}.
3468 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3470 PR rtl-optimization/79916
3471 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3472 regs (if any) to define how to gnerate SD moves when LRA is in
3475 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3477 PR rtl-optimization/85393
3478 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3479 * except.c (expand_dw2_landing_pad_for_region): Make static.
3480 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3481 a label and unconditional jump to old_bb, rather than
3482 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3485 PR rtl-optimization/85376
3486 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3487 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3488 instead of a specific value.
3490 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3491 Bin Cheng <bin.cheng@arm.com>
3493 PR tree-optimization/82965
3494 PR tree-optimization/83991
3495 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3496 by_profile_only parameter.
3497 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3498 information if the loop was predicted to iterate too many times.
3499 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3501 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3504 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3507 2018-04-13 Martin Liska <mliska@suse.cz>
3508 Jakub Jelinek <jakub@redhat.com>
3511 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3512 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3513 * builtins.c (expand_builtin_memory_copy_args): Use
3514 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3515 handle dest_addr == pc_rtx.
3517 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3520 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3521 asked to not generate direct moves.
3522 (fix_trunc<mode>si2_stfiwx): Similar.
3523 (fix_trunc<mode>si2_internal): Similar.
3525 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3528 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3529 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3530 lookup if dest in some wider mode is known to be const0_rtx and
3531 if so, record permanent equivalence for it to be ZERO_EXTEND of
3532 the narrower mode destination.
3534 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3536 * lto-streamer-out.c (output_function): Revert 259346.
3537 * omp-expand.c (expand_oacc_for): Likewise.
3539 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3541 PR rtl-optimization/85354
3542 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3543 * sel-sched.c (sel_global_init): ... here.
3545 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3548 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3549 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3550 mode for PE-COFF targets.
3551 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3552 (i386_pe_asm_lto_end): Likewise.
3553 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3554 (TARGET_ASM_LTO_END): Likewise.
3555 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3556 (i386_pe_asm_lto_start): New function.
3557 (i386_pe_asm_lto_end): Likewise.
3559 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3560 Richard Biener <rguenther@suse.de>
3563 * lto-streamer-out.c (output_function): Fix CFG loop state before
3565 * omp-expand.c (expand_oacc_for): Handle calls to internal
3566 functions like regular functions.
3568 2018-04-12 Richard Biener <rguenther@suse.de>
3571 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3572 for the early LTO debug to properly generate references to it
3573 during DIE emission. Do not re-use that for the skeleton for
3575 (dwarf2out_early_finish): Likewise.
3577 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3580 * config/i386/sse.md
3581 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3582 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3583 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3584 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3585 and output is a reg, avoid creating invalid lowpart subreg, but
3586 instead split into a 512-bit move. Don't split if not AVX512VL,
3587 input is xmm16+ reg and output is a mem.
3588 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3589 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3590 xmm16+ reg and output is a mem.
3592 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3594 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3595 also for flag_dwarf2_cfi_asm.
3597 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3599 PR rtl-optimization/85342
3600 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3601 a bool scalar var inside of the loop instead. Don't try to update
3602 recog_data.operand after failed apply_change_group.
3604 2018-04-12 Tom de Vries <tom@codesourcery.com>
3607 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3608 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3609 array with flexible array member as array without given dimension.
3610 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3611 argument for undefined param to true.
3613 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3616 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3617 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3618 from PowerPC section.
3619 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3620 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3621 help text that is too long.
3622 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3623 help text that is too long.
3624 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3625 help text that is too long.
3627 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3629 * config/alpha/alpha.md (stack_probe_internal): Rename
3630 from "probe_stack". Update all callers.
3632 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3634 PR rtl-optimization/84566
3635 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3636 sched_macro_fuse_insns.
3638 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3641 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3642 (compute_block_dependences): ... from here.
3644 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3646 PR tree-optimization/85331
3647 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3648 from int to HOST_WIDE_INT.
3650 2018-04-11 Martin Jambor <mjambor@suse.cz>
3653 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3654 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3655 not the same as the source val.
3656 (cgraph_edge_brings_value_p): New parameter.
3657 (gather_edges_for_value): Pass destination value to
3658 cgraph_edge_brings_value_p.
3659 (perhaps_add_new_callers): Likewise.
3660 (get_info_about_necessary_edges): Likewise and exclude values brought
3661 only by self-recursive edges.
3662 (create_specialized_node): Redirect only clones of self-calling edges.
3663 (+self_recursive_pass_through_p): New function.
3664 (find_more_scalar_values_for_callers_subset): Use it.
3665 (find_aggregate_values_for_callers_subset): Likewise.
3666 (known_aggs_to_agg_replacement_list): Removed.
3667 (decide_whether_version_node): Re-calculate known constants for all
3668 remaining context clones.
3670 2018-04-11 Richard Biener <rguenther@suse.de>
3673 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3674 from early DWARF output.
3675 (dwarf2out_early_finish): Output line info unconditionally into
3676 early DWARF and add reference to it.
3678 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3681 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3682 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3683 other than V2DFmode using iptr mode attribute.
3684 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3686 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3688 PR rtl-optimization/84659
3689 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3691 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3694 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3696 (output_loc_list): Pass address of a dummy size variable even in the
3697 locview handling loop.
3698 (index_location_lists): Add comment on why skip_loc_list_entry can't
3701 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3704 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3707 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3710 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3711 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3712 and -mstring-compare-inline-limit.
3714 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3717 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3718 for stack clash protection in a register whenever we need it to be in
3721 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3723 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3724 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3726 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3729 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3731 (mlong-double-): Ditto.
3732 * config/rs6000/sysv4.opt (msdata=): Ditto.
3733 (mtls-size=): Ditto.
3735 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3737 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3738 erroneous entries for
3739 "vector int vec_ldl (int, long int *)", and
3740 "vector unsigned int vec_ldl (int, unsigned long int *)".
3741 Add comments and entries for
3742 "vector bool char vec_ldl (int, bool char *)",
3743 "vector bool short vec_ldl (int, bool short *)",
3744 "vector bool int vec_ldl (int, bool int *)",
3745 "vector bool long long vec_ldl (int, bool long long *)",
3746 "vector pixel vec_ldl (int, pixel *)",
3747 "vector long long vec_ldl (int, long long *)",
3748 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3749 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3750 type tree bool_long_long_type_node and correct definition of
3751 bool_V2DI_type_node to make reference to this new type tree.
3752 (rs6000_mangle_type): Replace erroneous reference to
3753 bool_long_type_node with bool_long_long_type_node.
3754 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3755 comments to emphasize sign distinctions for char and int types and
3756 replace RS6000_BTI_bool_long constant with
3757 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3758 use of RS6000_BTI_pixel.
3759 (bool_long_type_node): Remove this macro definition.
3760 (bool_long_long_type_node): New macro definition
3762 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3764 PR rtl-optimization/85300
3765 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3766 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3767 simplify_unary_operation fails.
3769 2018-04-10 Martin Liska <mliska@suse.cz>
3771 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3772 cgraph_edge and ipa_ref.
3774 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3778 * config/i386/sse.md
3779 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3780 computation of the VEC_MERGE selector from mask.
3781 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3782 Fix decoding of the VEC_MERGE selector into mask.
3784 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3786 PR tree-optimization/85286
3787 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3789 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3791 * final.c (final_1): Set insn_last_address as well as
3792 insn_current_address.
3794 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3797 * explow.c (emit_stack_probe): Call validize_mem on memory location
3798 before passing it to gen_probe_stack. Create address operand and
3799 legitimize it for the probe_stack_address case.
3801 2018-04-09 Jan Hubicka <jh@suse.cz>
3804 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3805 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3806 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3808 (free_lang_data): Rebuild type inheritance graph.
3810 2018-04-09 Martin Sebor <msebor@redhat.com>
3812 * invoke.texi (-finline-small-functions): Mention other optimization
3814 (-findirect-inlining, -fpartial-inlining): Same.
3815 (-finline-functions-called-once): Same.
3816 (-freorder-blocks-and-partition): Same.
3818 2018-04-09 Jan Hubicka <jh@suse.cz>
3821 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3822 jumps; choose last target that matches the criteria (i.e.
3823 no partition changes for non-crossing jumps).
3824 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3825 support for redirecting crossing jumps to non-crossing.
3827 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3829 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3830 also for naked functions.
3832 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3834 * config/arc/arc.md (add_shift): New pattern.
3835 (add_shift2): Likewise.
3836 (sub_shift): Likewise.
3837 (sub_shift_cmp0_noout): Likewise.
3838 (compare_si_ashiftsi): Likewise.
3839 (xbfu_cmp0_noout): New combine pattern.
3840 (xbfu_cmp0"): Likewise.
3841 (movsi_set_cc_insn): Place the predicable variant first.
3842 (commutative_binary_cmp0_noout): Remove clobber.
3843 (commutative_binary_cmp0): New pattern.
3844 (noncommutative_binary_cmp0): Likewise.
3845 (noncommutative_binary_cmp0_noout): Likewise.
3846 (noncommutative_binary_comparison_result_used): Removed.
3847 (rsub_cmp0): New pattern.
3848 (rsub_cmp0_noout): Likewise.
3849 (extzvsi): Changed, keep only meaningful variants.
3850 (SQH, SEZ): New iterators.
3851 (SQH_postfix): New mode attribute.
3852 (SEZ_prefix): New code attribute.
3853 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3854 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3855 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3857 (noncommutative_operator): Check the availability of barrel
3860 2018-04-09 Richard Biener <rguenther@suse.de>
3862 PR tree-optimization/85284
3863 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3864 Only use the niter constraining form of simple_iv when the exit
3867 2018-04-09 Tom de Vries <tom@codesourcery.com>
3870 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3871 (define_expand "*memory_barrier"): New define_expand.
3872 (define_insn "memory_barrier"): New insn.
3874 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3876 PR rtl-optimization/80463
3877 PR rtl-optimization/83972
3878 PR rtl-optimization/83480
3880 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3881 correct producer for the insn.
3882 (tidy_control_flow): Fixup seqnos in case of debug insns.
3884 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3886 PR rtl-optimization/83913
3888 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3889 different sched-times when merging exprs.
3891 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3893 PR rtl-optimization/83962
3895 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3896 tidy_fallthru_edge and tidy_control_flow.
3898 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3900 PR rtl-optimization/83530
3902 * sel-sched.c (force_next_insn): New global variable.
3903 (remove_insn_for_debug): When force_next_insn is true, also leave only
3904 next insn in the ready list.
3905 (sel_sched_region): When the region wasn't scheduled, make another pass
3906 over it with force_next_insn set to 1.
3908 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3910 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3912 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3913 for interrupt control.
3914 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3915 functions for interrupt control.
3916 * config/nds32/nds32-intrinsic.md: Likewise.
3917 * config/nds32/nds32_intrinsic.h: Likewise.
3918 * config/nds32/nds32.h (nds32_builtins): Likewise.
3920 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3922 * config/nds32/nds32.c (nds32_init_machine_status,
3923 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3924 strict_aligned_p field.
3925 (nds32_expand_to_rtl_hook): New function.
3926 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3927 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3929 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3930 Chung-Ju Wu <jasonwucj@gmail.com>
3932 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3933 * config/nds32/nds32-n7.md: New file.
3934 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3935 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3937 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3938 * config/nds32/nds32.md (pipeline_model): Add n7.
3939 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3940 * config/nds32/pipelines.md: Include n7 settings.
3942 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3943 Chung-Ju Wu <jasonwucj@gmail.com>
3945 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3946 * config/nds32/nds32-e8.md: New file.
3947 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3948 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3950 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3951 * config/nds32/nds32.md (pipeline_model): Add e8.
3952 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3953 * config/nds32/pipelines.md: Include e8 settings.
3955 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3956 Chung-Ju Wu <jasonwucj@gmail.com>
3958 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3959 * config/nds32/nds32-n8.md: New file.
3960 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3961 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3963 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3964 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3965 * config/nds32/nds32.md (pipeline_model): Add n8.
3966 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3967 * config/nds32/pipelines.md: Include n8 settings.
3969 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3970 Chung-Ju Wu <jasonwucj@gmail.com>
3972 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3973 * config/nds32/nds32-n9-2r1w.md: New file.
3974 * config/nds32/nds32-n9-3r2w.md: New file.
3975 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3976 nds32_register_ports): New or modify for cpu n9.
3977 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3979 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3980 * config/nds32/nds32-utils.c: New file.
3981 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3982 TARGET_MUL_SLOW): Define.
3983 * config/nds32/nds32.md (pipeline_model): New attribute.
3984 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3985 New options that support cpu n9.
3986 * config/nds32/pipelines.md: Include n9 settings.
3987 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3989 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3991 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3992 information if necessary.
3993 (output_cond_branch_compare_zero): Likewise.
3994 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3995 (nds32_target_alignment): Refine for alignment.
3996 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3997 (FUNCTION_BOUNDARY): Modify.
3998 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4000 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4002 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4004 * config/nds32/constants.md (unspec_volatile_element): Add values for
4005 TLB operation and data prefetch.
4006 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4007 functions for TLB operation and data prefetch.
4008 * config/nds32/nds32-intrinsic.md: Likewise.
4009 * config/nds32/nds32_intrinsic.h: Likewise.
4010 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4011 (nds32_print_operand): Likewise.
4012 * config/nds32/nds32.h (nds32_builtins): Likewise.
4014 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4015 Andrew Pinski <pinsika@gcc.gnu.org>
4018 * match.pd: Use constant_boolean_node of correct type instead of
4019 boolean_true_node or boolean_false_node for simplifying
4020 pointer comparisons to zero.
4022 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4024 PR tree-optimization/80021
4025 * tree.c (verify_type_variant): Make error call in verify_variant_match
4026 translatable and remove final full stop.
4028 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4030 * config/nds32/constants.md (unspec_volatile_element): Add
4031 UNSPEC_VOLATILE_EH_RETURN.
4032 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4033 nds32_output_stack_pop): Support dwarf exception handling process.
4034 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4035 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4036 exception handling process.
4037 (nds32_compute_stack_frame): Likewise.
4038 (nds32_return_addr_rtx): Likewise.
4039 (nds32_initial_elimination_offset): Likewise.
4040 (nds32_expand_prologue): Likewise.
4041 (nds32_expand_epilogue): Likewise.
4042 (nds32_dynamic_chain_address): New function.
4043 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4045 (DYNAMIC_CHAIN_ADDRESS): Define.
4046 (EH_RETURN_DATA_REGNO): Define.
4047 (EH_RETURN_STACKADJ_RTX): Define.
4048 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4049 patterns for dwarf exception handling.
4051 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4053 * config/nds32/nds32.h: Clean up obsolete macros.
4055 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4057 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4058 Add enum values for particular instructions.
4059 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4060 particular intrinsic functions.
4061 * config/nds32/nds32-intrinsic.md: Likewise.
4062 * config/nds32/nds32_intrinsic.h: Likewise.
4063 * config/nds32/nds32.h (nds32_builtins): Likewise.
4064 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4065 (btst, ave): New patterns for particular instructions.
4067 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4069 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4070 Add enum values for atomic load/store and memory sync.
4071 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4073 * config/nds32/nds32-intrinsic.md: Likewise.
4074 * config/nds32/nds32_intrinsic.h: Likewise.
4075 * config/nds32/nds32.h (nds32_builtins): Likewise.
4077 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4079 PR tree-optimization/85257
4080 * fold-const.c (native_encode_vector): If not all elts could fit
4081 and off is -1, return 0 rather than offset.
4082 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4083 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4084 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4085 adjust buffer in native_interpret_expr call.
4087 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4089 * config/nds32/constants.md (unspec_volatile_element): Add cache
4090 control enum values.
4091 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4092 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4093 * config/nds32/nds32.c (nds32_cctl_names): New.
4094 (nds32_print_operand): Handle cache control register names.
4095 * config/nds32/nds32.h (nds32_builtins): New enum values.
4096 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4098 * config/nds32/nds32.md (type): Add mmu.
4099 * config/nds32/pipelines.md (simple_insn): Add mmu.
4101 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4103 * config/nds32/nds32.md (type): Remove call.
4104 * config/nds32/pipelines.md (simple_insn): Likewise.
4106 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4108 * config/nds32/constants.md (unspec_volatile_element): Add
4109 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4110 UNSPEC_VOLATILE_FMFCFG.
4111 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4112 description for fmfcfg and fmfcsr.
4113 (bdesc_1arg): Add fmtcsr.
4114 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4115 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4116 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4117 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4118 unspec_fmfcfg): New patterns.
4119 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4120 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4121 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4122 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4123 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4124 __nds32__fmfcfg): Define.
4126 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4128 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4129 intrinsic register names.
4130 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4131 intrinsic register enum values and macros.
4133 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4135 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4136 for load/store addressing form.
4137 (nds32_print_operand_address): Likewise.
4139 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4142 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4143 based on LABEL_REF. Remove useless assertion.
4144 (pic_address_needs_scratch): Fix formatting.
4145 (sparc_legitimize_pic_address): Minor tweaks.
4146 (sparc_delegitimize_address): Adjust assertion accordingly.
4147 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4148 into symbolic_operand.
4149 (movsi_high_pic_label_ref): Likewise.
4150 (movsi_lo_sum_pic_label_ref): Likewise.
4151 (movdi_pic_label_ref): Likewise.
4152 (movdi_high_pic_label_ref): Likewise.
4153 (movdi_lo_sum_pic_label_ref): Likewise.
4155 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4157 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4158 custom LIB_SPEC setup.
4160 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4161 Kito Cheng <kito.cheng@gmail.com>
4163 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4164 * config/riscv/freebsd.h: New.
4166 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4168 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4169 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4172 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4173 Kito Cheng <kito.cheng@gmail.com>
4175 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4176 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4177 * config/nds32/nds32-protos.h (nds32_output_call,
4178 nds32_output_return): Declare.
4179 * config/nds32/nds32.md: Refine all the call and return patterns.
4181 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4184 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4185 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4187 PR rtl-optimization/84872
4188 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4189 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4192 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4194 * expr.c (copy_blkmode_to_reg): Revert 254862.
4195 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4197 2018-04-06 Richard Biener <rguenther@suse.de>
4200 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4201 after seeing a component reference with an adjacent field. Treat
4202 refs to arrays at struct end of external decls similar to
4203 refs to unconstrained commons.
4205 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4208 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4209 look through SAVE_EXPRs with non-side-effects argument. Adjust
4211 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4214 2018-04-06 Richard Biener <rguenther@suse.de>
4217 * alias.c (find_base_term): New wrapper around find_base_term
4218 unwinding CSELIB_VAL_PTR changes.
4219 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4222 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4224 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4226 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4227 constant definitions.
4228 ("nop"): lr 0,0 -> nopr r0
4229 ("nop_lr0", "nop_lr1"): New insn definitions.
4231 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4233 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4234 NDS32_V3PUSH_AVAILABLE_P macro.
4236 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4237 Chung-Ju Wu <jasonwucj@gmail.com>
4239 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4240 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4241 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4242 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4243 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4244 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4245 * config/nds32/constraints.md: New constraints and checking for hard
4246 float configuration.
4247 * config/nds32/iterators.md: New mode iterator and attribute for hard
4248 float configuration.
4249 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4251 * config/nds32/nds32-fpu.md: New file.
4252 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4253 deal with hard float code generation.
4254 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4256 (abi_type, float_reg_number): New enum type.
4257 * config/nds32/nds32-predicates.c: New predicates for hard float.
4258 * config/nds32/nds32-protos.h: Declare functions for hard float.
4259 * config/nds32/nds32.c: Implementation for hard float configuration.
4260 * config/nds32/nds32.h: Definitions for hard float configuration.
4261 * config/nds32/nds32.md: Include hard float machine description and
4262 modify patterns for hard float configuration.
4263 * config/nds32/nds32.opt: New options for hard float configuration.
4264 * config/nds32/predicates.md: New predicates for hard float
4267 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4269 * common/config/nds32/nds32-common.c
4270 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4272 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4275 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4276 CONSTRUCTOR_ELT (ctor, ...)->value.
4278 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4281 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4283 2018-04-05 Tom de Vries <tom@codesourcery.com>
4286 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4289 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4290 Kito Cheng <kito.cheng@gmail.com>
4292 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4293 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4294 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4295 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4297 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4298 Kito Cheng <kito.cheng@gmail.com>
4300 * config/nds32/constraints.md (Ufe): New memory constraint.
4301 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4302 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4303 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4305 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4306 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4308 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4310 * config/nds32/nds32.md: Use optimize_size in the condition for
4311 alu-shift instructions.
4313 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4315 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4317 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4319 * config/nds32/nds32.md (negsi2): Refine pattern.
4321 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4322 Chung-Ju Wu <jasonwucj@gmail.com>
4324 * config/nds32/iterators.md (shift_rotate): New code iterator.
4325 (shift): New code attribute.
4326 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4327 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4328 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4329 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4330 bit-wise operations.
4331 (andsi3, *andsi3): Ditto.
4332 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4333 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4334 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4335 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4336 nds32_ior_operand, nds32_xor_operand): New predicates.
4338 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4340 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4341 (addsi3, subsi3): ... this.
4343 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4345 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4347 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4349 * config/nds32/nds32.md: Adjust indention.
4351 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4353 * config/nds32/nds32.md (feature): New attribute.
4355 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4357 * config/nds32/nds32.md (subtype): New attribute.
4359 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4362 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4363 expansion to perform a bitwise AND of the argument followed by a
4364 boolean negation of the result.
4366 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4368 PR rtl-optimization/84878
4369 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4370 the basic block. Assert the use reference is not artificial and that
4371 it has an associated insn.
4373 2018-04-04 Michael Matz <matz@suse.de>
4375 * builtins.c (compute_objsize): Pass correct operand
4376 to array_at_struct_end_p.
4378 2018-04-04 Richard Biener <rguenther@suse.de>
4381 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4382 from contexts for DINFO_LEVEL_TERSE and below.
4384 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4386 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4387 resiter_operand condition.
4388 * config/nds32/nds32.md (*move<mode>): Ditto.
4390 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4391 Monk Chiang <sh.chiang04@gmail.com>
4393 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4395 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4397 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4399 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4400 Kito Cheng <kito.cheng@gmail.com>
4402 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4403 nds32_cond_code_str, output_cond_branch,
4404 output_cond_branch_compare_zero, nds32_expand_cbranch,
4405 nds32_expand_cstore, nds32_expand_movcc,
4406 nds32_output_cbranchsi4_equality_zero,
4407 nds32_output_cbranchsi4_equality_reg,
4408 nds32_output_cbranchsi4_equality_reg_or_const_int,
4409 nds32_output_cbranchsi4_greater_less_zero: New functions.
4410 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4411 nds32_expand_cstore, nds32_expand_movcc,
4412 nds32_output_cbranchsi4_equality_zero,
4413 nds32_output_cbranchsi4_equality_reg,
4414 nds32_output_cbranchsi4_equality_reg_or_const_int,
4415 nds32_output_cbranchsi4_greater_less_zero): Declare.
4416 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4417 nds32_rimm11s_operand): New predicates.
4418 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4419 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4422 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4424 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4425 * config/nds32/nds32.md: Ditto.
4426 * config/nds32/pipelines.md: Ditto.
4428 2018-04-04 Richard Biener <rguenther@suse.de>
4430 PR tree-optimization/85168
4431 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4432 propagating abnormals.
4434 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4436 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4438 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4439 Kito Cheng <kito.cheng@gmail.com>
4441 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4442 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4443 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4444 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4445 * config/nds32/nds32.md (sibcall_internal): New.
4446 (sibcall_register): Remove.
4447 (sibcall_immediate): Remove.
4448 (sibcall_value_internal): New.
4449 (sibcall_value_register): Remove.
4450 (sibcall_value_immediate): Remove.
4451 * config/nds32/predicates.md (nds32_general_register_operand): New.
4452 (nds32_call_address_operand): New.
4454 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4456 PR rtl-optimization/85167
4457 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4458 bb_defs if *split_p, instead preinitialize it to NULL.
4460 PR tree-optimization/85156
4461 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4462 evaluating the argument multiple times.
4464 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4466 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4468 (_mm_cvtpd_ps): Likewise.
4469 (_mm_cvttpd_epi32): Likewise.
4470 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4471 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4472 vector, pixel, and bool following altivec.h include.
4474 2018-04-03 Martin Sebor <msebor@redhat.com>
4476 * doc/extend.texi (Common Function Attributes): Clarify.
4477 (const attribute): Likewise.
4478 (pure attribute): Likewise.
4480 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4483 * config/i386/i386.c (ix86_expand_vector_set): Use
4484 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4486 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4488 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4489 instructions when changing rounding bits to preserve precision bits
4490 in the x87 control word.
4492 2018-04-03 Martin Liska <mliska@suse.cz>
4494 PR tree-optimization/82491
4495 * rtl.h (strip_offset_and_add): Replace += suboffset with
4496 poly_uint64 () + suboffset.
4498 2018-03-29 Martin Liska <mliska@suse.cz>
4499 Martin Jambor <mjambor@suse.cz>
4502 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4503 param_type is not an integral or pointer type.
4505 2018-04-03 Richard Biener <rguenther@suse.de>
4507 * sese.h (recompute_all_dominators): Remove.
4509 2018-04-02 Martin Sebor <msebor@redhat.com>
4511 * doc/invoke.texi (-Wrestrict): Fix typos.
4513 2018-04-02 Jim Wilson <jimw@sifive.com>
4515 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4516 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4517 (<optab>di3, <optab>si3_extend): Likewise.
4518 (<optab>si3_mask, <optab>si3_mask_1): New.
4519 (<optab>di3_mask, <optab>di3_mask_1): New.
4520 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4521 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4522 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4524 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4526 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4529 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4531 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4532 (nds32_canonicalize_comparison): New function.
4534 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4535 Kito Cheng <kito.cheng@gmail.com>
4536 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4538 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4539 * config/nds32/constants.md (unspec_volatile_element): Add
4540 UNSPEC_VOLATILE_RELAX_GROUP.
4541 * config/nds32/nds32-relax-opt.c: New file.
4542 * config/nds32/nds32-predicates.c
4543 (nds32_symbol_load_store_p): New function.
4544 * config/nds32/nds32-protos.h
4545 (nds32_symbol_load_store_p): Declare function.
4546 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4547 * config/nds32/nds32.c
4548 (nds32_register_pass): New function to register pass.
4549 (nds32_register_passes): New function to register passes.
4550 * config/nds32/nds32.md (relax_group): New pattern.
4551 * config/nds32/nds32.opt (mrelax-hint): New option.
4552 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4554 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4556 * config/nds32/t-nds32: Modify files dependency.
4558 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4560 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4561 (PROFILE_HOOK): Define its implementation.
4563 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4565 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4566 type and 32-bit size.
4568 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4571 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4572 (V_128_256): New mode iterator.
4573 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4574 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4575 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4577 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4578 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4580 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4583 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4584 NaN inputs correctly.
4586 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4589 * config/rs6000/vsx.md (??r): New mode attribute.
4590 (*vsx_mov<mode>_64bit): Use it.
4591 (*vsx_mov<mode>_32bit): Likewise.
4593 2018-03-30 Martin Sebor <msebor@redhat.com>
4595 PR tree-optimization/84818
4596 * builtins.c (check_access): Use warning_n.
4598 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4601 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4603 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4606 2018-03-30 Julia Koval <julia.koval@intel.com>
4609 * x86-tune.def (movx, partial_reg_dependency): Enable for
4612 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4615 * lra-constraints.c (process_alt_operands): Move setting
4616 this_alternative_matches below.
4618 2018-03-29 Martin Liska <mliska@suse.cz>
4621 * doc/invoke.texi: Document how LTO works with debug info.
4622 Describe auto-load support of binutils. Mention 'x86-64'
4623 as valid option value of -march option.
4625 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4627 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4630 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4631 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4632 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4635 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4638 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4639 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4640 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4641 for RS6000_BTM_POWERPC64.
4642 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4643 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4644 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4649 2018-03-28 Carl Love <cel@us.ibm.com>
4652 2017-09-27 Carl Love <cel@us.ibm.com>
4654 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4655 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4656 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4659 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4661 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4662 instead of __vector bool.
4663 (_mm_max_pu8): Likewise.
4664 (_mm_min_pi16): Likewise.
4666 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4669 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4670 (DIVWEUO): Likewise.
4672 (DIVDEUO): Likewise.
4673 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4674 DIVWEUO and DIVDEUO.
4675 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4676 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4677 (div_extend): Likewise.
4678 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4680 (__builtin_divweuo): Likewise.
4681 (__builtin_divdeo): Likewise.
4682 (__builtin_divdeuo): Likewise.
4684 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4687 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4688 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4690 PR tree-optimization/82004
4691 * gimple-match-head.c (optimize_pow_to_exp): New function.
4692 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4693 Don't fold to exp if optimize_pow_to_exp is false.
4695 2018-03-28 Martin Liska <mliska@suse.cz>
4698 * calls.c (initialize_argument_information): Fix trailing space.
4699 * common.opt: Fix typo and provide better explanation for
4700 -fsanitize-coverage option.
4701 * config/i386/i386.opt: Fix typo.
4703 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4704 Martin Liska <mliska@suse.cz>
4707 * gimplify.c (asan_poison_variable): Don't do the check for
4708 gimplify_omp_ctxp here.
4709 (gimplify_decl_expr): Do it here.
4710 (gimplify_target_expr): Likewise.
4712 2018-03-28 Martin Liska <mliska@suse.cz>
4715 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4716 chkp_type_bounds_count if MPX is not enabled.
4718 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4720 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4722 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4725 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4726 function to create the function decl for complex long double
4727 multiply and divide for -mabi=ieeelongdouble.
4728 (init_float128_ieee): Call it.
4730 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4733 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4734 -fcf-protection=branch -mibt.
4735 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4737 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4740 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4742 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4745 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4746 extern array declarations.
4748 2018-03-27 Richard Biener <rguenther@suse.de>
4751 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4752 explicit single_use checks.
4754 2018-03-27 Richard Biener <rguenther@suse.de>
4756 PR tree-optimization/85082
4757 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4760 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4762 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4763 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4764 Turn on fasynchronous-unwind-tables and funwind-tables.
4766 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4769 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4770 (*bmi_blsr_<mode>_ccz): Ditto.
4772 2018-03-26 Tom de Vries <tom@codesourcery.com>
4774 PR tree-optimization/85063
4775 * omp-general.c (offloading_function_p): New function. Factor out
4777 * omp-offload.c (pass_omp_target_link::gate): ... here.
4778 * omp-general.h (offloading_function_p): Declare.
4779 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4780 with attribute omp declare target for offloading functions.
4782 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4784 PR tree-optimization/84005
4785 * tree-data-ref.h (get_base_for_alignment): Declare.
4786 * tree-data-ref.c (get_base_for_alignment_1): New function.
4787 (get_base_for_alignment): Likewise.
4788 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4789 get_base_for_alignment to find a suitable base object, instead
4790 of always using drb->base_address.
4792 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4795 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4796 known size by default.
4798 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4801 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4802 and non BLKmode operands.
4804 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4807 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4808 Clean up attributes.
4810 2018-03-23 Richard Biener <rguenther@suse.de>
4813 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4814 we are going to emit early debug for LTO.
4816 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4819 * function.c (match_asm_constraints_1): Don't optimize if input
4820 doesn't satisfy general_operand predicate for output's mode.
4823 * alias.c (write_dependence_p): Don't require for x_canonicalized
4824 non-VOIDmode if x has VOIDmode.
4827 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4828 just don't try to optimize it rather than assert it never happens.
4830 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4832 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4833 macro expansions for definition of ST_INTERNAL_<mode> and
4834 LD_INTERNAL_<mode> builtins.
4835 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4837 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4839 (altivec_expand_st_builtin): Likewise.
4840 (altivec_expand_builtin): Remove calls to deleted functions.
4841 (rs6000_address_for_altivec): Delete this function.
4842 * config/rs6000/vector.md: Remove expands for
4843 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4845 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4848 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4849 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4850 re-computing once computed.
4851 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4852 (arm_init_machine_status): Initialize
4853 machine->static_chain_stack_bytes.
4855 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4858 * doc/extend.texi: Add four new prototypes for vec_ld.
4859 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4860 definitions for more logical presentation.
4861 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4862 entries for V1TI variants of __builtin_altivec_ld builtin.
4863 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4864 handling of V1TI variant of LVX icode pattern.
4865 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4866 (rs6000_gimple_fold_builtin): Likewise.
4867 (altivec_init_builtins): Add code to define
4868 __builtin_altivec_lvx_v1ti function.
4870 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4873 * function.c (match_asm_constraints_1): Don't do the optimization
4874 if input isn't a REG, SUBREG, MEM or constant.
4876 2018-03-22 Tom de Vries <tom@codesourcery.com>
4878 PR tree-optimization/84956
4879 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4880 bb_has_abnormal_pred.
4882 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4885 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4886 DECL_INITIAL (decl) to decl at the end.
4887 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4890 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4892 * doc/extend.texi (__builtin_tgmath): Document when complex
4893 integer types are treated as _Complex _Float64.
4895 2018-03-21 Tom de Vries <tom@codesourcery.com>
4897 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4899 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4901 PR tree-optimization/84960
4902 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4903 if it is ENTRY block, move them into single succ of ENTRY in that case.
4905 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4907 PR tree-optimization/84811
4908 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4909 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4910 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4911 (known_subrange_p): Update accordingly. Cast each value involved
4912 in the size comparison, rather than casting the result of the
4915 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4917 PR tree-optimization/84982
4918 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4919 by flipping the least significant bit rather than all bits from
4920 bitpos to bitpos + bitsize - 1.
4922 2018-03-21 Nathan Sidwell <nathan@acm.org>
4924 * doc/extend.texi (Deprecated Features): Remove mention of
4925 long-deleted deprecations.
4927 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4930 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4931 * configure: Regenerate.
4933 2018-03-21 Tom de Vries <tom@codesourcery.com>
4935 PR tree-optimization/83126
4936 * tree-parloops.c (num_phis): New function.
4937 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4939 2018-03-21 Nathan Sidwell <nathan@acm.org>
4941 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4942 mention anon-struct/union members and trailing attributes.
4944 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4946 PR tree-optimization/84969
4947 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4948 builtin memset partitions if they set different rhs values.
4950 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4952 PR rtl-optimization/84989
4953 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4954 VEC_DUPLICATE with scalar result mode.
4956 2018-03-21 Martin Liska <mliska@suse.cz>
4959 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4960 not intended return statement.
4962 2018-03-21 Martin Liska <mliska@suse.cz>
4965 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4966 (chkp_find_bound_slots_1): Limit number of iterations.
4968 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4971 * Minor grammar fixes for x86 options.
4973 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4976 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4977 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4979 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4982 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4983 (altivec_lvx_<mode>_1op): Likewise.
4984 (altivec_stvx_<mode>_2op): Likewise.
4985 (altivec_stvx_<mode>_1op): Likewise.
4986 (altivec_lvx_<VM2:mode>): New define_expand.
4987 (altivec_stvx_<VM2:mode>): Likewise.
4988 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4989 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4990 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4991 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4992 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4993 (rs6000_gen_lvx): Likewise.
4994 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4995 (altivec_expand_stv_builtin): Likewise.
4996 (altivec_expand_builtin): Likewise.
4997 * config/rs6000/vector.md: Likewise.
4999 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5002 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5005 2018-03-20 Richard Biener <rguenther@suse.de>
5008 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5009 sign-conversions as zero, fall back to standard scalar_stmt
5012 2018-03-20 Martin Liska <mliska@suse.cz>
5015 * predict.c (rebuild_frequencies): Handle case when we have
5016 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5018 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5021 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5022 flag_section_anchors.
5023 * varasm.c (use_blocks_for_decl_p): Remove hack for
5024 dw2_force_const_mem.
5027 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5029 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5030 be created, use lowpart_subreg of operands[0] rather than operands[0]
5032 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5033 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5034 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5035 and n constraint instead of aarch64_shift_imm_di and Usd.
5036 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5037 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5039 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5042 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5043 to favor GPR over NEON registers.
5044 (<shift>di3_neon): Likewise.
5046 2018-03-20 Tom de Vries <tom@codesourcery.com>
5049 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5050 (nvptx_process_pars): Emit bar.sync asap and alap.
5052 2018-03-20 Tom de Vries <tom@codesourcery.com>
5055 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5056 seen_label if seen_label is already set.
5058 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5061 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5062 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5063 Use 1U instead of 1. Formatting fixes.
5066 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5067 instead of TREE_TYPE (s1) for the return value.
5069 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5071 PR tree-optimization/84946
5072 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5073 bitsize + bitsize in poly_uint64 rather than poly_int64.
5076 * dwarf2asm.c: Include fold-const.c.
5077 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5078 of decl rather than decl itself.
5080 PR rtl-optimization/84643
5081 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5083 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5086 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5087 calling assemble_variable.
5089 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5092 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5093 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5095 2018-03-19 Jim Wilson <jimw@sifive.com>
5098 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5099 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5100 (riscv_first_stack_step): Likewise.
5101 (riscv_option_override): Use STACK_BOUNDARY instead of
5103 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5105 (BIGGEST_ALIGNMENT): Set to 128.
5106 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5107 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5110 2018-03-19 Richard Biener <rguenther@suse.de>
5112 PR tree-optimization/84933
5113 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5114 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5116 2018-03-19 Richard Biener <rguenther@suse.de>
5118 PR tree-optimization/84859
5119 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5120 (cond_if_else_store_replacement): Perform sinking operation on
5121 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5122 Generalize what a BB with a single eligible store is.
5124 2018-03-19 Richard Biener <rguenther@suse.de>
5126 PR tree-optimization/84929
5127 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5128 chrec_is_positive against non-chrec arg.
5130 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5133 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5135 2018-03-18 Martin Liska <mliska@suse.cz>
5137 PR rtl-optimization/84635
5138 * regrename.c (build_def_use): Use matches_mode only when
5141 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5143 PR tree-optimization/84913
5144 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5145 vectorize chains of COND_EXPRs.
5147 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5149 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5151 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5153 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5155 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5157 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5159 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5160 Kito Cheng <kito.cheng@gmail.com>
5162 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5163 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5164 (nds32_adjust_reg_alloc_order): New function.
5165 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5167 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5169 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5170 nds32_print_operand, nds32_print_operand_address): Use
5171 HOST_WIDE_INT_PRINT_DEC instead.
5173 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5175 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5177 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5180 * config/i386/i386.c (initial_ix86_tune_features,
5181 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5183 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5184 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5185 rather than 1u << ix86_tune. Formatting fix.
5186 (ix86_option_override_internal): Change ix86_arch_mask from
5187 unsigned int to unsigned HOST_WIDE_INT, initialize to
5188 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5189 (ix86_function_specific_restore): Likewise.
5191 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5194 * postreload.c (reload_combine_recognize_pattern): Perform
5195 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5196 truncate_int_for_mode the result for the destination's mode.
5199 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5200 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5202 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5205 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5208 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5209 regno_allocno_class_array and sorted_pseudos.
5210 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5211 insns where regno is used.
5213 2018-03-16 Martin Liska <mliska@suse.cz>
5216 * multiple_target.c (create_dispatcher_calls): Redirect
5217 reference in the symbol table.
5219 2018-03-16 Martin Liska <mliska@suse.cz>
5222 * multiple_target.c (create_dispatcher_calls): Redirect also
5225 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5229 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5230 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5231 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5233 2018-03-16 Julia Koval <julia.koval@intel.com>
5235 * doc/invoke.texi (Skylake Server): Add CLWB.
5236 Cannonlake): Remove CLWB.
5238 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5240 PR tree-optimization/84841
5241 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5243 (FLOAT_ONE_CONST_TYPE): Define.
5244 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5245 (sort_by_operand_rank): Put entries with higher constant_type last
5246 rather than first to match comments.
5248 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5250 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5253 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5256 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5257 check and use address of target always.
5259 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5262 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5263 (indirect_thunk_bnd_needed): Likewise.
5264 (indirect_thunks_used): Likewise.
5265 (indirect_thunks_bnd_used): Likewise.
5266 (indirect_return_needed): New.
5267 (indirect_return_bnd_needed): Likewise.
5268 (output_indirect_thunk_function): Add a bool argument for
5270 (output_indirect_thunk_function): Don't generate alias for
5271 function return thunk.
5272 (ix86_code_end): Call output_indirect_thunk_function to generate
5273 function return thunks.
5274 (ix86_output_function_return): Set indirect_return_bnd_needed
5275 and indirect_return_needed instead of indirect_thunk_bnd_needed
5276 and indirect_thunk_needed.
5278 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5280 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5281 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5282 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5284 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5285 Paul Hua <paul.hua.gm@gmail.com>
5288 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5290 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5292 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5293 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5296 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5299 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5300 instead of GET_MODE_SIZE when comparing Units.
5302 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5305 * varasm.c (hash_section): Return an unchangeble hash value
5306 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5307 Return !aarch64_can_use_per_function_literal_pools_p ().
5309 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5312 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5313 rather than address of cmode as last argument to prepare_cmp_insn.
5315 2018-03-15 Julia Koval <julia.koval@intel.com>
5317 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5318 F_AVX512VNNI, F_AVX512BITALG): New.
5320 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5323 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5324 insn for floating-point loads and stores.
5326 2018-03-14 Carl Love <cel@us.ibm.com>
5328 * config/rs6000/rs6000-c.c: Add macro definitions for
5329 ALTIVEC_BUILTIN_VEC_PERMXOR.
5330 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5331 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5332 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5333 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5335 * config/doc/extend.texi: Add prototypes for vec_permxor.
5337 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5340 * diagnostic-show-locus.c (class layout_point): Convert m_line
5341 from int to linenum_type.
5342 (line_span::comparator): Use linenum "compare" function when
5343 comparing line numbers.
5344 (test_line_span): New function.
5345 (layout_range::contains_point): Convert param "row" from int to
5347 (layout_range::intersects_line_p): Likewise.
5348 (layout::will_show_line_p): Likewise.
5349 (layout::print_source_line): Likewise.
5350 (layout::should_print_annotation_line_p): Likewise.
5351 (layout::print_annotation_line): Likewise.
5352 (layout::print_leading_fixits): Likewise.
5353 (layout::annotation_line_showed_range_p): Likewise.
5354 (struct line_corrections): Likewise for field m_row.
5355 (line_corrections::line_corrections): Likewise for param "row".
5356 (layout::print_trailing_fixits): Likewise.
5357 (layout::get_state_at_point): Likewise.
5358 (layout::get_x_bound_for_row): Likewise.
5359 (layout::print_line): Likewise.
5360 (diagnostic_show_locus): Likewise for locals "last_line" and
5362 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5363 * input.c (selftest::test_linenum_comparisons): New function.
5364 (selftest::input_c_tests): Call it.
5365 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5366 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5367 * selftest.h (ASSERT_GT): New macro.
5368 (ASSERT_GT_AT): New macro.
5369 (ASSERT_LT): New macro.
5370 (ASSERT_LT_AT): New macro.
5372 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5374 PR rtl-optimization/84780
5375 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5377 2018-03-14 Martin Liska <mliska@suse.cz>
5379 * tree.c (record_node_allocation_statistics): Use
5380 get_stats_node_kind.
5381 (get_stats_node_kind): New function extracted from
5382 record_node_allocation_statistics.
5383 (free_node): Use get_stats_node_kind.
5385 2018-03-14 Richard Biener <rguenther@suse.de>
5387 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5388 that the value-set of ANTIC_IN doesn't grow.
5391 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5393 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5394 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5396 2018-03-14 Julia Koval <julia.koval@intel.com>
5398 * config.gcc (icelake-client, icelake-server): New.
5400 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5401 (initial_ix86_arch_features): Ditto.
5402 (PTA_SKYLAKE): Add SGX.
5403 (PTA_ICELAKE): Remove.
5404 (PTA_ICELAKE_CLIENT): New.
5405 (PTA_ICELAKE_SERVER): New.
5406 (ix86_option_override_internal): Split up icelake on icelake client and
5408 (get_builtin_code_for_version): Ditto.
5409 (fold_builtin_cpu): Ditto.
5410 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5411 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5412 * config/i386/i386.h (processor_type): Ditto.
5413 * doc/invoke.texi: Ditto.
5415 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5418 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5419 INTEGER_CST offset, add it together with bitpos / 8 and
5420 sign extend based on POINTER_SIZE.
5424 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5427 * config/i386/constraints.md (Yc): New register constraint.
5428 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5429 Use Yc constraint for alternative 2 of operand 0. Remove
5430 preferred_for_speed attribute.
5432 2018-03-14 Richard Biener <rguenther@suse.de>
5434 PR tree-optimization/84830
5435 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5436 with the old one to avoid oscillations.
5438 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5441 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5443 (assign_by_spills): Return a flag of reload assignment failure.
5444 Do not process the reload assignment failures. Do not spill other
5445 reload pseudos if they has the same reg class. Update n if
5447 (lra_assign): Add a return arg. Set up from the result of
5448 assign_by_spills call.
5449 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5450 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5451 usage_insns if it is not NULL.
5452 (spill_hard_reg_in_range): New function.
5453 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5454 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5455 function prototypes.
5456 (lra_assign): Change prototype.
5457 * lra.c (lra): Add code to deal with fails by splitting hard reg
5460 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5462 * config/riscv/riscv.opt (mrelax): New option.
5463 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5464 "norelax" when riscv_mrelax is disabled.
5465 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5467 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5470 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5471 reassociation for int modes.
5473 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5475 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5476 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5478 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5479 * config/aarch64/aarch64-sve.md
5480 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5481 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5482 (*extend<mode><Vwide>2): Rename to...
5483 (aarch64_sve_extend<mode><Vwide>2): ...this.
5484 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5485 renaming the old pattern to...
5486 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5488 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5489 define_expand, renaming the old pattern to...
5490 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5491 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5492 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5493 account when deciding which SVE instruction the optab should use.
5494 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5496 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5498 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5499 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5500 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5501 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5502 (tlsdesc_small_<mode>): Turn a define_expand and use
5503 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5504 (tlsdesc_small_advsimd_<mode>): ...this.
5505 (tlsdesc_small_sve_<mode>): New pattern.
5507 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5509 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5510 (UNSPEC_UMUL_HIGHPART): New constants.
5511 (MUL_HIGHPART): New int iteraor.
5512 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5513 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5515 (*<su>mul<mode>3_highpart): New define_insn.
5517 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5520 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5523 2018-03-13 Martin Liska <mliska@suse.cz>
5526 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5528 (sem_item_optimizer::~sem_item_optimizer): Release it.
5529 (sem_item_optimizer::merge_classes): Register variable aliases.
5530 (sem_item_optimizer::fixup_pt_set): New function.
5531 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5532 * ipa-icf.h: Declare new variables and functions.
5534 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5537 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5538 integer_pow2p@2 and test integer_pow2p in condition.
5539 (A < 0 ? C : 0): Similarly for @1.
5542 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5543 characters starting at p contain '\0' character, don't look beyond
5547 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5548 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5551 * reg-stack.c (change_stack): Change update_end var from int to
5552 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5553 also call set_block_for_insn on the newly added insns and rescan.
5556 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5557 on the last operand.
5560 * tree.c (stabilize_reference_1): Return save_expr (e) for
5561 STATEMENT_LIST even if it doesn't have side-effects.
5563 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5565 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5567 2018-03-12 Renlin Li <renlin.li@arm.com>
5569 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5570 aarch64_output_scalar_simd_mov_immediate.
5572 2018-03-12 Martin Sebor <msebor@redhat.com>
5574 PR tree-optimization/83456
5575 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5576 for perfectly overlapping calls to memcpy.
5577 (gimple_fold_builtin_memory_chk): Same.
5578 (gimple_fold_builtin_strcpy): Handle no-warning.
5579 (gimple_fold_builtin_stxcpy_chk): Same.
5580 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5582 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5584 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5585 parameter. Use it for SFmode.
5586 (rs6000_function_arg_advance_1): Adjust.
5587 (rs6000_function_arg): Adjust.
5588 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5590 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5592 PR rtl-optimization/84169
5593 PR rtl-optimization/84780
5594 * combine.c (can_combine_p): Check for a 2-insn combination whether
5595 the destination register is used between the two insns, too.
5597 2018-03-12 Richard Biener <rguenther@suse.de>
5599 PR tree-optimization/84803
5600 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5601 for refs DR analysis didn't process.
5603 2018-03-12 Richard Biener <rguenther@suse.de>
5605 PR tree-optimization/84777
5606 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5607 force-vectorize loops ignore whether we are optimizing for size.
5609 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5611 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5612 (TARGET_MD_ASM_ADJUST): Define.
5614 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5615 Kito Cheng <kito.cheng@gmail.com>
5616 Chung-Ju Wu <jasonwucj@gmail.com>
5618 * config/nds32/nds32.c (nds32_compute_stack_frame,
5619 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5620 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5621 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5622 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5623 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5624 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5625 * config/nds32/nds32.md (prologue, epilogue): Use macro
5626 NDS32_V3PUSH_AVAILABLE_P to do checking.
5628 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5631 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5632 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5633 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5634 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5635 addition of most attributes on !orig_type_die or the attribute not
5636 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5638 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5639 Chung-Ju Wu <jasonwucj@gmail.com>
5641 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5643 * config/nds32/nds32.opt (mvh): New option.
5645 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5646 Chung-Ju Wu <jasonwucj@gmail.com>
5648 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5650 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5651 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5654 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5655 Chung-Ju Wu <jasonwucj@gmail.com>
5657 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5659 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5660 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5662 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5663 Kito Cheng <kito.cheng@gmail.com>
5664 Chung-Ju Wu <jasonwucj@gmail.com>
5666 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5667 UNSPEC_FFMISM and UNSPEC_FLMISM.
5668 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5669 for ffb, ffmism and flmism.
5670 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5671 (unspec_ffmism): Ditto.
5672 (unspec_flmism): Ditto.
5673 (nds32_expand_builtin_impl): Check if string extension is available.
5674 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5675 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5677 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5680 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5683 * lra-assigns.c (assign_by_spills): Return a flag of reload
5684 assignment failure. Do not process the reload assignment
5685 failures. Do not spill other reload pseudos if they has the same
5687 (lra_assign): Add a return arg. Set up from the result of
5688 assign_by_spills call.
5689 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5690 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5691 usage_insns if it is not NULL.
5692 (spill_hard_reg_in_range): New function.
5693 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5694 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5695 function prototypes.
5696 (lra_assign): Change prototype.
5697 * lra.c (lra): Add code to deal with fails by splitting hard reg
5700 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5703 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5705 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5708 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5709 (dw_val_node): Add val_symbolic_view.
5710 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5711 (symview_upper_bound): New.
5712 (new_line_info_table): Initialize symviews_since_reset.
5713 (dwarf2out_source_line): Count symviews_since_reset and set
5714 symview_upper_bound.
5715 (dw_val_equal_p): Handle symview.
5716 (add_AT_symview): New.
5717 (print_dw_val): Handle symview.
5718 (attr_checksum, attr_checksum_ordered): Likewise.
5719 (same_dw_val_p, size_of_die): Likewise.
5720 (value_format, output_die): Likewise.
5721 (add_high_low_attributes): Use add_AT_symview for entry_view.
5722 (dwarf2out_finish): Reset symview_upper_bound, clear
5725 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5728 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5729 Add strict argument and use it.
5730 (rs6000_split_multireg_move): Update for new strict argument.
5731 (mem_operand_gpr): Disallow all non-offsettable addresses.
5732 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5734 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5737 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5738 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5739 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5742 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5743 decl, use remap_type if we want to use the type.
5745 2018-03-09 Martin Sebor <msebor@redhat.com>
5747 PR tree-optimization/84526
5748 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5750 (builtin_access::generic_overlap): Be prepared to handle non-array
5753 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5755 PR rtl-optimization/84682
5756 * lra-constraints.c (process_address_1): Check is_address flag
5757 for address constraints.
5758 (process_alt_operands): Likewise.
5759 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5760 preprocess_constraints.
5761 * recog.h (preprocess_constraints): Add oploc parameter.
5763 * recog.c (preprocess_constraints): Test address_operand for
5764 CT_ADDRESS constraints.
5766 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5769 * lra-assigns.c (assign_by_spills): Return a flag of reload
5770 assignment failure. Do not process the reload assignment
5771 failures. Do not spill other reload pseudos if they has the same
5773 (lra_assign): Add a return arg. Set up from the result of
5774 assign_by_spills call.
5775 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5776 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5777 usage_insns if it is not NULL.
5778 (spill_hard_reg_in_range): New function.
5779 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5780 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5781 function prototypes.
5782 (lra_assign): Change prototype.
5783 * lra.c (lra): Add code to deal with fails by splitting hard reg
5786 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5789 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5790 Accept complain bool parameter. Only emit errors if it is true.
5791 (arm_parse_cpu_option_name): Likewise.
5792 (arm_target_thumb_only): Adjust callers of the above.
5793 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5794 prototype to take a default true bool parameter.
5795 (arm_parse_arch_option_name): Likewise.
5797 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5798 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5802 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5803 * configure: Regenerate.
5804 * configure.ac ("linker --version-script option"): New.
5805 ("linker soname option"): New.
5807 2018-03-09 Richard Biener <rguenther@suse.de>
5809 PR tree-optimization/84775
5810 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5811 immediate uses of predicate stmts and mark them modified.
5814 PR tree-optimization/84178
5815 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5817 (version_loop_for_if_conversion): Delay update_ssa call.
5818 (tree_if_conversion): Delay update_ssa until after predicate
5821 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5824 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5825 when the function accesses prior frames.
5827 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5830 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5831 gen_llsym, otherwise call maybe_gen_llsym.
5834 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5835 has ',' character inside of it.
5837 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5840 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5841 as clobbering CC_REGNUM.
5843 2018-03-08 Richard Biener <rguenther@suse.de>
5846 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5847 (follow_copies_to_constant): Do not follow SSA names registered
5850 2018-03-08 Richard Biener <rguenther@suse.de>
5852 PR tree-optimization/84178
5853 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5855 (version_loop_for_if_conversion): Delay update_ssa call.
5856 (tree_if_conversion): Delay update_ssa until after predicate
5859 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5861 PR tree-optimization/84178
5862 * tree-if-conv.c (release_bb_predicate): Remove the
5863 the assertion that the stmts have NULL use_ops.
5864 Discard the statements, asserting that they haven't
5865 yet been added to a BB.
5867 2018-03-08 Richard Biener <rguenther@suse.de>
5869 PR tree-optimization/84746
5870 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5871 (phi_translate): Pass in destination ANTIC_OUT set.
5872 (phi_translate_1): Likewise. For a simplified result lookup
5873 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5874 (phi_translate_set): Adjust.
5875 (do_pre_regular_insertion): Likewise.
5876 (do_pre_partial_partial_insertion): Likewise.
5878 2018-03-08 Martin Liska <mliska@suse.cz>
5880 PR gcov-profile/84735
5881 * doc/gcov.texi: Document usage of profile files.
5882 * gcov-io.h: Document changes in the format.
5884 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5888 * dwarf2out.c (struct dw_line_info_table): Update comments for
5890 (FORCE_RESET_NEXT_VIEW): New.
5891 (FORCE_RESETTING_VIEW_P): New.
5892 (RESETTING_VIEW_P): Check for -1 too.
5893 (ZERO_VIEW_P): Likewise.
5894 (new_line_info_table): Force-reset next view.
5895 (dwarf2out_begin_function): Likewise.
5896 (dwarf2out_source_line): Simplify zero_view_p initialization.
5897 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5898 view directly. Omit view when omitting .loc at line 0.
5900 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5902 PR tree-optimization/84740
5903 * tree-switch-conversion.c (process_switch): Call build_constructors
5904 only if info.phi_count is non-zero.
5906 PR tree-optimization/84739
5907 * tree-tailcall.c (find_tail_calls): Check call arguments against
5908 DECL_ARGUMENTS (current_function_decl) rather than
5909 DECL_ARGUMENTS (func) when checking for tail recursion.
5911 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5913 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5914 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5915 Volker Reichelt's entry and add entries for people that perform
5916 GCC fuzzy testing and report numerous bugs.
5918 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5921 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5922 readonly data in sdata, if that is disabled.
5923 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5924 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5925 -mreadonly-in-sdata option.
5927 2018-03-07 Martin Sebor <msebor@redhat.com>
5929 PR tree-optimization/84468
5930 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5931 basic block when looking for nul assignment.
5933 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5936 * except.h (output_function_exception_table): Adjust prototype.
5937 * except.c (output_function_exception_table): Remove FNNAME parameter
5938 and add SECTION parameter. Ouput one part of the table at a time.
5939 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5940 the first part of the exception table and emit unwind directives.
5941 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5942 (i386_pe_seh_cold_init): Likewise.
5943 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5944 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5945 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5946 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5947 * config/i386/winnt.c: Include except.h.
5948 (struct seh_frame_state): Add reg_offset, after_prologue and
5949 in_cold_section fields.
5950 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5951 (i386_pe_seh_cold_init): New function.
5952 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5953 to seh->in_cold_section.
5954 (seh_emit_push): Record the offset of the push.
5955 (seh_emit_save): Record the offet of the save.
5956 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5957 Test seh->after_prologue to disregard the epilogue.
5958 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5959 (i386_pe_end_cold_function): New function.
5961 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5964 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5965 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5968 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5970 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5971 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5974 * multiple_target.c: Include tree-inline.h and intl.h.
5975 (expand_target_clones): Diagnose and fail if node->definition and
5976 !tree_versionable_function_p (node->decl).
5978 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5980 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5982 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5983 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5984 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5986 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5989 * combine.c (try_combine): Use reg_or_subregno instead of handling
5990 just paradoxical SUBREGs and REGs.
5992 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5994 * config/arc/arc.c (arc_finalize_pic): Remove function.
5995 (arc_must_save_register): We use single base PIC register, remove
5996 checks to save/restore the PIC register.
5997 (arc_expand_prologue): Likewise.
5998 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6000 (arc_verify_short): Likewise.
6001 (arc_attr_type): Likewise.
6002 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6003 (walk_stores): Likewise.
6004 (arc_address_cost): Make it static.
6005 (arc_verify_short): Likewise.
6006 (branch_dest): Likewise.
6007 (arc_attr_type): Likewise.
6008 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6009 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6010 (arc_final_prescan_insn): Remove inserting the nops due to
6011 hardware hazards. It is done in reorg step.
6012 (insn_length_variant_t): Remove.
6013 (insn_length_parameters_t): Likewise.
6014 (arc_insn_length_parameters): Likewise.
6015 (arc_get_insn_variants): Likewise.
6016 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6018 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6021 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6024 PR tree-optimization/84687
6025 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6027 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6029 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6031 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6032 Rename to ppc_speculation_barrier.
6033 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6034 __builtin_ppc_speculation_barrier.
6036 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6039 * combine.c (combine_simplify_rtx): Don't try to simplify if
6040 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6043 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6045 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6046 to 32 bytes when compiling for POWER9.
6048 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6051 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6052 regparm >= 3 with no arg reg available also for calls with
6053 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6056 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6058 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6060 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6063 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6065 2018-03-05 Richard Biener <rguenther@suse.de>
6067 PR tree-optimization/84486
6068 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6069 When inserting a __builtin_assume_aligned call set the LHS
6070 SSA name alignment info accordingly.
6072 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6074 PR tree-optimization/84114
6075 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6076 Avoid reassociation of FLOAT_MODE addition.
6078 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6080 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6081 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6082 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6083 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6084 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6085 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6086 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6088 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6089 __builtin_ia32_wbinvd): New builtins.
6090 (SPECIAL_ARGS2): New.
6091 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6092 (SPECIAL_ARGS2): New.
6093 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6094 (ix86_valid_target_attribute_inner_p): Ditto.
6095 (ix86_init_mmx_sse_builtins): Add special_args2.
6096 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6097 TARGET_WBNOINVD_P): New.
6098 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6099 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6100 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6101 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6102 * config/i386/pconfigintrin.h: New file.
6103 * config/i386/wbnoinvdintrin.h: Ditto.
6104 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6105 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6107 2018-03-05 Richard Biener <rguenther@suse.de>
6109 PR tree-optimization/84670
6110 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6112 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6113 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6114 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6115 doesn't grow if all successors (recursively) were visited at least
6118 2018-03-05 Richard Biener <rguenther@suse.de>
6120 PR tree-optimization/84650
6121 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6122 if executed in the loop pipeline.
6124 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6126 * doc/configfiles.texi (Configuration Files): Move info about
6127 conditionalizing $target-protos.h to...
6128 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6129 differs from $target-protos.h.
6131 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6132 Chung-Ju Wu <jasonwucj@gmail.com>
6134 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6135 * config/nds32/nds32-multiple.md (setmemsi): Define.
6136 * config/nds32/nds32-memory-manipulation.c
6137 (nds32_gen_dup_4_byte_to_word_value): New.
6138 (emit_setmem_word_loop): New.
6139 (emit_setmem_byte_loop): New.
6140 (nds32_expand_setmem_loop): New.
6141 (nds32_expand_setmem_loop_v3m): New.
6142 (nds32_expand_setmem_unroll): New.
6143 (nds32_expand_setmem): New.
6145 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6146 Chung-Ju Wu <jasonwucj@gmail.com>
6148 * config/nds32/nds32-memory-manipulation.c
6149 (nds32_emit_load_store): New.
6150 (nds32_emit_post_inc_load_store): New.
6151 (nds32_emit_mem_move): New.
6152 (nds32_emit_mem_move_block): New.
6153 (nds32_expand_movmemsi_loop_unknown_size): New.
6154 (nds32_expand_movmemsi_loop_known_size): New.
6155 (nds32_expand_movmemsi_loop): New.
6156 (nds32_expand_movmemsi_unroll): New.
6157 (nds32_expand_movmemqi): Rename ...
6158 (nds32_expand_movmemsi): ... to this.
6159 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6160 (movmemsi): ... to this.
6161 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6162 (nds32_expand_movmemsi): ... to this.
6164 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6165 Monk Chiang <sh.chiang04@gmail.com>
6166 Chung-Ju Wu <jasonwucj@gmail.com>
6168 * config/nds32/nds32-protos.h
6169 (nds32_expand_load_multiple): New arguments.
6170 (nds32_expand_store_multiple): Ditto.
6171 (nds32_valid_multiple_load_store): Rename ...
6172 (nds32_valid_multiple_load_store_p): ... to this.
6173 * config/nds32/nds32-memory-manipulation.c
6174 (nds32_expand_load_multiple): Refine implementation.
6175 (nds32_expand_store_multiple): Ditto.
6176 * config/nds32/nds32-multiple.md
6177 (load_multiple): Update nds32_expand_load_multiple interface.
6178 (store_multiple): Update nds32_expand_store_multiple interface.
6179 * config/nds32/nds32-predicates.c
6180 (nds32_valid_multiple_load_store): Rename ...
6181 (nds32_valid_multiple_load_store_p): ... to this and refine
6183 * config/nds32/predicates.md
6184 (nds32_load_multiple_and_update_address_operation): New predicate.
6185 (nds32_store_multiple_and_update_address_operation): New predicate.
6187 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6188 Chung-Ju Wu <jasonwucj@gmail.com>
6190 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6191 (combo): New attribute.
6192 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6194 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6196 * config/nds32/nds32.opt: Change -mcmodel= default value.
6198 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6199 Monk Chiang <sh.chiang04@gmail.com>
6200 Chung-Ju Wu <jasonwucj@gmail.com>
6202 * config/nds32/constants.md (unspec_element): New enum.
6203 * config/nds32/constraints.md (Umw): New constraint.
6204 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6205 * config/nds32/nds32-intrinsic.md: Likewise.
6206 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6207 (nds32_valid_smw_lwm_base_p): New.
6208 (nds32_output_smw_single_word): New.
6209 (nds32_output_lmw_single_word): New.
6210 (nds32_expand_unaligned_load): New.
6211 (nds32_expand_unaligned_store): New.
6212 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6213 (nds32_output_smw_single_word): Declare.
6214 (nds32_output_lmw_single_word): Declare.
6215 (nds32_expand_unaligned_load): Declare.
6216 (nds32_expand_unaligned_store): Declare.
6217 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6218 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6219 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6220 NDS32_BUILTIN_UASTORE_DW.
6221 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6224 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6225 Kito Cheng <kito.cheng@gmail.com>
6226 Chung-Ju Wu <jasonwucj@gmail.com>
6228 * config/nds32/nds32-intrinsic.c
6229 (nds32_expand_builtin_null_ftype_reg): Delete.
6230 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6231 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6232 (nds32_read_argument): New.
6233 (nds32_legitimize_target): Ditto.
6234 (nds32_legitimize_argument): Ditto.
6235 (nds32_check_constant_argument): Ditto.
6236 (nds32_expand_unop_builtin): Ditto.
6237 (nds32_expand_unopimm_builtin): Ditto.
6238 (nds32_expand_binop_builtin): Ditto.
6239 (nds32_builtin_decl_impl): Ditto.
6240 (builtin_description): Ditto.
6241 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6242 (nds32_init_builtins_impl): Ditto.
6243 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6244 (nds32_builtin_decl): New.
6245 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6246 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6248 2018-03-02 Jeff Law <law@redhat.com>
6250 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6251 (redundant_insn, fill_simple_delay_slots): Likewise.
6252 (fill_slots_from_thread): Likewise.
6253 * resource.c (mark_referenced_resources): Likewise.
6254 (mark_set_resources, find_dead_or_set_registers): Likewise.
6256 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6258 * substring-locations.h (format_warning_va): Formatting fix for
6260 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6262 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6263 * substring-locations.c: Include intl.h.
6264 (format_warning_va): Turned into small wrapper around
6265 format_warning_n_va, renamed to ...
6266 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6267 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6269 (format_warning_at_substring_n): New function.
6270 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6271 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6272 format_warning_at_substring with just a shorter name instead of
6273 const function pointer.
6274 (fmtwarn_n): New function.
6275 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6276 appropriate, get rid of all the fmtstr temporaries, move conditionals
6277 with G_() wrapped string literals directly into fmtwarn arguments,
6278 cast dir.len to (int), formatting fixes.
6280 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6282 * doc/invoke.texi: Remove "Cilk Plus" references.
6284 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6285 Richard Biener <rguenther@suse.de>
6288 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6289 for error or warning attributes if CALL_FROM_THUNK_P is set.
6292 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6295 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6296 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6299 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6300 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6302 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6303 instead of -1U in last predictors element's probability member.
6305 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6308 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6309 arguments if they are comparable.
6311 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6313 PR tree-optimization/84634
6314 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6315 masks and masked_loop_p with a single loop_masks, making sure it's
6316 null for bb vectorization.
6318 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6320 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6321 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6322 loop->force_vectorize to check whether there is no alias.
6324 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6327 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6329 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6331 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6332 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6333 instead of a loop around prev_real_insn.
6334 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6338 * config/i386/i386.c (ix86_print_operand): Use conditional
6339 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6342 2018-03-02 Richard Biener <rguenther@suse.de>
6344 PR tree-optimization/84427
6345 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6346 (bitmap_set_subtract_values): Rewrite to handle multiple
6349 (prune_clobbered_mems): Likewise.
6350 (phi_translate): Take edge instead of pred/phiblock.
6351 (phi_translate_1): Likewise.
6352 (phi_translate_set): Likewise. Insert all translated
6353 exprs for a value into the set, keeping possibly multiple
6354 expressions per value.
6355 (compute_antic_aux): Adjust for phi_translate changes.
6356 When intersecting union the expressions and prune those
6357 not in the final value set, keeping possibly multiple
6358 expressions per value. Do not use value-insertion
6359 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6360 all expressions. Add verification that the value-sets
6361 only shrink during iteration.
6362 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6363 (do_pre_regular_insertion): Likewise.
6364 (do_pre_partial_partial_insertion): Likewise.
6366 2018-03-02 Richard Biener <rguenther@suse.de>
6369 * config/darwin.c (saved_debug_info_level): New static global.
6370 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6371 (darwin_asm_lto_end): Restore debug info generation settings.
6373 2018-03-01 Martin Liska <mliska@suse.cz>
6376 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6379 2018-03-01 Richard Biener <rguenther@suse.de>
6382 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6383 type annotation with LTO when debug was disabled at compile-time.
6385 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6387 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6389 (mips_final_postscan_insn): Likewise.
6391 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6393 PR rtl-optimization/84528
6394 * alias.c (init_alias_target): Add commentary.
6395 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6396 a unique base value if the frame pointer is not eliminated
6397 to the stack pointer.
6399 2018-03-01 Tom de Vries <tom@codesourcery.com>
6401 PR rtl-optimization/83327
6402 * lra-int.h (hard_regs_spilled_into): Declare.
6403 * lra.c (hard_regs_spilled_into): Define.
6404 (init_reg_info): Init hard_regs_spilled_into.
6405 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6406 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6407 (process_bb_lives): Handle hard_regs_spilled_into.
6408 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6409 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6411 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6413 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6414 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6415 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6416 * config/rs6000/aix72.h: New file.
6418 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6420 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6421 instead of warning_at with conditional singular and plural messages
6425 * stor-layout.c (update_alignment_for_field): For
6426 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6427 && !DECL_PACKED (field), do the alignment update, just use
6428 only desired_align instead of MAX (type_align, desired_align)
6430 (place_field): Don't do known_align < desired_align handling
6431 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6432 is non-NULL, instead do it after rli->prev_field handling and
6433 only if not within a bitfield word. For DECL_PACKED (field)
6434 use type_align of BITS_PER_UNIT.
6436 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6438 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6439 superfluous parentheses and trailing spaces.
6441 2018-02-28 Richard Biener <rguenther@suse.de>
6443 PR tree-optimization/84584
6444 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6445 SCoPs with fake exit edge.
6447 2018-02-28 Martin Liska <mliska@suse.cz>
6450 * timevar.c (timer::print): Fix format to properly print 100%
6453 2018-02-28 Richard Biener <rguenther@suse.de>
6456 * genmatch.c (capture_info::walk_match): Do not mark
6457 captured expressions without operands as expr_p given
6458 they act more like predicates and should be subject to
6459 "lost tail" side-effect preserving.
6461 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6463 PR rtl-optimization/81611
6464 * auto-inc-dec.c (attempt_change): Move dead note from
6465 mem_insn if it's the next use of regno
6466 (find_address): Take address use of reg holding
6467 non-incremented value. Add parm to limit search to the named
6469 (merge_in_block): Attempt to use a mem insn that is the next
6470 use of the original regno.
6472 2018-02-27 Martin Sebor <msebor@redhat.com>
6475 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6476 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6478 2018-02-27 Martin Sebor <msebor@redhat.com>
6480 PR translation/84207
6481 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6482 n argument to unsigned HOST_WIDE_INT.
6483 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6484 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6485 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6486 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6488 2018-02-27 Richard Biener <rguenther@suse.de>
6490 PR tree-optimization/84512
6491 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6492 Do not use the estimate returned from record_stmt_cost for
6493 the scalar iteration cost but sum properly using add_stmt_cost.
6495 2018-02-27 Richard Biener <rguenther@suse.de>
6497 PR tree-optimization/84466
6498 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6499 Adjust last change to less strictly validate use operands.
6501 2018-02-27 Martin Liska <mliska@suse.cz>
6503 PR gcov-profile/84548
6504 * gcov.c (process_file): Allow partial overlap and consider it
6505 also as group functions.
6506 (output_lines): Properly calculate range of lines for a group.
6508 2018-02-27 Martin Liska <mliska@suse.cz>
6510 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6511 'ggc' suffixes. Change first column width.
6512 (timer::print): Fix formatting of the column.
6514 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6516 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6517 preserve inline entry blocks for the sake of debug inline
6518 entry point markers alone.
6519 (remove_unused_locals): Suggest in comments a better place to
6520 force the preservation of inline entry blocks that are
6521 otherwise unused, but do not preserve them.
6523 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6525 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6527 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6530 * config/i386/constraints.md (Bs): Replace
6531 ix86_indirect_branch_register with
6532 TARGET_INDIRECT_BRANCH_REGISTER.
6534 * config/i386/i386.md (indirect_jump): Likewise.
6535 (tablejump): Likewise.
6536 (*sibcall_memory): Likewise.
6537 (*sibcall_value_memory): Likewise.
6538 Peepholes of indirect call and jump via memory: Likewise.
6539 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6540 (*sibcall_value_GOT_32): Likewise.
6541 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6542 (GOT_memory_operand): Likewise.
6543 (call_insn_operand): Likewise.
6544 (sibcall_insn_operand): Likewise.
6545 (GOT32_symbol_operand): Likewise.
6546 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6548 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6550 PR rtl-optimization/83496
6551 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6552 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6553 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6554 redundant insn, if any.
6555 (relax_delay_slots): Likewise.
6556 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6558 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6560 PR tree-optimization/83965
6561 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6562 that grouped statements are part of a reduction chain. Return
6563 true if the statement is not marked as a reduction itself but
6565 (vect_recog_dot_prod_pattern): Don't check whether the statement
6566 is part of a group here.
6567 (vect_recog_sad_pattern): Likewise.
6568 (vect_recog_widen_sum_pattern): Likewise.
6570 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6573 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6575 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6578 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6580 (ix86_output_indirect_function_return): New prototype.
6581 (ix86_split_simple_return_pop_internal): Likewise.
6582 * config/i386/i386.c (indirect_return_via_cx): New.
6583 (indirect_return_via_cx_bnd): Likewise.
6584 (indirect_thunk_name): Handle return va CX_REG.
6585 (output_indirect_thunk_function): Create alias for
6586 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6587 (ix86_output_indirect_jmp): Remove the bool argument.
6588 (ix86_output_indirect_function_return): New function.
6589 (ix86_split_simple_return_pop_internal): Likewise.
6590 * config/i386/i386.md (*indirect_jump): Don't pass false
6591 to ix86_output_indirect_jmp.
6592 (*tablejump_1): Likewise.
6593 (simple_return_pop_internal): Change it to define_insn_and_split.
6594 Call ix86_split_simple_return_pop_internal to split it for
6596 (simple_return_indirect_internal): Call
6597 ix86_output_indirect_function_return instead of
6598 ix86_output_indirect_jmp.
6600 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6603 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6604 memset and value initialization afterwards.
6606 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6608 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6610 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6613 * common/config/aarch64/aarch64-common.c
6614 (aarch_option_optimization_table[]): Switch
6615 off fomit-frame-pointer
6617 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6618 Chung-Ju Wu <jasonwucj@gmail.com>
6620 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6622 (store_multiple): Ditto.
6624 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6626 * config.gcc: Add --with-cpu support for nds32 target.
6627 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6628 * config/nds32/nds32.opt: Add -mcpu= option.
6630 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6632 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6633 isel=yes): Warn for these deprecated options.
6635 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6637 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6638 ISA_2_5_MASKS_EMBEDDED.
6640 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6642 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6643 p->max as pointers rather than using iterative_hash_expr.
6645 2018-02-23 Carl Love <cel@us.ibm.com>
6647 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6648 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6650 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6651 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6652 P8V_BUILTIN_VEC_VUNSIGNED2.
6654 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6657 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6658 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6660 * lra-constraints.c (curr_insn_transform): Set up
6661 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6663 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6664 * lra-eliminations.c (spill_pseudos): Ditto.
6665 (process_insn_for_elimination): Ditto.
6666 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6667 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6668 LRA_NON_CLOBBERED_ALT.
6670 2018-02-22 Martin Sebor <msebor@redhat.com>
6672 PR tree-optimization/84480
6673 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6674 to maybe_diag_stxncpy_trunc. Call it.
6675 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6676 from gimple_fold_builtin_strcpy. Print inlining stack.
6677 (handle_builtin_stxncpy): Print inlining stack.
6678 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6680 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6683 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6684 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6685 and -fcheck-pointer-bounds are used together.
6686 (indirect_thunk_prefix): New enum.
6687 (indirect_thunk_need_prefix): New function.
6688 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6689 "_nt" instead of "_bnd" for NOTRACK prefix.
6690 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6691 (output_indirect_thunk_function): Likewise.
6693 (ix86_code_end): Update output_indirect_thunk_function calls.
6694 (ix86_output_indirect_branch_via_reg): Replace
6695 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6696 (ix86_output_indirect_branch_via_push): Likewise.
6697 (ix86_output_function_return): Likewise.
6698 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6699 incompatible with -fcf-protection=branch and
6700 -fcheck-pointer-bounds.
6702 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6705 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6706 Change gcc_assert call to output_operand_lossage.
6708 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6710 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6712 2018-02-22 DJ Delorie <dj@redhat.com>
6713 Sebastian Perta <sebastian.perta@renesas.com>
6714 Oleg Endo <olegendo@gcc.gnu.org>
6716 * config/rx/rx.c (rx_rtx_costs): New function.
6717 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6719 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6721 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6723 2018-02-22 Martin Liska <mliska@suse.cz>
6726 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6727 Add "native" as a possible value.
6729 2018-02-22 Martin Liska <mliska@suse.cz>
6732 * config/i386/i386.c (ix86_option_override_internal):
6733 Add "native" as a possible value for -march and -mtune.
6735 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6738 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6739 to all type variants.
6741 PR tree-optimization/84503
6742 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6743 width as info->bitpos + info->bitsize - start.
6744 (merged_store_group::merge_overlapping): Simplify width computation.
6745 (check_no_overlap): New function.
6746 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6747 start + width and last_order of the group, fail if check_no_overlap
6749 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6750 to group if check_no_overlap fails.
6752 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6754 * config/rs6000/altivec.md: Delete contraint arguments to
6755 define_expand, define_split, and define_peephole2, and in
6756 define_insn_and_split if always unused.
6757 * config/rs6000/darwin.md: Ditto.
6758 * config/rs6000/dfp.md: Ditto.
6759 * config/rs6000/rs6000.md: Ditto.
6760 * config/rs6000/sync.md: Ditto.
6761 * config/rs6000/vector.md: Ditto.
6762 * config/rs6000/vsx.md: Ditto.
6764 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6766 * config/rs6000/altivec.md: Write output control strings as braced
6767 blocks instead of double-quoted strings.
6768 * config/rs6000/darwin.md: Ditto.
6769 * config/rs6000/rs6000.md: Ditto.
6770 * config/rs6000/vector.md: Ditto.
6771 * config/rs6000/vsx.md: Ditto.
6773 2018-02-21 Jason Merrill <jason@redhat.com>
6775 PR c++/84314 - ICE with templates and fastcall attribute.
6776 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6778 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6780 * ipa-cp.c (determine_versionability): Fix comment typos.
6782 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6785 * ipa-cp.c (determine_versionability): Do not version functions caling
6788 2018-02-21 Martin Liska <mliska@suse.cz>
6791 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6792 Add "native" as a possible value.
6793 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6794 the macro when native cpu detection is available.
6796 2018-02-21 Martin Liska <mliska@suse.cz>
6799 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6800 Add "native" as a possible value.
6801 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6802 when native cpu detection is available.
6804 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6805 Martin Sebor <msebor@redhat.com>
6807 PR tree-optimization/84478
6808 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6810 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6811 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6812 support which is conservatively correct, for 2 only stay conservative
6813 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6814 argument to the 2 argument get_range_strlen, adjust 6 arg
6815 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6817 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6818 (gimple_fold_builtin_strlen): Pass true as last argument to
6821 2018-02-20 Martin Sebor <msebor@redhat.com>
6824 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6825 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6826 (builtin_memref::builtin_memref): Factor out parts into
6827 set_base_and_offset and call it.
6829 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6832 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6833 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6834 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6835 search at the associated MODE_INT.
6837 2018-02-20 Jeff Law <law@redhat.com>
6840 PR tree-optimization/81592
6842 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6843 for range data rather than using global data.
6844 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6845 range data rather than using global data.
6846 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6847 pass it to children as needed.
6848 (struct directive::fmtresult): Similarly.
6849 (struct directive::set_width): Similarly.
6850 (struct directive::set_precision): Similarly.
6851 (format_integer, format_directive, parse_directive): Similarly.
6852 (format_none): Accept unnamed vr_values parameter.
6853 (format_percent, format_floating, format_character): Similarly.
6854 (format_string, format_plain): Similarly.
6855 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6856 the EVRP range analyzer for range data rather than using global data.
6857 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6858 gimple-ssa-evrp-analyze.h
6859 (class sprintf_dom_walker): Add after_dom_children member function.
6860 Add evrp_range_analyzer member.
6861 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6862 range analyzer as needed.
6863 (sprintf_dom_walker::after_dom_children): New member function.
6864 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6866 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6867 (evrp_range_analyzer::pop_to_marker): Likewise.
6869 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6871 PR tree-optimization/84419
6872 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6873 with the required type if its current type is compatible but
6876 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6879 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6880 after vectorization.
6882 2018-02-20 Martin Liska <mliska@suse.cz>
6885 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6886 possible values if we don't have a hint.
6888 2018-02-20 Martin Liska <mliska@suse.cz>
6892 * final.c (shorten_branches): Build align_tab array with one
6894 * opts.c (finish_options): Add alignment option limit check.
6895 (MAX_CODE_ALIGN): Likewise.
6896 (MAX_CODE_ALIGN_VALUE): Likewise.
6897 * doc/invoke.texi: Document maximum allowed option value for
6898 all -falign-* options.
6900 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6903 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6904 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6905 * var-tracking.c (emit_note_insn_var_location): Remove all references
6906 to NOTE_INSN_CALL_ARG_LOCATION.
6907 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6908 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6909 Use copy_rtx_if_shared.
6910 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6911 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6912 (dwarf2out_var_location): Remove handling of
6913 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6915 * final.c (final_scan_insn): Remove all references to
6916 NOTE_INSN_CALL_ARG_LOCATION.
6917 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6918 before dumping final insns.
6919 * except.c (emit_note_eh_region_end): Remove all references to
6920 NOTE_INSN_CALL_ARG_LOCATION.
6921 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6922 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6923 * config/arc/arc.c (hwloop_optimize): Likewise.
6924 * config/arm/arm.c (create_fix_barrier): Likewise.
6925 * config/s390/s390.c (s390_chunkify_start): Likewise.
6926 * config/sh/sh.c (find_barrier): Likewise.
6927 * config/i386/i386.c (rest_of_insert_endbranch,
6928 ix86_seh_fixup_eh_fallthru): Likewise.
6929 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6930 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6931 * config/frv/frv.c (frv_function_prologue): Likewise.
6932 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6934 (note_outside_basic_block_p): Remove all references to
6935 NOTE_INSN_CALL_ARG_LOCATION.
6936 * gengtype.c (adjust_field_rtx_def): Likewise.
6937 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6939 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6940 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6943 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6946 PR tree-optimization/84452
6947 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6948 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6951 2018-02-19 Martin Liska <mliska@suse.cz>
6954 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6956 2018-02-19 Martin Liska <mliska@suse.cz>
6957 Richard Sandiford <richard.sandiford@linaro.org>
6959 PR tree-optimization/82491
6960 * gimple-fold.c (get_base_constructor): Make earlier bail out
6963 2018-02-19 Carl Love <cel@us.ibm.com>
6965 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6966 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6968 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6969 P8V_BUILTIN_VEC_NEG.
6971 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6973 * config/rl78/rl78.md (movdf): New define expand.
6975 2018-02-19 Martin Liska <mliska@suse.cz>
6978 * doc/invoke.texi: Fix typo.
6979 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6981 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6983 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6984 handle rs6000_single_float and rs6000_double_float specially for
6987 2018-02-16 Jeff Law <law@redhat.com>
6989 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6992 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6995 * ipa-inline.c (inline_small_functions): Fix a typo.
6997 2018-02-16 Nathan Sidwell <nathan@acm.org>
6999 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7001 2018-02-16 Carl Love <cel@us.ibm.com>
7003 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7004 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7005 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7006 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7007 expansion to P8V_BUILTIN_VEC_FLOAT2.
7009 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7011 PR rtl-optimization/70023
7012 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7013 src_regno into account.
7015 2018-02-16 Carl Love <cel@us.ibm.com>
7017 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7018 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7019 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7020 * config/rs6000/rs6000.c: Remove case statements for
7021 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7022 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7023 and P9V_BUILTIN_VEC_VINSERT4B.
7024 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7025 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7026 * config/rs6000/vsx.md:
7027 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7030 2018-02-16 Carl Love <cel@us.ibm.com>
7032 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7034 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7036 * config/rs6000/rs6000-c.c: Add the definitions for
7037 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7038 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7039 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7040 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7041 definition for insert4b and define insn *insert3b_internal.
7042 * doc/extend.texi: Add documentation for vec_extract4b.
7044 2018-02-16 Nathan Sidwell <nathan@acm.org>
7046 * doc/extend.texi (Backwards Compatibility): Mention friend
7047 injection. Note for-scope is deprecated.
7048 * doc/invoke.texi (-ffriend-injection): Deprecate.
7050 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7052 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7053 that moved to I2, also allow destinations that are a paradoxical
7054 subreg (instead of a normal reg).
7056 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7059 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7062 2018-02-16 Richard Biener <rguenther@suse.de>
7064 PR tree-optimization/84037
7065 PR tree-optimization/84016
7067 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7068 Adjust vec_construct for the fact we need additional higher latency
7069 128bit inserts for AVX256 and AVX512 vector builds.
7070 (ix86_add_stmt_cost): Scale vector construction cost for
7073 2018-02-16 Richard Biener <rguenther@suse.de>
7075 PR tree-optimization/84417
7076 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7077 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7078 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7080 2018-02-16 Martin Liska <mliska@suse.cz>
7083 * internal-fn.def (ASAN_CHECK): Set proper flags.
7084 (ASAN_MARK): Likewise.
7086 2018-02-16 Julia Koval <julia.koval@intel.com>
7088 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7089 from PTA_CANNONLAKE.
7091 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7094 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7095 Use ++iter rather than iter++ for std::list iterators.
7096 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7097 defer deleting them until all nodes in the forest are processed. Do
7098 free even leaf nodes. Change to_process into auto_vec.
7101 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7102 * vec.h (vec_default_construct): Use memset instead of placement new
7103 if BROKEN_VALUE_INITIALIZATION is defined.
7104 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7105 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7108 PR rtl-optimization/83723
7109 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7110 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7111 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7113 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7115 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7117 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7119 PR rtl-optimization/81443
7120 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7121 from inner REGs to paradoxical SUBREGs.
7123 2018-02-16 Richard Biener <rguenther@suse.de>
7125 PR tree-optimization/84399
7126 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7127 For operands we can analyze at their definition make sure we can
7128 analyze them at each use as well.
7130 2018-02-16 Richard Biener <rguenther@suse.de>
7132 PR tree-optimization/84190
7133 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7134 volatile accesses if the decl isn't volatile.
7136 2018-02-15 Jason Merrill <jason@redhat.com>
7138 PR c++/84314 - ICE with templates and fastcall attribute.
7139 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7140 TYPE_CANONICAL on an existing type.
7142 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7144 PR tree-optimization/84383
7145 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7146 dstoff nor call operand_equal_p if dstbase is NULL.
7148 PR tree-optimization/84334
7149 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7150 also a CONSTANT_CLASS_P, punt.
7152 2018-02-14 Jim Wilson <jimw@sifive.com>
7154 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7155 first SMALL_OPERAND check. New local min_second_step. Move assert
7156 to where locals are set. Add TARGET_RVC support.
7157 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7159 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7161 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7163 2018-02-14 Martin Sebor <msebor@redhat.com>
7165 PR tree-optimization/83698
7166 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7167 arrays constrain the offset range to their bounds.
7168 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7169 (builtin_access::overlap): Avoid setting the size of overlap if it's
7171 (maybe_diag_overlap): Also consider arrays when deciding what values
7172 of offsets to include in diagnostics.
7174 2018-02-14 Martin Sebor <msebor@redhat.com>
7177 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7178 that correspond to the kind of a declaration.
7180 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7183 * config/pa/pa.md: Load address of PIC label using the linkage table
7184 if the label is nonlocal.
7186 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7188 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7189 warning message if user requests -maltivec=be.
7190 * doc/invoke.texi: Document deprecation of -maltivec=be.
7192 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7195 * config/rs6000/rs6000-c.c: Update definitions for
7196 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7197 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7199 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7202 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7203 add _get_ssp intrinsics. Remove argument from
7204 __builtin_ia32_rdssp[d|q].
7205 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7206 * config/i386/i386-builtin.def: Remove argument from
7207 __builtin_ia32_rdssp[d|q].
7208 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7209 ix86_expand_special_args_builtin for _rdssp[d|q].
7210 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7211 Clear register before usage.
7212 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7213 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7215 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7217 PR tree-optimization/84357
7218 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7219 operand 1 of an ARRAY_REF too.
7221 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7224 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7225 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7227 (set_of_reg): New struct.
7228 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7229 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7230 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7232 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7233 Split into bitclr, bitset, bitinvert patterns if appropriate.
7234 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7235 use rx_fuse_in_memory_bitop.
7236 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7237 to named insn, correct maximum insn length.
7239 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7242 * machmode.def: Define a complex mode for PARTIAL_INT.
7243 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7245 * doc/rtl.texi: Document CSPImode.
7246 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7248 (msp430_hard_regno_nregs_with_padding): Likewise.
7250 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7253 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7255 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7257 PR rtl-optimization/84169
7258 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7259 we generated a parallel as new i3 and we split that to new i2 and i3
7260 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7261 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7262 those to i2, not i1. Partially rewrite this scan code.
7264 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7267 * stor-layout.c (place_field): For variable length fields, adjust
7268 offset_align afterwards not just based on the field's alignment,
7269 but also on the size.
7272 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7273 of exps and logs in the use_exp2 case.
7275 2018-02-13 Jeff Law <law@redhat.com>
7277 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7280 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7283 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7287 * common.opt (gas-loc-support, gas-locview-support): New.
7288 (ginline-points, ginternal-reset-location-views): New.
7289 * doc/invoke.texi: Document them. Use @itemx where intended.
7290 (gvariable-location-views): Adjust.
7291 * target.def (reset_location_view): New.
7292 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7293 (TARGET_RESET_LOCATION_VIEW): New.
7294 * doc/tm.texi: Rebuilt.
7295 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7296 (dwarf2out_default_as_locview_support): New.
7297 (output_asm_line_debug_info): Use option variables.
7298 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7299 (output_loc_list): Likewise.
7300 (add_high_low_attributes): Check option variables.
7301 Don't output entry view attribute in strict mode.
7302 (gen_inlined_subroutine_die): Check option variables.
7303 (dwarf2out_inline_entry): Likewise.
7304 (init_sections_and_labels): Likewise.
7305 (dwarf2out_early_finish): Likewise.
7306 (maybe_reset_location_view): New, from...
7307 (dwarf2out_var_location): ... here. Call it.
7308 * debug.h (dwarf2out_default_as_loc_support): Declare.
7309 (dwarf2out_default_as_locview_support): Declare.
7310 * hooks.c (hook_int_rtx_insn_0): New.
7311 * hooks.h (hook_int_rtx_insn_0): Declare.
7312 * toplev.c (process_options): Take -gas-loc-support and
7313 -gas-locview-support from dwarf2out. Enable
7314 -gvariable-location-views by default only with locview
7315 assembler support. Enable -ginternal-reset-location-views by
7316 default only if the target defines the corresponding hook.
7317 Enable -ginline-points by default if location views are
7318 enabled; force it disabled if statement frontiers are
7320 * tree-inline.c (expand_call_inline): Check option variables.
7321 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7323 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7325 PR tree-optimization/84321
7326 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7327 handling. Also check whether the anti-range contains any values
7328 that satisfy the mask; switch to a VR_RANGE if not.
7330 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7333 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7335 2018-02-13 Martin Jambor <mjambor@suse.cz>
7338 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7339 of call statements, also set location of a load to a temporary.
7341 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7343 * config/rl78/rl78.c (add_vector_labels): New function.
7344 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7345 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7346 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7347 which checks that no arguments are passed.
7348 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7349 * doc/extend.texi: Documentation for the new attribute.
7351 2018-02-13 Andreas Schwab <schwab@suse.de>
7353 * config/riscv/linux.h (CPP_SPEC): Define.
7355 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7358 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7359 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7360 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7361 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7362 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7363 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7364 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7365 temporarily for AES and PCLMUL builtins.
7367 PR tree-optimization/84339
7368 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7369 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7373 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7374 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7375 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7377 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7379 * omp-simd-clone.h: New file.
7380 * omp-simd-clone.c: Include omp-simd-clone.h.
7381 (expand_simd_clones): No longer static.
7382 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7383 cgraph.h and omp-simd-clone.h.
7384 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7385 (vect_recog_widen_shift_pattern): Formatting fix.
7386 (vect_pattern_recog_1): Don't check optab for calls.
7389 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7390 operands[2] into a REG before using gen_lowpart on it.
7392 2018-02-12 Jeff Law <law@redhat.com>
7395 * config/sh/sh.c (find_barrier): Consider a sibling call
7398 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7399 successfully back substituting a reg.
7401 2018-02-12 Richard Biener <rguenther@suse.de>
7403 PR tree-optimization/84037
7404 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7405 parameter, move visited init to caller.
7406 (vect_slp_analyze_operations): Separate cost from validity
7407 check, initialize visited once for all instances.
7408 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7410 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7412 (vect_model_promotion_demotion_cost): Likewise.
7413 (vectorizable_bswap): Guard cost modeling with !slp_node
7414 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7416 (vectorizable_call): Likewise.
7417 (vectorizable_conversion): Likewise.
7418 (vectorizable_assignment): Likewise.
7419 (vectorizable_shift): Likewise.
7420 (vectorizable_operation): Likewise.
7421 (vectorizable_store): Likewise.
7422 (vectorizable_load): Likewise.
7423 (vectorizable_condition): Likewise.
7424 (vectorizable_comparison): Likewise.
7426 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7429 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7430 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7431 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7433 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7436 * params.def (inline-min-speedup): Increase from 8 to 15.
7437 (max-inline-insns-auto): Decrease from 40 to 30.
7438 * ipa-split.c (consider_split): Add some buffer for function to
7439 be considered inlining candidate.
7440 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7443 2018-02-12 Richard Biener <rguenther@suse.de>
7445 PR tree-optimization/84037
7446 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7447 matched stmts if we cannot swap the non-matched ones.
7449 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7451 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7452 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7453 _mm_maskz_scalef_round_ss): New intrinsics.
7454 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7455 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7456 __builtin_ia32_scalefss_round): Remove.
7457 (__builtin_ia32_scalefsd_mask_round,
7458 __builtin_ia32_scalefss_mask_round): New intrinsics.
7459 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7460 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7461 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7462 "<round_constraint>")): Changed to ...
7463 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7464 "<round_scalar_constraint>")): ... this.
7465 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7466 %0, %1, %2<round_op3>}"): Changed to ...
7467 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7468 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7469 %2<round_scalar_mask_op3>}"): ... this.
7470 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7472 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7474 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7475 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7476 (_mm_maskz_sqrt_round_ss): New intrinsics.
7477 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7478 (__builtin_ia32_sqrtsd_mask_round)
7479 (__builtin_ia32_sqrtss_mask_round): New builtins.
7480 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7481 (__builtin_ia32_sqrtss_round): Remove.
7482 (__builtin_ia32_sqrtsd_mask_round)
7483 (__builtin_ia32_sqrtss_mask_round): New builtins.
7484 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7485 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7486 ((match_operand:VF_128 1 "vector_operand"
7487 "xBm,<round_constraint>")): Changed to ...
7488 ((match_operand:VF_128 1 "vector_operand"
7489 "xBm,<round_scalar_constraint>")): ... this.
7490 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7491 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7492 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7493 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7494 %<iptr>1<round_scalar_mask_op3>}): ... this.
7495 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7496 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7498 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7501 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7502 Cast vec_cmpeq result to correct type.
7503 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7504 Cast vec_cmpgt result to correct type.
7506 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7508 * final.c (final_scan_insn_1): Renamed from...
7509 (final_scan_insn): ... this. New wrapper, to recover
7510 seen from the outermost call in recursive ones.
7511 * config/sparc/sparc.c (output_return): Drop seen from call.
7512 (output_sibcall): Likewise.
7513 * config/visium/visium.c (output_branch): Likewise.
7515 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7517 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7520 2018-02-10 Alan Modra <amodra@gmail.com>
7523 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7524 Specify LR as an input.
7526 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7529 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7530 remove_member_access_dummy_vars): New functions.
7531 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7532 lower_omp_1, execute_lower_omp): Use them.
7534 PR rtl-optimization/84308
7535 * shrink-wrap.c (spread_components): Release todo vector.
7537 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7539 PR rtl-optimization/57193
7540 * ira-color.c (struct allocno_color_data): Add member
7541 conflict_allocno_hard_prefs.
7542 (update_conflict_allocno_hard_prefs): New.
7543 (bucket_allocno_compare_func): Add a preference based on
7544 conflict_allocno_hard_prefs.
7545 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7546 (color_allocnos): Remove a dead code. Initiate
7547 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7549 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7552 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7553 constraint from =wa to wa. Avoid a subreg on the output operand,
7554 instead use a pseudo and subreg it in a move.
7555 (p9_xxbrd_<mode>): Changed to ...
7556 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7557 (p9_xxbrd_v2df): New expander.
7558 (p9_xxbrw_<mode>): Changed to ...
7559 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7560 (p9_xxbrw_v4sf): New expander.
7562 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7564 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7566 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7569 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7570 multiply in 32-bit mode.
7571 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7572 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7575 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7577 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7578 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7579 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7580 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7582 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7585 * dwarf2out.c (is_trivial_indirect_ref): New function.
7586 (dwarf2out_late_global_decl): Do not generate a location
7587 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7588 and that are not defined in the current unit.
7590 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7592 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7593 instead of a libcall for UNORDERED.
7595 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7598 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7599 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7601 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7604 * config/s390/s390.c (s390_set_current_function): Invoke
7605 s390_indirect_branch_settings also if fndecl didn't change.
7607 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7609 * config/rs6000/rs6000.md (blockage): Set length to zero.
7611 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7613 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7615 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7618 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7619 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7623 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7624 PARALLEL incoming that failed vt_get_decl_and_offset check.
7627 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7628 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7630 (get_variable_section): For decls in named .bss* sections pass true as
7631 second argument to bss_initializer_p.
7633 2018-02-09 Marek Polacek <polacek@redhat.com>
7634 Jakub Jelinek <jakub@redhat.com>
7637 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7638 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7639 Sync some changes from cxx_fold_indirect_ref.
7641 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7643 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7645 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7646 (BLOCK_INLINE_ENTRY_LABEL): New.
7647 (dwarf2out_var_location): Disregard inline entry markers.
7648 (inline_entry_data): New struct.
7649 (inline_entry_data_hasher): New hashtable type.
7650 (inline_entry_data_hasher::hash): New.
7651 (inline_entry_data_hasher::equal): New.
7652 (inline_entry_data_table): New variable.
7653 (add_high_low_attributes): Add DW_AT_entry_pc and
7654 DW_AT_GNU_entry_view attributes if a pending entry is found
7655 in inline_entry_data_table. Add old entry_pc attribute only
7656 if debug nonbinding markers are disabled.
7657 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7658 markers are enabled.
7659 (block_within_block_p, dwarf2out_inline_entry): New.
7660 (dwarf2out_finish): Check that no entries remained in
7661 inline_entry_data_table.
7662 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7663 (final_scan_insn, notice_source_line): Likewise.
7664 (rest_of_clean_state): Skip inline entry markers.
7665 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7667 * gimple.c (gimple_build_debug_inline_entry): New.
7668 * gimple.h (enum gimple_debug_subcode): Add
7669 GIMPLE_DEBUG_INLINE_ENTRY.
7670 (gimple_build_debug_inline_entry): Declare.
7671 (gimple_debug_inline_entry_p): New.
7672 (gimple_debug_nonbind_marker_p): Adjust.
7673 * insn-notes.def (INLINE_ENTRY): New.
7674 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7675 inline entry marker notes.
7676 (print_insn): Likewise.
7677 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7678 (INSN_DEBUG_MARKER_KIND): Likewise.
7679 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7680 * tree-inline.c (expand_call_inline): Build and insert
7681 debug_inline_entry stmt.
7682 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7683 inline entry blocks early, if nonbind markers are enabled.
7684 (dump_scope_block): Dump fragment info.
7685 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7686 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7687 (gimple_build_debug_inline_entry): New.
7688 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7689 Enable/disable inline entry points too.
7690 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7691 (DEBUG_INSN): Describe inline entry markers.
7693 * common.opt (gvariable-location-views): New.
7694 (gvariable-location-views=incompat5): New.
7695 * config.in: Rebuilt.
7696 * configure: Rebuilt.
7697 * configure.ac: Test assembler for view support.
7698 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7699 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7700 * dwarf2out.c (var_loc_view): New typedef.
7701 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7702 (dwarf2out_locviews_in_attribute): New.
7703 (dwarf2out_locviews_in_loclist): New.
7704 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7705 (enum dw_line_info_opcode): Add LI_adv_address.
7706 (struct dw_line_info_table): Add view.
7707 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7708 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7709 (zero_view_p): New variable.
7710 (ZERO_VIEW_P): New macro.
7711 (output_asm_line_debug_info): New.
7712 (struct var_loc_node): Add view.
7713 (add_AT_view_list, AT_loc_list): New.
7714 (add_var_loc_to_decl): Add view param. Test it against last.
7715 (new_loc_list): Add view params. Record them.
7716 (AT_loc_list_ptr): Handle loc and view lists.
7717 (view_list_to_loc_list_val_node): New.
7718 (print_dw_val): Handle dw_val_class_view_list.
7719 (size_of_die): Likewise.
7720 (value_format): Likewise.
7721 (loc_list_has_views): New.
7722 (gen_llsym): Set vl_symbol too.
7723 (maybe_gen_llsym, skip_loc_list_entry): New.
7724 (dwarf2out_maybe_output_loclist_view_pair): New.
7725 (output_loc_list): Output view list or entries too.
7726 (output_view_list_offset): New.
7727 (output_die): Handle dw_val_class_view_list.
7728 (output_dwarf_version): New.
7729 (output_compilation_unit_header): Use it.
7730 (output_skeleton_debug_sections): Likewise.
7731 (output_rnglists, output_line_info): Likewise.
7732 (output_pubnames, output_aranges): Update version comments.
7733 (output_one_line_info_table): Output view numbers in asm comments.
7734 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7735 Call maybe_gen_llsym.
7736 (loc_list_from_tree_1): Adjust.
7737 (add_AT_location_description): Create view list attribute if
7738 needed, check it's absent otherwise.
7739 (convert_cfa_to_fb_loc_list): Adjust.
7740 (maybe_emit_file): Call output_asm_line_debug_info for test.
7741 (dwarf2out_var_location): Reset views as needed. Precompute
7742 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7743 attribute. Set view.
7744 (new_line_info_table): Reset next view.
7745 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7746 (dwarf2out_source_line): Likewise. Output view resets and labels to
7747 the assembler, or select appropriate line info opcodes.
7748 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7749 (optimize_string_length): Catch it. Adjust.
7750 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7751 dw_val_class_view_list, and remove it if no longer needed.
7752 (hash_loc_list): Hash view numbers.
7753 (loc_list_hasher::equal): Compare them.
7754 (optimize_location_lists): Check whether a view list symbol is
7755 needed, and whether the locview attribute is present, and
7756 whether they match. Remove the locview attribute if no longer
7758 (index_location_lists): Call skip_loc_list_entry for test.
7759 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7760 Use output_dwarf_version.
7761 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7762 (struct dw_val_node): Add val_view_list.
7763 * final.c (SEEN_NEXT_VIEW): New.
7764 (set_next_view_needed): New.
7765 (clear_next_view_needed): New.
7766 (maybe_output_next_view): New.
7767 (final_start_function): Rename to...
7768 (final_start_function_1): ... this. Take pointer to FIRST,
7769 add SEEN parameter. Emit param bindings in the initial view.
7770 (final_start_function): Reintroduce SEEN-less interface.
7771 (final): Rename to...
7772 (final_1): ... this. Take SEEN parameter. Output final pending
7773 next view at the end.
7774 (final): Reintroduce seen-less interface.
7775 (final_scan_insn): Output pending next view before switching
7776 sections or ending a block. Mark the next view as needed when
7777 outputting variable locations. Notify debug backend of section
7778 changes, and of location view changes.
7779 (rest_of_handle_final): Adjust.
7780 * toplev.c (process_options): Autodetect value for debug variable
7781 location views option. Warn on incompat5 without -gdwarf-5.
7782 * doc/invoke.texi (gvariable-location-views): New.
7783 (gvariable-location-views=incompat5): New.
7784 (gno-variable-location-views): New.
7786 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7788 PR tree-optimization/84136
7789 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7790 that the result of find_edge is non-NULL.
7792 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7795 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7796 storing integer register in SImode. Fix cost of 256 and 512
7797 byte aligned SSE register store.
7799 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7801 * config/i386/i386.c (ix86_multiplication_cost): Fix
7802 multiplication cost for TARGET_AVX512DQ.
7804 2018-02-08 Marek Polacek <polacek@redhat.com>
7806 PR tree-optimization/84238
7807 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7810 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7812 PR tree-optimization/84265
7813 * tree-vect-stmts.c (vectorizable_store): Don't treat
7814 VMAT_CONTIGUOUS accesses as grouped.
7815 (vectorizable_load): Likewise.
7817 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7819 PR tree-optimization/81635
7820 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7821 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7822 (test_round_for_mask): New functions.
7823 (wide_int_cc_tests): Call test_round_for_mask.
7824 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7825 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7826 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7827 range returned by get_range_info.
7829 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7832 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7833 * symtab.c: Include builtins.h
7834 (symtab_node::output_to_lto_symbol_table_p): Move here
7835 from lto-streamer-out.c:output_symbol_p.
7836 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7837 (output_symbol_p): Move all logic to symtab.c
7838 (produce_symtab): Update.
7840 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7842 * config/s390/s390-opts.h (enum indirect_branch): Define.
7843 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7844 (s390_indirect_branch_via_thunk)
7845 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7846 (enum s390_indirect_branch_type): Define.
7847 * config/s390/s390.c (struct s390_frame_layout, struct
7848 machine_function): Remove.
7849 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7850 (indirect_branch_table_label_no, indirect_branch_table_name):
7852 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7853 (enum s390_indirect_branch_option): Define.
7854 (s390_return_addr_from_memory): New function.
7855 (s390_handle_string_attribute): New function.
7856 (s390_attribute_table): Add new attribute handler.
7857 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7858 (s390_indirect_branch_via_thunk): New function.
7859 (s390_indirect_branch_via_inline_thunk): New function.
7860 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7861 sibling call optimization for non z10 compiles.
7862 (s390_emit_call): Force indirect branch target to be a single
7863 register. Add r1 clobber for non-z10 compiles.
7864 (s390_emit_epilogue): Emit return jump via return_use expander.
7865 (s390_reorg): Handle JUMP_INSNs as execute targets.
7866 (s390_option_override_internal): Perform validity checks for the
7867 new command line options.
7868 (s390_indirect_branch_attrvalue): New function.
7869 (s390_indirect_branch_settings): New function.
7870 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7871 (s390_output_indirect_thunk_function): New function.
7872 (s390_code_end): Implement target hook.
7873 (s390_case_values_threshold): Implement target hook.
7874 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7876 * config/s390/s390.h (struct s390_frame_layout)
7877 (struct machine_function): Move here from s390.c.
7878 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7879 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7880 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7881 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7882 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7883 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7884 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7885 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7886 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7887 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7888 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7889 (mnemonic attribute): Add values which aren't recognized
7891 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7892 pattern for branch conversion. Fix mnemonic attribute.
7893 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7894 indirect branch via thunk if requested.
7895 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7896 ("*indirect_jump"): Disable for branch conversion using out of
7898 ("indirect_jump_via_thunk<mode>_z10")
7899 ("indirect_jump_via_thunk<mode>")
7900 ("indirect_jump_via_inlinethunk<mode>_z10")
7901 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7902 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7903 ("casesi_jump_via_inlinethunk<mode>_z10")
7904 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7905 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7906 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7907 ("*indirect2_jump"): Disable for branch conversion.
7908 ("casesi_jump"): Turn into expander and expand patterns for branch
7910 ("return_use"): New expander.
7911 ("*return"): Emit return via thunk and rename it to ...
7912 ("*return<mode>"): ... this one.
7913 * config/s390/s390.opt: Add new options and and enum for the
7916 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7918 * lra-constraints.c (match_reload): Unconditionally use
7919 gen_lowpart_SUBREG, rather than selecting between that
7920 and equivalent gen_rtx_SUBREG code.
7922 2018-02-08 Richard Biener <rguenther@suse.de>
7924 PR tree-optimization/84233
7925 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7926 changed flag instead of boguously re-using phi_inserted.
7928 2018-02-08 Martin Jambor <mjambor@suse.cz>
7930 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7931 static local variables.
7933 2018-02-08 Richard Biener <rguenther@suse.de>
7935 PR tree-optimization/84278
7936 * tree-vect-stmts.c (vectorizable_store): When looking for
7937 smaller vector types to perform grouped strided loads/stores
7938 make sure the mode is supported by the target.
7939 (vectorizable_load): Likewise.
7941 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7943 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7944 Increase LDP/STP opportunities by adding adjacent callee-saves.
7946 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7948 PR rtl-optimization/84068
7949 PR rtl-optimization/83459
7950 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7952 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7954 PR tree-optimization/84224
7955 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7956 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7959 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7962 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7963 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7964 count, remove one USE.
7966 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7968 * doc/install.texi (Configuration): Document the
7969 --with-long-double-format={ibm,ieee} PowerPC configuration
7973 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7974 Convert from define_expand to be define_insn_and_split. Rework
7975 float/double/_Float128 conversions to QI/HI/SImode to work with
7976 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7977 conversions to QI/HImode types did a store and then a load to
7978 truncate the value. For conversions to VSX registers, don't split
7979 the insn, instead emit the code directly. Use the code iterator
7980 any_fix to combine signed and unsigned conversions.
7981 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7982 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7983 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7984 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7985 (fix_<mode>di2_hw): Likewise.
7986 (fixuns_<mode>di2_hw): Likewise.
7987 (fix_<mode>si2_hw): Likewise.
7988 (fixuns_<mode>si2_hw): Likewise.
7989 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7990 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7991 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7992 fix<uns>_trunc<SFDF:mode>si2_p8.
7993 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7995 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7996 (fix<uns>_<mode>_mem): Likewise.
7997 (fctiw<u>z_<mode>_mem): Likewise.
7998 (fix<uns>_<mode>_mem): Likewise.
7999 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8000 the register allocator from doing a direct move to the GPRs to do
8001 a store, and instead use the ISA 3.0 store byte/half-word from
8002 vector register instruction. For IEEE 128-bit floating point,
8003 also optimize stores of 32-bit ints.
8004 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8006 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8008 * genextract.c (push_pathstr_operand): New function to support
8010 (walk_rtx): Call push_pathstr_operand.
8011 (print_path): Support [a-zA-Z].
8013 2018-02-07 Richard Biener <rguenther@suse.de>
8015 PR tree-optimization/84037
8016 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8017 (cse_and_gimplify_to_preheader): Declare.
8018 (vect_get_place_in_interleaving_chain): Likewise.
8019 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8021 (_loop_vec_info::~_loop_vec_info): Delete it.
8022 (cse_and_gimplify_to_preheader): New function.
8023 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8024 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8025 (vectorizable_load): Likewise. For grouped stores always base
8026 the IV on the first element.
8027 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8028 condition before gimplifying.
8030 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8032 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8033 *DIV_EXPR and *MOD_EXPR.
8035 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8038 * config/i386/i386.c (ix86_option_override_internal): Mask out
8039 the CF_SET bit when checking -fcf-protection.
8041 2018-02-07 Tom de Vries <tom@codesourcery.com>
8044 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8047 2018-02-07 Richard Biener <rguenther@suse.de>
8049 PR tree-optimization/84204
8050 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8053 PR tree-optimization/84205
8054 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8055 special-case isl_ast_op_zdiv_r.
8057 PR tree-optimization/84223
8058 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8059 Only add conditions from within the region.
8060 (gather_bbs::after_dom_children): Adjust.
8062 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8065 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8066 * config/avr/avr.md: Only post-reload split REG-REG moves if
8067 either register is GENERAL_REG_P.
8069 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8071 PR tree-optimization/84235
8072 * tree-ssa-scopedtables.c
8073 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8074 if the subtraction is performed in floating point type where NaNs are
8075 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8076 build 1. Formatting fix.
8078 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8081 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8082 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8083 and skip it regardless of bb boundaries. Use CALL_P macro,
8084 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8085 unnecessarily, formatting fix.
8087 2018-02-06 Michael Collison <michael.collison@arm.com>
8089 * config/arm/thumb2.md:
8090 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8091 (*thumb_mov_notscc): Ditto.
8093 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8096 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8097 unsigned_fix, not "s".
8099 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8101 * configure.ac (gcc_fn_eh_frame_ro): New function.
8102 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8103 correct .eh_frame permissions.
8104 * configure: Regenerate.
8106 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8108 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8111 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8113 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8114 Display warning message for -mno-speculate-indirect-jumps.
8116 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8118 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8120 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8122 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8124 PR tree-optimization/84225
8125 * tree-eh.c (find_trapping_overflow): Only call
8126 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8128 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8131 * config/i386/i386.c: Reimplement the check of possible options
8132 -mibt/-mshstk conbination. Change error messages.
8133 * doc/invoke.texi: Fix a typo: remove extra '='.
8135 2018-02-06 Marek Polacek <polacek@redhat.com>
8137 PR tree-optimization/84228
8138 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8140 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8143 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8144 emitted arch directives.
8145 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8146 __ARM_FEATURE_COPROC before changing architectures.
8148 2018-02-06 Richard Biener <rguenther@suse.de>
8150 * config/i386/i386.c (print_reg): Fix typo.
8151 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8153 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8155 * configure: Regenerate.
8157 2018-02-05 Martin Sebor <msebor@redhat.com>
8159 PR tree-optimization/83369
8160 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8163 2018-02-05 Martin Liska <mliska@suse.cz>
8165 * doc/invoke.texi: Cherry-pick upstream r323995.
8167 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8169 * ira.c (ira_init_register_move_cost): Adjust comment.
8171 2018-02-05 Martin Liska <mliska@suse.cz>
8173 PR gcov-profile/84137
8174 * doc/gcov.texi: Fix typo in documentation.
8176 2018-02-05 Martin Liska <mliska@suse.cz>
8178 PR gcov-profile/83879
8179 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8180 using dlopen functionality.
8182 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8184 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8185 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8186 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8187 _mm_maskz_range_round_ss): New intrinsics.
8188 (__builtin_ia32_rangesd128_round)
8189 (__builtin_ia32_rangess128_round): Remove.
8190 (__builtin_ia32_rangesd128_mask_round,
8191 __builtin_ia32_rangess128_mask_round): New builtins.
8192 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8193 __builtin_ia32_rangess128_round): Remove.
8194 (__builtin_ia32_rangesd128_mask_round,
8195 __builtin_ia32_rangess128_mask_round): New builtins.
8196 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8197 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8198 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8199 "<round_saeonly_constraint>")): Changed to ...
8200 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8201 "<round_saeonly_scalar_constraint>")): ... this.
8202 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8203 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8204 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8205 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8206 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8208 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8210 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8212 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8213 Remove all values except native, 8540 and 8548.
8215 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8217 * config/i386/i386.c (ix86_output_function_return): Pass
8218 INVALID_REGNUM, instead of -1, as invalid register number to
8219 indirect_thunk_name and output_indirect_thunk.
8221 2018-02-02 Julia Koval <julia.koval@intel.com>
8223 * config.gcc: Add -march=icelake.
8224 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8225 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8226 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8227 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8228 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8229 (processor_target_table): Add icelake.
8230 (ix86_option_override_internal): Handle new PTAs.
8231 (get_builtin_code_for_version): Handle icelake.
8232 (M_INTEL_COREI7_ICELAKE): New.
8233 (fold_builtin_cpu): Handle icelake.
8234 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8235 * doc/invoke.texi: Add -march=icelake.
8237 2018-02-02 Julia Koval <julia.koval@intel.com>
8239 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8240 to wide_int_bitmask.
8241 * wide-int-bitmask.h: New.
8243 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8246 * config/i386/i386.md: Replace Pmode with word_mode in
8247 builtin_setjmp_setup and builtin_longjmp to support x32.
8249 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8253 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8255 (rs6000_supported_cpu_names): New static variable.
8256 (linux_cpu_translation_table): Likewise.
8257 (elf_platform) <cpu>: Define new static variable and use it.
8258 Translate kernel AT_PLATFORM name to canonical name if needed.
8259 Error if platform name is unknown.
8261 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8264 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8266 2018-02-01 Jeff Law <law@redhat.com>
8269 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8270 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8271 the scratch if RELEASE_VIA_POP is false.
8272 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8273 If we have to save a temporary register, decrement SIZE appropriately.
8274 Pass new arguments to release_scratch_register_on_entry.
8275 (ix86_adjust_stack_and_probe): Likewise.
8276 (ix86_emit_probe_stack_range): Pass new arguments to
8277 release_scratch_register_on_entry.
8279 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8281 PR rtl-optimization/84157
8282 * combine.c (change_zero_ext): Use REG_P predicate in
8283 front of HARD_REGISTER_P predicate.
8285 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8287 * config/avr/avr.c (avr_option_override): Move disabling of
8288 -fdelete-null-pointer-checks to...
8289 * common/config/avr/avr-common.c (avr_option_optimization_table):
8292 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8294 PR tree-optimization/81635
8295 * tree-data-ref.c (split_constant_offset_1): For types that
8296 wrap on overflow, try to use range info to prove that wrapping
8299 2018-02-01 Renlin Li <renlin.li@arm.com>
8302 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8304 (aarch64_register_move_cost): Likewise.
8305 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8307 (REG_CLASS_NAMES): Likewise.
8308 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8309 TAILCALL_ADDR_REGS. Remove IP registers.
8310 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8312 2018-02-01 Richard Biener <rguenther@suse.de>
8314 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8315 for specifying RPO order and allow NULL for that.
8316 * domwalk.c (dom_walker::dom_walker): Likewise.
8317 (dom_walker::walk): Handle NULL RPO order.
8318 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8320 (rewrite_update_dom_walker): Likewise.
8321 (mark_def_dom_walker): Likewise.
8323 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8325 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8326 (aarch64_maybe_expand_sve_subreg_move): Declare.
8327 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8328 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8330 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8331 that are semantically a reverse operation.
8332 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8333 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8334 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8336 (aarch64_can_change_mode_class): For big-endian, forbid changes
8337 between two SVE modes if they have different element sizes.
8339 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8341 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8342 the TImode handling for big-endian targets.
8344 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8346 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8347 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8349 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8350 Remove BSWAP handing for big-endian targets and use the form of
8351 LD1RQ appropariate for the mode.
8353 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8355 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8356 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8359 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8362 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8363 check for operands that need to go through aarch64_sve_reload_be.
8365 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8367 PR tree-optimization/81661
8368 PR tree-optimization/84117
8369 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8370 * tree-eh.c: Include gimplify.h.
8371 (find_trapping_overflow, replace_trapping_overflow,
8372 rewrite_to_non_trapping_overflow): New functions.
8373 * tree-vect-loop.c: Include tree-eh.h.
8374 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8375 * tree-data-ref.c: Include tree-eh.h.
8376 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8378 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8380 PR rtl-optimization/84123
8381 * combine.c (change_zero_ext): Check if hard register satisfies
8382 can_change_dest_mode before calling gen_lowpart_SUBREG.
8384 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8387 * ira.c (ira_init_register_move_cost): Remove assert.
8389 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8391 PR rtl-optimization/84071
8392 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8393 * doc/tm.texi: Regenerate.
8395 2018-01-31 Richard Biener <rguenther@suse.de>
8397 PR tree-optimization/84132
8398 * tree-data-ref.c (analyze_miv_subscript): Properly
8399 check whether evolution_function_is_affine_multivariate_p
8400 before calling gcd_of_steps_may_divide_p.
8402 2018-01-31 Julia Koval <julia.koval@intel.com>
8405 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8406 * config/i386/i386.md (rdpid_rex64) New.
8407 (rdpid): Make 32bit only.
8409 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8412 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8413 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8415 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8418 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8420 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8422 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8424 PR rtl-optimization/84071
8425 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8426 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8428 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8430 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8431 (arc_attribute_table): Add 'aux' attribute.
8432 (arc_in_small_data_p): Consider aux like variables.
8433 (arc_is_aux_reg_p): New function.
8434 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8435 (arc_get_aux_arg): New function.
8436 (prepare_move_operands): Handle aux-register access.
8437 (arc_handle_aux_attribute): New function.
8438 * doc/extend.texi (ARC Variable attributes): Add subsection.
8440 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8442 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8443 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8444 (arc_attribute_table): Add 'uncached' attribute.
8445 (arc_print_operand): Print '.di' flag for uncached memory
8447 (arc_in_small_data_p): Do not consider for small data the uncached
8449 (arc_is_uncached_mem_p): New function.
8450 * config/arc/predicates.md (compact_store_memory_operand): Check
8451 for uncached memory accesses.
8452 (nonvol_nonimm_operand): Likewise.
8453 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8455 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8458 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8459 falign-loops=): Add Optimization flag.
8461 2018-01-30 Jeff Law <law@redhat.com>
8464 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8465 INT_REGISTERS_SAVED. Check it prior to calling
8466 get_scratch_register_on_entry.
8467 (ix86_adjust_stack_and_probe): Similarly.
8468 (ix86_emit_probe_stack_range): Similarly.
8469 (ix86_expand_prologue): Corresponding changes.
8471 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8474 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8475 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8477 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8480 * lra-constraints.c (curr_insn_transform): Process AND in the
8483 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8485 PR rtl-optimization/83986
8486 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8487 dependence against last_pending_memory_flush in addition to
8490 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8492 PR tree-optimization/81611
8493 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8496 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8499 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8502 2018-01-30 Richard Biener <rguenther@suse.de>
8503 Jakub Jelinek <jakub@redhat.com>
8505 PR tree-optimization/84111
8506 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8507 inner loops added during recursion, as they don't have up-to-date
8510 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8513 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8514 (can_inline_edge_by_limits_p): ... here.
8515 (can_early_inline_edge_p, check_callers,
8516 update_caller_keys, update_callee_keys, recursive_inlining,
8517 add_new_edges_to_heap, speculation_useful_p,
8518 inline_small_functions,
8519 inline_small_functions, flatten_function,
8520 inline_to_all_callers_1): Update.
8522 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8524 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8527 2018-01-30 Richard Biener <rguenther@suse.de>
8529 PR tree-optimization/83008
8530 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8531 invariant and constant vector uses in stmts when they need
8534 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8537 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8538 * configure: Regenerate.
8540 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8542 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8544 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8545 Use gen_rtx_REG rather than gen_lowpart.
8547 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8549 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8550 rather than 0 when creating partial subregs.
8552 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8554 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8557 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8560 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8561 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8562 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8563 flags. This restores the settings used before the 2017-07-24.
8564 Turning off pre increment/decrement/modify allows IVOPTS to
8565 optimize DF/SF loops where the index is an int.
8567 2018-01-29 Richard Biener <rguenther@suse.de>
8568 Kelvin Nilsen <kelvin@gcc.gnu.org>
8571 * tree-vect-stmts.c (vectorizable_call): Don't call
8572 targetm.vectorize_builtin_md_vectorized_function if callee is
8575 2018-01-22 Carl Love <cel@us.ibm.com>
8577 * doc/extend.tex: Fix typo in second arg in
8578 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8580 2018-01-29 Richard Biener <rguenther@suse.de>
8582 PR tree-optimization/84086
8583 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8584 (flush_ssaname_freelist): When SSA names were released reset
8585 the SCEV hash table.
8587 2018-01-29 Richard Biener <rguenther@suse.de>
8589 PR tree-optimization/84057
8590 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8591 removed paths when removing edges.
8593 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8595 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8596 -mfunction-return=@var{choice}.
8598 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8601 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8603 (layout::print_source_line): Likewise.
8604 (test_get_line_width_without_trailing_whitespace): Add test cases.
8606 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8609 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8612 2018-01-26 Jim Wilson <jimw@sifive.com>
8614 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8616 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8619 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8621 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8622 and CMP + SUB-immediate -> SUBS.
8624 2018-01-26 Martin Sebor <msebor@redhat.com>
8626 PR tree-optimization/83896
8627 * tree-ssa-strlen.c (get_string_len): Rename...
8628 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8629 Avoid assuming length is constant.
8630 (handle_char_store): Use HOST_WIDE_INT for string length.
8632 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8635 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8636 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8638 2018-01-26 Richard Biener <rguenther@suse.de>
8640 PR rtl-optimization/84003
8641 * dse.c (record_store): Only record redundant stores when
8642 the earlier store aliases at least all accesses the later one does.
8644 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8646 PR rtl-optimization/83985
8647 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8648 REG_CFA_RESTORE insns.
8649 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8650 REG_CFA_RESTORE insns here.
8653 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8654 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8656 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8658 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8659 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8660 (arc_init): Likewise.
8661 (arc_override_options): Likewise.
8662 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8664 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8666 * config/arc/arc.h (TARGET_DBNZ): Define.
8667 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8668 properly set the tune attribute.
8669 (dbnz): Use TARGET_DBNZ guard.
8670 * config/arc/arc.opt (mtune): Add core3 option.
8672 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8674 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8675 recognize new pic like addresses.
8676 (arc_delegitimize_address): Clean up.
8678 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8680 * config/arc/arc-arches.def: Option mrf16 valid for all
8682 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8683 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8684 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8685 * config/arc/arc-tables.opt: Regenerate.
8686 * config/arc/arc.c (arc_conditional_register_usage): Handle
8687 reduced register file case.
8688 (arc_file_start): Set must have build attributes.
8689 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8691 * config/arc/arc.opt (mrf16): Add new option.
8692 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8693 * config/arc/genmultilib.awk: Handle new mrf16 option.
8694 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8695 * config/arc/t-multilib: Regenerate.
8696 * doc/invoke.texi (ARC Options): Document mrf16 option.
8698 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8700 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8701 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8702 (arc_attribute_table): Add 'secure_call' attribute.
8703 (arc_print_operand): Print secure call operand.
8704 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8706 (arc_is_secure_call_p): New function. * config/arc/arc.md
8707 (call_i): Add support for sjli instruction.
8708 (call_value_i): Likewise.
8709 * config/arc/constraints.md (Csc): New constraint.
8711 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8712 John Eric Martin <John.Martin@emmicro-us.com>
8714 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8715 * config/arc/arc.c (_arc_jli_section): New struct.
8716 (arc_jli_section): New type.
8717 (rc_jli_sections): New static variable.
8718 (arc_handle_jli_attribute): New function.
8719 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8720 (arc_file_end): New function.
8721 (TARGET_ASM_FILE_END): Define.
8722 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8723 (arc_add_jli_section): New function.
8724 (jli_call_scan): Likewise.
8725 (arc_reorg): Call jli_call_scan.
8726 (arc_output_addsi): Remove 'S' from printing asm operand.
8727 (arc_is_jli_call_p): New function.
8728 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8730 (movhi_insn): Likewise.
8731 (movsi_insn): Likewise.
8732 (movsi_set_cc_insn): Likewise.
8733 (loadqi_update): Likewise.
8734 (load_zeroextendqisi_update): Likewise.
8735 (load_signextendqisi_update): Likewise.
8736 (loadhi_update): Likewise.
8737 (load_zeroextendhisi_update): Likewise.
8738 (load_signextendhisi_update): Likewise.
8739 (loadsi_update): Likewise.
8740 (loadsf_update): Likewise.
8741 (movsicc_insn): Likewise.
8742 (bset_insn): Likewise.
8743 (bxor_insn): Likewise.
8744 (bclr_insn): Likewise.
8745 (bmsk_insn): Likewise.
8746 (bicsi3_insn): Likewise.
8747 (cmpsi_cc_c_insn): Likewise.
8748 (movsi_ne): Likewise.
8749 (movsi_cond_exec): Likewise.
8750 (clrsbsi2): Likewise.
8761 (call_i): Remove 'S' asm letter, add jli instruction.
8762 (call_value_i): Likewise.
8763 * config/arc/arc.op (mjli-always): New option.
8764 * config/arc/constraints.md (Cji): New constraint.
8765 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8767 (subsf3_fpx): Likewise.
8768 (mulsf3_fpx): Likewise.
8769 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8771 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8773 * doc/invoke.texi (ARC): Document mjli-always option.
8775 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8777 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8778 avoid addition with 0 and use incw and decw where possible.
8780 2018-01-26 Richard Biener <rguenther@suse.de>
8782 PR tree-optimization/81082
8783 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8784 association if it requires casting to unsigned.
8785 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8786 from fold_plusminus_mult_expr to catch important cases late when
8787 range info is available.
8789 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8791 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8792 * configure.ac (hidden_linkonce): New test.
8793 * configure: Regenerate.
8794 * config.in: Regenerate.
8796 2018-01-26 Julia Koval <julia.koval@intel.com>
8798 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8799 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8800 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8801 _mm_mask_bitshuffle_epi64_mask): Fix type.
8802 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8803 USI_FTYPE_V4DI_V4DI_USI): Remove.
8804 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8805 __builtin_ia32_vpshufbitqmb256_mask,
8806 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8807 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8808 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8810 2018-01-26 Alan Modra <amodra@gmail.com>
8813 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8814 UNSPEC_VBPERMQ. Sort other unspecs.
8816 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8818 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8820 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8823 * predict.c (drop_profile): Do not push/pop cfun; update also
8825 (handle_missing_profiles): Fix logic looking for zero profiles.
8827 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8830 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8831 on functions with #pragma omp declare simd or functions with simd
8833 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8834 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8835 Remove trailing \n from warning_at calls.
8837 2018-01-25 Tom de Vries <tom@codesourcery.com>
8840 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8841 for neutered workers.
8843 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8846 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8847 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8849 2018-01-24 Jeff Law <law@redhat.com>
8852 * i386.c (get_probe_interval): Move to earlier point.
8853 (ix86_compute_frame_layout): If -fstack-clash-protection and
8854 the frame is larger than the probe interval, then use pushes
8855 to save registers rather than reg->mem moves.
8856 (ix86_expand_prologue): Remove conditional for int_registers_saved
8859 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8862 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8863 min/max for never referenced object.
8865 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8868 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8870 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8871 attributes from DECL_ATTRIBUTES (decl) without affecting
8872 DECL_ATTRIBUTES (current_function_decl).
8873 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8874 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8876 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8878 PR tree-optimization/83979
8879 * fold-const.c (fold_comparison): Use constant_boolean_node
8880 instead of boolean_{true,false}_node.
8882 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8884 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8887 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8889 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8890 Simplify the clause that sets the length attribute.
8891 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8892 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8893 clause that sets the length attribute.
8894 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8896 2018-01-24 Tom de Vries <tom@codesourcery.com>
8899 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8900 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8901 Add strict parameter.
8902 (prevent_branch_around_nothing): Insert dummy insn between branch to
8903 label and label with no ptx insn inbetween.
8904 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8906 2018-01-24 Tom de Vries <tom@codesourcery.com>
8909 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8910 for neutered threads in warp.
8911 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8913 2018-01-24 Richard Biener <rguenther@suse.de>
8915 PR tree-optimization/83176
8916 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8919 2018-01-24 Richard Biener <rguenther@suse.de>
8921 PR tree-optimization/82819
8922 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8923 code generating pluses that are no-ops in the target precision.
8925 2018-01-24 Richard Biener <rguenther@suse.de>
8928 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8930 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8932 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8933 to merge probabilities.
8934 * predict.c (probably_never_executed): Also mark as cold functions
8935 with global 0 profile and guessed local profile.
8936 * profile-count.c (profile_probability::combine_with_count): New
8938 * profile-count.h (profile_probability::operator*,
8939 profile_probability::operator*=, profile_probability::operator/,
8940 profile_probability::operator/=): Reduce precision to adjusted
8941 and set value to guessed on contradictory divisions.
8942 (profile_probability::combine_with_freq): Remove.
8943 (profile_probability::combine_wiht_count): Declare.
8944 (profile_count::force_nonzero):: Set to adjusted.
8945 (profile_count::probability_in):: Set quality to adjusted.
8946 * tree-ssa-tail-merge.c (replace_block_by): Use
8949 2018-01-23 Andrew Waterman <andrew@sifive.com>
8950 Jim Wilson <jimw@sifive.com>
8952 * config/riscv/riscv.c (riscv_stack_boundary): New.
8953 (riscv_option_override): Set riscv_stack_boundary. Handle
8954 riscv_preferred_stack_boundary_arg.
8955 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8956 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8957 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8958 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8959 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8960 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8962 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8965 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8966 of struct ix86_frame.
8967 (ix86_expand_epilogue): Likewise. Add a local variable for
8968 the reg_save_offset field in struct ix86_frame.
8970 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8972 PR tree-optimization/82604
8973 * tree-loop-distribution.c (enum partition_kind): New enum item
8974 PKIND_PARTIAL_MEMSET.
8975 (partition_builtin_p): Support above new enum item.
8976 (generate_code_for_partition): Ditto.
8977 (compute_access_range): Differentiate cases that equality can be
8978 proven at all loops, the innermost loops or no loops.
8979 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8980 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8981 (finalize_partitions, distribute_loop): Don't fuse partition of
8982 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8983 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8986 2018-01-23 Martin Liska <mliska@suse.cz>
8988 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8989 order to ignore the predictor.
8990 (PRED_POLYMORPHIC_CALL): Likewise.
8991 (PRED_RECURSIVE_CALL): Likewise.
8993 2018-01-23 Martin Liska <mliska@suse.cz>
8995 * tree-profile.c (tree_profiling): Print function header to
8996 aware reader which function we are working on.
8997 * value-prof.c (gimple_find_values_to_profile): Do not print
8998 not interesting value histograms.
9000 2018-01-23 Martin Liska <mliska@suse.cz>
9002 * profile-count.h (enum profile_quality): Add
9003 profile_uninitialized as the first value. Do not number values
9004 as they are zero based.
9005 (profile_count::verify): Update sanity check.
9006 (profile_probability::verify): Likewise.
9008 2018-01-23 Nathan Sidwell <nathan@acm.org>
9010 * doc/invoke.texi (ffor-scope): Deprecate.
9012 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9014 PR tree-optimization/83510
9015 * domwalk.c (set_all_edges_as_executable): New function.
9016 (dom_walker::dom_walker): Convert bool param
9017 "skip_unreachable_blocks" to enum reachability. Move setup of
9018 edge flags to set_all_edges_as_executable and only do it when
9019 reachability is REACHABLE_BLOCKS.
9020 * domwalk.h (enum dom_walker::reachability): New enum.
9021 (dom_walker::dom_walker): Convert bool param
9022 "skip_unreachable_blocks" to enum reachability.
9023 (set_all_edges_as_executable): New decl.
9024 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9025 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9027 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9028 but converting true to REACHABLE_BLOCKS.
9029 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9031 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9032 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9033 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9035 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9036 if check_all_array_refs will be called.
9038 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9040 * tree.c (selftest::test_location_wrappers): Add more test
9043 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9045 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9046 (selftest::test_bit_in_range): Likewise.
9048 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9051 * doc/sourcebuild.texi (vect_float): Say that the selector
9052 only describes the situation when -funsafe-math-optimizations is on.
9053 (vect_float_strict): Document.
9055 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9057 PR tree-optimization/83965
9058 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9059 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9060 instead of checking only for a reduction.
9061 (vect_recog_widen_sum_pattern): Likewise.
9063 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9065 * predict.c (probably_never_executed): Only use precise profile info.
9066 (compute_function_frequency): Skip after inlining hack since we now
9067 have quality checking.
9069 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9071 * profile-count.h (profile_probability::very_unlikely,
9072 profile_probability::unlikely, profile_probability::even): Set
9073 precision to guessed.
9075 2018-01-23 Richard Biener <rguenther@suse.de>
9077 PR tree-optimization/83963
9078 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9079 Properly terminate dominator walk when crossing the exit edge not
9080 when visiting its source block.
9082 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9085 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9086 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9088 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9090 PR tree-optimization/83957
9091 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9092 semicolon after for body surrounded by braces.
9094 PR tree-optimization/83081
9095 * profile-count.h (profile_probability::split): New method.
9096 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9097 Use profile_probability::split.
9098 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9099 when splitting a single conditional jump into 2.
9101 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9103 PR tree-optimization/69452
9104 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9107 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9109 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9110 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9111 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9113 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9115 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9116 * config/rl78/rl78.md (movdi): New define_expand.
9117 * config/rl78/rl78.c (rl78_split_movdi): New function.
9119 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9122 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9124 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9125 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9126 128-bit to produce an UNSPEC move to get the double word with the
9127 signbit and then a shift directly to do signbit.
9128 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9129 implementation with a new version that just does either a direct
9130 move or a regular move. Move memory interface to separate insns.
9131 Move insns so they are next to the expander.
9132 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9133 with signbit move. Split big and little endian case.
9134 (signbit<mode>2_dm_mem_le): Likewise.
9135 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9136 (signbit<mode>2_dm2): Likewise.
9138 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9140 * config/rl78/rl78.md (anddi3): New define_expand.
9142 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9144 * config/rl78/rl78.md (umindi3): New define_expand.
9146 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9148 * config/rl78/rl78.md (smindi3): New define_expand.
9150 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9152 * config/rl78/rl78.md (smaxdi3): New define_expand.
9154 2018-01-22 Carl Love <cel@us.ibm.com>
9156 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9157 LVX_V1TI): Add macro expansion.
9158 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9159 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9160 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9161 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9162 Change check to determine if the instruction is a byte reversing
9163 entry. Fix typo in comment.
9164 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9165 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9166 Add def_builtin calls for new builtins.
9167 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9168 Add define_insn expansion.
9170 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9172 * config/rl78/rl78.md (umaxdi3): New define_expand.
9174 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9176 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9177 for non-QImode registers.
9179 2018-01-22 Richard Biener <rguenther@suse.de>
9181 PR tree-optimization/83963
9182 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9183 including the loop exit block.
9184 (scop_detection::merge_sese): Likewise.
9185 (scop_detection::add_scop): Do it here instead.
9187 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9189 * doc/sourcebuild.texi (arm_softfloat): Document.
9191 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9194 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9195 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9196 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9198 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9199 David Edelsohn <dje.gcc@gmail.com>
9202 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9203 Change "crset eq" to "crset 2".
9204 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9205 (*call_indirect_aix<mode>_nospec): Likewise.
9206 (*call_value_indirect_aix<mode>_nospec): Likewise.
9207 (*call_indirect_elfv2<mode>_nospec): Likewise.
9208 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9209 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9210 change assembly output from . to $.
9211 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9212 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9213 (*tablejump<mode>_internal1_nospec): Likewise.
9215 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9218 * config/sh/sh_optimize_sett_clrt.cc:
9219 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9221 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9223 PR tree-optimization/83940
9224 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9225 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9226 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9227 use it to pass back the definition type.
9228 (vect_check_store_rhs): Likewise rhs_dt_out.
9229 (vect_build_gather_load_calls): Add a mask_dt argument and use
9230 it instead of a call to vect_is_simple_use.
9231 (vectorizable_store): Update calls to vect_check_load_store_mask
9232 and vect_check_store_rhs. Use the dt returned by the latter instead
9233 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9234 instead of calls to vect_is_simple_use. Pass the scalar rather
9235 than the vector operand to vect_is_simple_use when handling
9236 second and subsequent copies of an rhs value.
9237 (vectorizable_load): Update calls to vect_check_load_store_mask
9238 and vect_build_gather_load_calls. Use the cached mask_dt and
9239 gs_info.offset_dt instead of calls to vect_is_simple_use.
9241 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9244 * tree-emutls.c: Include gimplify.h.
9245 (lower_emutls_2): New function.
9246 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9247 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9248 it before further processing.
9251 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9252 UINTVAL (trueop1) instead of INTVAL (op1).
9254 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9258 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9259 INCOMING_FRAME_SP_OFFSET if not defined.
9260 (scan_trace): Add ENTRY argument. If true and
9261 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9262 emit a note to adjust the CFA offset.
9263 (create_cfi_notes): Adjust scan_trace callers.
9264 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9265 INCOMING_FRAME_SP_OFFSET in the CIE.
9266 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9267 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9269 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9270 * doc/tm.texi: Regenerated.
9272 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9274 PR rtl-optimization/83147
9275 * lra-constraints.c (remove_inheritance_pseudos): Use
9276 lra_substitute_pseudo_within_insn.
9278 2018-01-19 Tom de Vries <tom@codesourcery.com>
9279 Cesar Philippidis <cesar@codesourcery.com>
9282 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9284 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9287 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9288 spaces for function labels.
9290 2018-01-19 Martin Liska <mliska@suse.cz>
9292 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9293 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9294 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9295 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9296 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9297 (PRED_CONST_RETURN): Change from 69 to 65.
9298 (PRED_NULL_RETURN): Change from 91 to 71.
9299 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9300 (PRED_LOOP_GUARD): Change from 66 to 73.
9302 2018-01-19 Martin Liska <mliska@suse.cz>
9304 * predict.c (predict_insn_def): Add new assert.
9305 (struct branch_predictor): Change type to signed integer.
9306 (test_prediction_value_range): Amend test to cover
9308 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9309 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9310 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9311 (PRED_LOOP_IV_COMPARE): Likewise.
9312 * predict.h (PROB_UNINITIALIZED): Define new constant.
9314 2018-01-19 Martin Liska <mliska@suse.cz>
9316 * predict.c (dump_prediction): Add new format for
9317 analyze_brprob.py script which is enabled with -details
9319 * profile-count.h (precise_p): New function.
9321 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9323 PR tree-optimization/83922
9324 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9325 there are no statements that need masking.
9326 (vect_active_double_reduction_p): New function.
9327 (vect_analyze_loop_operations): Use it when handling phis that
9328 are not in the loop header.
9330 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9332 PR tree-optimization/83914
9333 * tree-vect-loop.c (vectorizable_induction): Don't convert
9334 init_expr or apply the peeling adjustment for inductions
9335 that are nested within the vectorized loop.
9337 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9339 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9342 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9346 * function.h (gimplify_parameters): Add gimple_seq * argument.
9347 * function.c: Include gimple.h and options.h.
9348 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9349 for the added local temporaries if needed.
9350 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9351 if there are any parameter cleanups, wrap whole body into a
9352 try/finally with the cleanups.
9354 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9357 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9358 Use GET_MODE_CLASS for scalar floating point.
9360 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9364 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9365 Fix call of call_cgraph_insertion_hooks.
9367 2018-01-18 Martin Sebor <msebor@redhat.com>
9369 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9371 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9374 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9377 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9380 * common.opt: (-ffile-prefix-map): New option.
9381 * opts.c (common_handle_option): Defer it.
9382 * opts-global.c (handle_common_deferred_options): Handle it.
9383 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9384 * file-prefix-map.h: New file.
9385 (remap_debug_filename, add_debug_prefix_map): ...here.
9386 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9387 * final.c (debug_prefix_map, add_debug_prefix_map
9388 remap_debug_filename): Move to...
9389 * file-prefix-map.c: New file.
9390 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9391 generalize, get rid of alloca(), use strrchr() instead of strchr().
9392 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9393 Implement in terms of add_prefix_map().
9394 (remap_macro_filename, remap_debug_filename): Implement in term of
9396 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9397 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9398 * dbxout.c: Include file-prefix-map.h.
9399 * varasm.c: Likewise.
9400 * vmsdbgout.c: Likewise.
9401 * xcoffout.c: Likewise.
9402 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9403 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9404 * doc/invoke.texi (-ffile-prefix-map): Document.
9405 (-fdebug-prefix-map): Update description.
9407 2018-01-18 Martin Liska <mliska@suse.cz>
9409 * config/i386/i386.c (indirect_thunk_name): Document that also
9411 (output_indirect_thunk): Document why both instructions
9412 (pause and lfence) are generated.
9414 2018-01-18 Richard Biener <rguenther@suse.de>
9416 PR tree-optimization/83887
9417 * graphite-scop-detection.c
9418 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9419 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9420 (scop_detection::merge_sese): Re-implement with a flood-fill
9421 algorithm that properly finds a SESE region if it exists.
9423 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9426 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9427 pointer_diff optimizations use view_convert instead of convert.
9429 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9431 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9432 Generate different code for -mno-speculate-indirect-jumps.
9433 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9434 (*call_indirect_aix<mode>): Disable for
9435 -mno-speculate-indirect-jumps.
9436 (*call_indirect_aix<mode>_nospec): New define_insn.
9437 (*call_value_indirect_aix<mode>): Disable for
9438 -mno-speculate-indirect-jumps.
9439 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9440 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9441 -mno-speculate-indirect-jumps.
9442 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9444 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9446 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9447 long double type, set the flags for noting the default long double
9448 type, even if we don't pass or return a long double type.
9450 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9453 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9456 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9458 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9459 support for merge[hl].
9460 (fold_mergehl_helper): New helper function.
9461 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9462 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9463 (altivec_vmrglw_direct): Add xxmrglw insn.
9465 2018-01-17 Andrew Waterman <andrew@sifive.com>
9467 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9468 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9470 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9473 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9474 call the lto_location_cache before reading the
9475 DECL_SOURCE_LOCATION of the types.
9477 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9478 Richard Sandiford <richard.sandiford@linaro.org>
9480 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9481 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9482 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9483 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9484 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9486 * config/aarch64/constraints.md (aarch64_movti_operand):
9488 * config/aarch64/predicates.md (Uti): Add new constraint.
9490 2018-01-17 Carl Love <cel@us.ibm.com>
9492 * config/rs6000/vsx.md (define_expand xl_len_r,
9493 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9494 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9496 (define_expand, define_insn): Move the shift left from the
9497 define_insn to the define_expand for lxvl and stxvl instructions.
9498 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9499 and XL_LEN_R definitions to PURE.
9501 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9503 * config/i386/i386.c (indirect_thunk_name): Declare regno
9504 as unsigned int. Compare regno with INVALID_REGNUM.
9505 (output_indirect_thunk): Ditto.
9506 (output_indirect_thunk_function): Ditto.
9507 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9508 in the call to output_indirect_thunk_function.
9510 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9513 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9514 rather than the size of inner_type to determine the stack slot size
9515 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9517 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9520 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9523 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9525 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9526 endian Linux systems to optionally enable multilibs for selecting
9527 the long double type if the user configured an explicit type.
9528 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9529 have no long double multilibs if not defined.
9530 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9531 warn if the user used -mabi={ieee,ibm}longdouble and we built
9532 multilibs for long double.
9533 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9534 appropriate multilib option.
9535 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9537 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9538 for building long double multilibs.
9539 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9541 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9543 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9546 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9548 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9551 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9554 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9557 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9559 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9560 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9562 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9564 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9565 different rtl trees depending on TARGET_64BIT.
9566 (rs6000_gen_lvx): Likewise.
9568 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9570 * config/visium/visium.md (nop): Tweak comment.
9571 (hazard_nop): Likewise.
9573 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9575 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9576 -mspeculate-indirect-jumps.
9577 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9578 for -mno-speculate-indirect-jumps.
9579 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9580 (*call_value_indirect_elfv2<mode>): Disable for
9581 -mno-speculate-indirect-jumps.
9582 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9583 (indirect_jump): Emit different RTL for
9584 -mno-speculate-indirect-jumps.
9585 (*indirect_jump<mode>): Disable for
9586 -mno-speculate-indirect-jumps.
9587 (*indirect_jump<mode>_nospec): New define_insn.
9588 (tablejump): Emit different RTL for
9589 -mno-speculate-indirect-jumps.
9590 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9591 (tablejumpsi_nospec): New define_expand.
9592 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9593 (tablejumpdi_nospec): New define_expand.
9594 (*tablejump<mode>_internal1): Disable for
9595 -mno-speculate-indirect-jumps.
9596 (*tablejump<mode>_internal1_nospec): New define_insn.
9597 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9600 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9602 * caller-save.c (insert_save): Drop unnecessary parameter. All
9605 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9606 Richard Biener <rguenth@suse.de>
9609 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9610 return early, inline manually is_gimple_sizepos. Make sure if we
9611 call gimplify_expr we don't end up with a gimple constant.
9612 * tree.c (variably_modified_type_p): Don't return true for
9613 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9614 * gimplify.h (is_gimple_sizepos): Remove.
9616 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9618 PR tree-optimization/83857
9619 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9620 vectorizable_live_operation for pure SLP statements.
9621 (vectorizable_live_operation): Handle PHIs.
9623 2018-01-16 Richard Biener <rguenther@suse.de>
9625 PR tree-optimization/83867
9626 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9627 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9629 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9632 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9633 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9634 If off is not INTEGER_CST, issue a may not be aligned warning
9635 rather than isn't aligned. Use isn%'t rather than isn't.
9636 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9638 <case MULT_EXPR>: Improve the case when bottom and one of the
9639 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9640 operand, in that case check if the other operand is multiple of
9641 bottom divided by the INTEGER_CST operand.
9643 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9646 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9647 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9648 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9649 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9650 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9651 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9652 (pa_function_arg_size): New function.
9654 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9656 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9657 in a separate statement.
9659 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9661 PR tree-optimization/83847
9662 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9663 group gathers and scatters.
9665 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9667 PR rtl-optimization/86620
9668 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9670 PR rtl-optimization/83213
9671 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9672 to last if both are JUMP_INSNs.
9674 PR tree-optimization/83843
9675 * gimple-ssa-store-merging.c
9676 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9677 store_immediate_info for bswap/nop orig_stores.
9679 2018-01-15 Andrew Waterman <andrew@sifive.com>
9681 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9683 <UDIV>: Increase cost if !TARGET_DIV.
9685 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9687 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9688 (define_attr "cr_logical_3op"): New.
9689 (cceq_ior_compare): Adjust.
9690 (cceq_ior_compare_complement): Adjust.
9691 (*cceq_rev_compare): Adjust.
9692 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9693 (is_cracked_insn): Adjust.
9694 (insn_must_be_first_in_group): Adjust.
9695 * config/rs6000/40x.md: Adjust.
9696 * config/rs6000/440.md: Adjust.
9697 * config/rs6000/476.md: Adjust.
9698 * config/rs6000/601.md: Adjust.
9699 * config/rs6000/603.md: Adjust.
9700 * config/rs6000/6xx.md: Adjust.
9701 * config/rs6000/7450.md: Adjust.
9702 * config/rs6000/7xx.md: Adjust.
9703 * config/rs6000/8540.md: Adjust.
9704 * config/rs6000/cell.md: Adjust.
9705 * config/rs6000/e300c2c3.md: Adjust.
9706 * config/rs6000/e500mc.md: Adjust.
9707 * config/rs6000/e500mc64.md: Adjust.
9708 * config/rs6000/e5500.md: Adjust.
9709 * config/rs6000/e6500.md: Adjust.
9710 * config/rs6000/mpc.md: Adjust.
9711 * config/rs6000/power4.md: Adjust.
9712 * config/rs6000/power5.md: Adjust.
9713 * config/rs6000/power6.md: Adjust.
9714 * config/rs6000/power7.md: Adjust.
9715 * config/rs6000/power8.md: Adjust.
9716 * config/rs6000/power9.md: Adjust.
9717 * config/rs6000/rs64.md: Adjust.
9718 * config/rs6000/titan.md: Adjust.
9720 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9722 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9723 ix86_indirect_branch_register logic.
9725 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9727 * config/i386/constraints.md (Bs): Update
9728 ix86_indirect_branch_register check. Don't check
9729 ix86_indirect_branch_register with GOT_memory_operand.
9731 * config/i386/predicates.md (GOT_memory_operand): Don't check
9732 ix86_indirect_branch_register here.
9733 (GOT32_symbol_operand): Likewise.
9735 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9737 * config/i386/predicates.md (constant_call_address_operand):
9738 Rewrite ix86_indirect_branch_register logic.
9739 (sibcall_insn_operand): Likewise.
9741 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9743 * config/i386/constraints.md (Bs): Replace
9744 ix86_indirect_branch_thunk_register with
9745 ix86_indirect_branch_register.
9747 * config/i386/i386.md (indirect_jump): Likewise.
9748 (tablejump): Likewise.
9749 (*sibcall_memory): Likewise.
9750 (*sibcall_value_memory): Likewise.
9751 Peepholes of indirect call and jump via memory: Likewise.
9752 * config/i386/i386.opt: Likewise.
9753 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9754 (GOT_memory_operand): Likewise.
9755 (call_insn_operand): Likewise.
9756 (sibcall_insn_operand): Likewise.
9757 (GOT32_symbol_operand): Likewise.
9759 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9762 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9763 type rather than type addr's type points to.
9764 (expand_omp_atomic_mutex): Likewise.
9765 (expand_omp_atomic): Likewise.
9767 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9770 * config/i386/i386.c (output_indirect_thunk_function): Use
9771 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9772 for __x86_return_thunk.
9774 2018-01-15 Richard Biener <rguenther@suse.de>
9777 * expmed.c (extract_bit_field_1): Fix typo.
9779 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9782 * config/arm/iterators.md (VF): New mode iterator.
9783 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9784 Remove integer-related logic from pattern.
9785 (neon_vabd<mode>_3): Likewise.
9787 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9790 * common.opt (fstrict-overflow): No longer an alias.
9791 (fwrapv-pointer): New option.
9792 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9793 also for pointer types based on flag_wrapv_pointer.
9794 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9795 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9796 opts->x_flag_wrapv got set.
9797 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9798 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9799 POINTER_TYPE_OVERFLOW_UNDEFINED.
9800 * match.pd: Likewise in address comparison pattern.
9801 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9803 2018-01-15 Richard Biener <rguenther@suse.de>
9806 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9807 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9808 Reset type names to their identifier if their TYPE_DECL doesn't
9809 have linkage (and thus is used for ODR and devirt).
9810 (save_debug_info_for_decl): Remove.
9811 (save_debug_info_for_type): Likewise.
9812 (add_tree_to_fld_list): Adjust.
9813 * tree-pretty-print.c (dump_generic_node): Make dumping of
9814 type names more robust.
9816 2018-01-15 Richard Biener <rguenther@suse.de>
9818 * BASE-VER: Bump to 8.0.1.
9820 2018-01-14 Martin Sebor <msebor@redhat.com>
9823 * builtins.c (check_access): Avoid warning when the no-warning bit
9826 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9828 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9829 * ira-color (allocno_hard_regs_compare): Likewise.
9831 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9834 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9835 Use .pushsection/.popsection.
9837 2018-01-14 Martin Sebor <msebor@redhat.com>
9840 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9842 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9844 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9845 entry from extra_headers.
9846 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9847 extra_headers, make the list bitwise identical to the i?86-*-* one.
9849 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9851 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9852 -mcmodel=large with -mindirect-branch=thunk,
9853 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9854 -mfunction-return=thunk-extern.
9855 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9856 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9857 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9859 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9861 * config/i386/i386.c (print_reg): Print the name of the full
9862 integer register without '%'.
9863 (ix86_print_operand): Handle 'V'.
9864 * doc/extend.texi: Document 'V' modifier.
9866 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9868 * config/i386/constraints.md (Bs): Disallow memory operand for
9869 -mindirect-branch-register.
9871 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9872 (GOT_memory_operand): Likewise.
9873 (call_insn_operand): Likewise.
9874 (sibcall_insn_operand): Likewise.
9875 (GOT32_symbol_operand): Likewise.
9876 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9877 for -mindirect-branch-register.
9878 (tablejump): Likewise.
9879 (*sibcall_memory): Likewise.
9880 (*sibcall_value_memory): Likewise.
9881 Disallow peepholes of indirect call and jump via memory for
9882 -mindirect-branch-register.
9883 (*call_pop): Replace m with Bw.
9884 (*call_value_pop): Likewise.
9885 (*sibcall_pop_memory): Replace m with Bs.
9886 * config/i386/i386.opt (mindirect-branch-register): New option.
9887 * doc/invoke.texi: Document -mindirect-branch-register option.
9889 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9891 * config/i386/i386-protos.h (ix86_output_function_return): New.
9892 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9893 set function_return_type.
9894 (indirect_thunk_name): Add ret_p to indicate thunk for function
9896 (output_indirect_thunk_function): Pass false to
9897 indirect_thunk_name.
9898 (ix86_output_indirect_branch_via_reg): Likewise.
9899 (ix86_output_indirect_branch_via_push): Likewise.
9900 (output_indirect_thunk_function): Create alias for function
9901 return thunk if regno < 0.
9902 (ix86_output_function_return): New function.
9903 (ix86_handle_fndecl_attribute): Handle function_return.
9904 (ix86_attribute_table): Add function_return.
9905 * config/i386/i386.h (machine_function): Add
9906 function_return_type.
9907 * config/i386/i386.md (simple_return_internal): Use
9908 ix86_output_function_return.
9909 (simple_return_internal_long): Likewise.
9910 * config/i386/i386.opt (mfunction-return=): New option.
9911 (indirect_branch): Mention -mfunction-return=.
9912 * doc/extend.texi: Document function_return function attribute.
9913 * doc/invoke.texi: Document -mfunction-return= option.
9915 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9917 * config/i386/i386-opts.h (indirect_branch): New.
9918 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9919 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9920 with local indirect jump when converting indirect call and jump.
9921 (ix86_set_indirect_branch_type): New.
9922 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9923 (indirectlabelno): New.
9924 (indirect_thunk_needed): Likewise.
9925 (indirect_thunk_bnd_needed): Likewise.
9926 (indirect_thunks_used): Likewise.
9927 (indirect_thunks_bnd_used): Likewise.
9928 (INDIRECT_LABEL): Likewise.
9929 (indirect_thunk_name): Likewise.
9930 (output_indirect_thunk): Likewise.
9931 (output_indirect_thunk_function): Likewise.
9932 (ix86_output_indirect_branch_via_reg): Likewise.
9933 (ix86_output_indirect_branch_via_push): Likewise.
9934 (ix86_output_indirect_branch): Likewise.
9935 (ix86_output_indirect_jmp): Likewise.
9936 (ix86_code_end): Call output_indirect_thunk_function if needed.
9937 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9939 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9940 (ix86_attribute_table): Add indirect_branch.
9941 * config/i386/i386.h (machine_function): Add indirect_branch_type
9942 and has_local_indirect_jump.
9943 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9945 (tablejump): Likewise.
9946 (*indirect_jump): Use ix86_output_indirect_jmp.
9947 (*tablejump_1): Likewise.
9948 (simple_return_indirect_internal): Likewise.
9949 * config/i386/i386.opt (mindirect-branch=): New option.
9950 (indirect_branch): New.
9953 (thunk-inline): Likewise.
9954 (thunk-extern): Likewise.
9955 * doc/extend.texi: Document indirect_branch function attribute.
9956 * doc/invoke.texi: Document -mindirect-branch= option.
9958 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9961 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9963 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9965 * ipa-inline.c (want_inline_small_function_p): Return false if
9966 inlining has already failed with CIF_FINAL_ERROR.
9967 (update_caller_keys): Call want_inline_small_function_p before
9969 (update_callee_keys): Likewise.
9971 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9973 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9975 (rs6000_quadword_masked_address_p): Likewise.
9976 (quad_aligned_load_p): Likewise.
9977 (quad_aligned_store_p): Likewise.
9978 (const_load_sequence_p): Add comment to describe the outer-most loop.
9979 (mimic_memory_attributes_and_flags): New function.
9980 (rs6000_gen_stvx): Likewise.
9981 (replace_swapped_aligned_store): Likewise.
9982 (rs6000_gen_lvx): Likewise.
9983 (replace_swapped_aligned_load): Likewise.
9984 (replace_swapped_load_constant): Capitalize argument name in
9985 comment describing this function.
9986 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9987 and stores that access quad-word aligned addresses and replace
9988 with stvx or lvx instructions when appropriate.
9989 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9990 New function prototype.
9991 (rs6000_quadword_masked_address_p): Likewise.
9992 (rs6000_gen_lvx): Likewise.
9993 (rs6000_gen_stvx): Likewise.
9994 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9995 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9996 when memory address is aligned.
9997 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9998 this split to select lvx instruction when memory address is aligned.
9999 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10000 instruction when memory address is aligned.
10001 (*vsx_le_perm_load_v16qi): Likewise.
10002 (four unnamed splitters): Modify to select the stvx instruction
10003 when memory is aligned.
10005 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10007 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10008 which appears in the queue multiple times.
10010 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10011 Alan Hayward <alan.hayward@arm.com>
10012 David Sherwood <david.sherwood@arm.com>
10014 * tree-vectorizer.h (vec_lower_bound): New structure.
10015 (_loop_vec_info): Add check_nonzero and lower_bounds.
10016 (LOOP_VINFO_CHECK_NONZERO): New macro.
10017 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10018 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10019 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10020 fields. Make seg_len the distance travelled, not including the
10022 (dr_direction_indicator): Declare.
10023 (dr_zero_step_indicator): Likewise.
10024 (dr_known_forward_stride_p): Likewise.
10025 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10027 (runtime_alias_check_p): Allow runtime alias checks with
10029 (operator ==): Compare access_size and align.
10030 (prune_runtime_alias_test_list): Rework for new distinction between
10031 the access_size and seg_len.
10032 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10034 (get_segment_min_max): New function.
10035 (create_intersect_range_checks): Use it.
10036 (dr_step_indicator): New function.
10037 (dr_direction_indicator): Likewise.
10038 (dr_zero_step_indicator): Likewise.
10039 (dr_known_forward_stride_p): Likewise.
10040 * tree-loop-distribution.c (data_ref_segment_size): Return
10041 DR_STEP * (niters - 1).
10042 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10044 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10045 (vect_preserves_scalar_order_p): New function, split out from...
10046 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10047 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10048 (vect_vfa_access_size): New function.
10049 (vect_vfa_align): Likewise.
10050 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10051 (dump_lower_bound): New function.
10052 (vect_check_lower_bound): Likewise.
10053 (vect_small_gap_p): Likewise.
10054 (vectorizable_with_step_bound_p): Likewise.
10055 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10056 depencies if the vectorization factor is 1. Convert the checks
10057 for nonzero steps into checks on the bounds of DR_STEP. Try using
10058 a bunds check for variable steps if the minimum required step is
10059 relatively small. Update calls to the dr_with_seg_len
10060 constructor and to vect_compile_time_alias.
10061 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10063 (vect_loop_versioning): Call it.
10064 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10066 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10068 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10069 Alan Hayward <alan.hayward@arm.com>
10070 David Sherwood <david.sherwood@arm.com>
10072 * doc/sourcebuild.texi (vect_scatter_store): Document.
10073 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10075 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10077 * genopinit.c (main): Add supports_vec_scatter_store and
10078 supports_vec_scatter_store_cached to target_optabs.
10079 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10080 IFN_MASK_SCATTER_STORE.
10081 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10083 * internal-fn.h (internal_store_fn_p): Declare.
10084 (internal_fn_stored_value_index): Likewise.
10085 * internal-fn.c (scatter_store_direct): New macro.
10086 (expand_scatter_store_optab_fn): New function.
10087 (direct_scatter_store_optab_supported_p): New macro.
10088 (internal_store_fn_p): New function.
10089 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10090 IFN_MASK_SCATTER_STORE.
10091 (internal_fn_mask_index): Likewise.
10092 (internal_fn_stored_value_index): New function.
10093 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10094 for scatter stores.
10095 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10096 * optabs-query.c (supports_vec_scatter_store_p): New function.
10097 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10098 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10099 true for scatter stores.
10100 (vect_gather_scatter_fn_p): Handle scatter stores too.
10101 (vect_check_gather_scatter): Consider using scatter stores if
10102 supports_vec_scatter_store_p.
10103 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10104 scatter stores too.
10105 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10106 internal_fn_stored_value_index.
10107 (check_load_store_masking): Handle scatter stores too.
10108 (vect_get_store_rhs): Make public.
10109 (vectorizable_call): Use internal_store_fn_p.
10110 (vectorizable_store): Handle scatter store internal functions.
10111 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10112 when deciding whether the end of the group has been reached.
10113 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10114 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10115 (mask_scatter_store<mode>): New insns.
10117 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10118 Alan Hayward <alan.hayward@arm.com>
10119 David Sherwood <david.sherwood@arm.com>
10121 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10122 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10123 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10125 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10126 Use vect_truncate_gather_scatter_offset if we can't treat the
10127 operation as a normal gather load or scatter store.
10128 (get_group_load_store_type): Take the gather_scatter_info
10129 as argument. Try using a gather load or scatter store for
10130 single-element groups.
10131 (get_load_store_type): Update calls to get_group_load_store_type
10132 and vect_use_strided_gather_scatters_p.
10134 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10135 Alan Hayward <alan.hayward@arm.com>
10136 David Sherwood <david.sherwood@arm.com>
10138 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10139 optional tree argument.
10140 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10142 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10143 but continue to use the current value as a fallback.
10144 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10145 to compare the updates.
10146 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10147 (get_load_store_type): Use it when handling a strided access.
10148 (vect_get_strided_load_store_ops): New function.
10149 (vect_get_data_ptr_increment): Likewise.
10150 (vectorizable_load): Handle strided gather loads. Always pass
10151 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10153 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10154 Alan Hayward <alan.hayward@arm.com>
10155 David Sherwood <david.sherwood@arm.com>
10157 * doc/md.texi (gather_load@var{m}): Document.
10158 (mask_gather_load@var{m}): Likewise.
10159 * genopinit.c (main): Add supports_vec_gather_load and
10160 supports_vec_gather_load_cached to target_optabs.
10161 * optabs-tree.c (init_tree_optimization_optabs): Use
10162 ggc_cleared_alloc to allocate target_optabs.
10163 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10164 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10166 * internal-fn.h (internal_load_fn_p): Declare.
10167 (internal_gather_scatter_fn_p): Likewise.
10168 (internal_fn_mask_index): Likewise.
10169 (internal_gather_scatter_fn_supported_p): Likewise.
10170 * internal-fn.c (gather_load_direct): New macro.
10171 (expand_gather_load_optab_fn): New function.
10172 (direct_gather_load_optab_supported_p): New macro.
10173 (direct_internal_fn_optab): New function.
10174 (internal_load_fn_p): Likewise.
10175 (internal_gather_scatter_fn_p): Likewise.
10176 (internal_fn_mask_index): Likewise.
10177 (internal_gather_scatter_fn_supported_p): Likewise.
10178 * optabs-query.c (supports_at_least_one_mode_p): New function.
10179 (supports_vec_gather_load_p): Likewise.
10180 * optabs-query.h (supports_vec_gather_load_p): Declare.
10181 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10182 and memory_type field.
10183 (NUM_PATTERNS): Bump to 15.
10184 * tree-vect-data-refs.c: Include internal-fn.h.
10185 (vect_gather_scatter_fn_p): New function.
10186 (vect_describe_gather_scatter_call): Likewise.
10187 (vect_check_gather_scatter): Try using internal functions for
10188 gather loads. Recognize existing calls to a gather load function.
10189 (vect_analyze_data_refs): Consider using gather loads if
10190 supports_vec_gather_load_p.
10191 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10192 (vect_get_gather_scatter_offset_type): Likewise.
10193 (vect_convert_mask_for_vectype): Likewise.
10194 (vect_add_conversion_to_patterm): Likewise.
10195 (vect_try_gather_scatter_pattern): Likewise.
10196 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10197 (vect_vect_recog_func_ptrs): Add it.
10198 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10199 internal_fn_mask_index and internal_gather_scatter_fn_p.
10200 (check_load_store_masking): Take the gather_scatter_info as an
10201 argument and handle gather loads.
10202 (vect_get_gather_scatter_ops): New function.
10203 (vectorizable_call): Check internal_load_fn_p.
10204 (vectorizable_load): Likewise. Handle gather load internal
10206 (vectorizable_store): Update call to check_load_store_masking.
10207 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10208 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10209 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10210 (aarch64_gather_scale_operand_d): New predicates.
10211 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10212 (mask_gather_load<mode>): New insns.
10214 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10215 Alan Hayward <alan.hayward@arm.com>
10216 David Sherwood <david.sherwood@arm.com>
10218 * optabs.def (fold_left_plus_optab): New optab.
10219 * doc/md.texi (fold_left_plus_@var{m}): Document.
10220 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10221 * internal-fn.c (fold_left_direct): Define.
10222 (expand_fold_left_optab_fn): Likewise.
10223 (direct_fold_left_optab_supported_p): Likewise.
10224 * fold-const-call.c (fold_const_fold_left): New function.
10225 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10226 * tree-parloops.c (valid_reduction_p): New function.
10227 (gather_scalar_reductions): Use it.
10228 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10229 (vect_finish_replace_stmt): Declare.
10230 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10231 (needs_fold_left_reduction_p): New function, split out from...
10232 (vect_is_simple_reduction): ...here. Accept reductions that
10233 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10234 (vect_force_simple_reduction): Also store the reduction type in
10235 the assignment's STMT_VINFO_REDUC_TYPE.
10236 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10237 (merge_with_identity): New function.
10238 (vect_expand_fold_left): Likewise.
10239 (vectorize_fold_left_reduction): Likewise.
10240 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10241 scalar phi in place for it. Check for target support and reject
10242 cases that would reassociate the operation. Defer the transform
10243 phase to vectorize_fold_left_reduction.
10244 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10245 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10246 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10248 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10250 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10251 call to ifc_temp_var.
10253 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10254 Alan Hayward <alan.hayward@arm.com>
10255 David Sherwood <david.sherwood@arm.com>
10257 * target.def (legitimize_address_displacement): Take the original
10258 offset as a poly_int.
10259 * targhooks.h (default_legitimize_address_displacement): Update
10261 * targhooks.c (default_legitimize_address_displacement): Likewise.
10262 * doc/tm.texi: Regenerate.
10263 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10264 as an argument, moving assert of ad->disp == ad->disp_term to...
10265 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10266 Try calling targetm.legitimize_address_displacement before expanding
10267 the address rather than afterwards, and adjust for the new interface.
10268 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10269 Match the new hook interface. Handle SVE addresses.
10270 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10271 new hook interface.
10273 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10275 * Makefile.in (OBJS): Add early-remat.o.
10276 * target.def (select_early_remat_modes): New hook.
10277 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10278 * doc/tm.texi: Regenerate.
10279 * targhooks.h (default_select_early_remat_modes): Declare.
10280 * targhooks.c (default_select_early_remat_modes): New function.
10281 * timevar.def (TV_EARLY_REMAT): New timevar.
10282 * passes.def (pass_early_remat): New pass.
10283 * tree-pass.h (make_pass_early_remat): Declare.
10284 * early-remat.c: New file.
10285 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10287 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10289 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10290 Alan Hayward <alan.hayward@arm.com>
10291 David Sherwood <david.sherwood@arm.com>
10293 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10294 vfm1 with a bound_epilog parameter.
10295 (vect_do_peeling): Update calls accordingly, and move the prologue
10296 call earlier in the function. Treat the base bound_epilog as 0 for
10297 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10298 this base when peeling for gaps.
10299 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10300 with fully-masked loops.
10301 (vect_estimate_min_profitable_iters): Handle the single peeled
10302 iteration in that case.
10304 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10305 Alan Hayward <alan.hayward@arm.com>
10306 David Sherwood <david.sherwood@arm.com>
10308 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10309 single-element interleaving even if the size is not a power of 2.
10310 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10311 accesses for single-element interleaving if the group size is
10314 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10315 Alan Hayward <alan.hayward@arm.com>
10316 David Sherwood <david.sherwood@arm.com>
10318 * doc/md.texi (fold_extract_last_@var{m}): Document.
10319 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10320 * optabs.def (fold_extract_last_optab): New optab.
10321 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10322 * internal-fn.c (fold_extract_direct): New macro.
10323 (expand_fold_extract_optab_fn): Likewise.
10324 (direct_fold_extract_optab_supported_p): Likewise.
10325 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10326 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10327 EXTRACT_LAST_REDUCTION.
10328 (get_initial_def_for_reduction): Do not create an initial vector
10329 for EXTRACT_LAST_REDUCTION reductions.
10330 (vectorizable_reduction): Leave the scalar phi in place for
10331 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10332 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10333 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10334 transform phase to vectorizable_condition.
10335 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10337 (vect_finish_stmt_generation): ...here.
10338 (vect_finish_replace_stmt): New function.
10339 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10340 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10342 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10344 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10345 Alan Hayward <alan.hayward@arm.com>
10346 David Sherwood <david.sherwood@arm.com>
10348 * doc/md.texi (extract_last_@var{m}): Document.
10349 * optabs.def (extract_last_optab): New optab.
10350 * internal-fn.def (EXTRACT_LAST): New internal function.
10351 * internal-fn.c (cond_unary_direct): New macro.
10352 (expand_cond_unary_optab_fn): Likewise.
10353 (direct_cond_unary_optab_supported_p): Likewise.
10354 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10355 loops using EXTRACT_LAST.
10356 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10357 (extract_last_<mode>): ...this optab.
10358 (vec_extract<mode><Vel>): Update accordingly.
10360 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10361 Alan Hayward <alan.hayward@arm.com>
10362 David Sherwood <david.sherwood@arm.com>
10364 * target.def (empty_mask_is_expensive): New hook.
10365 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10366 * doc/tm.texi: Regenerate.
10367 * targhooks.h (default_empty_mask_is_expensive): Declare.
10368 * targhooks.c (default_empty_mask_is_expensive): New function.
10369 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10370 if the target says that empty masks are expensive.
10371 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10373 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10375 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10376 Alan Hayward <alan.hayward@arm.com>
10377 David Sherwood <david.sherwood@arm.com>
10379 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10380 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10381 (vect_use_loop_mask_for_alignment_p): New function.
10382 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10383 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10384 niters_skip argument. Make sure that the first niters_skip elements
10385 of the first iteration are inactive.
10386 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10387 Update call to vect_set_loop_masks_directly.
10388 (get_misalign_in_elems): New function, split out from...
10389 (vect_gen_prolog_loop_niters): ...here.
10390 (vect_update_init_of_dr): Take a code argument that specifies whether
10391 the adjustment should be added or subtracted.
10392 (vect_update_init_of_drs): Likewise.
10393 (vect_prepare_for_masked_peels): New function.
10394 (vect_do_peeling): Skip prologue peeling if we're using a mask
10395 instead. Update call to vect_update_inits_of_drs.
10396 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10398 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10399 alignment. Do not include the number of peeled iterations in
10400 the minimum threshold in that case.
10401 (vectorizable_induction): Adjust the start value down by
10402 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10403 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10404 Take the number of skipped iterations into account when calculating
10406 * tree-vect-stmts.c (vect_gen_while_not): New function.
10408 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10409 Alan Hayward <alan.hayward@arm.com>
10410 David Sherwood <david.sherwood@arm.com>
10412 * doc/sourcebuild.texi (vect_fully_masked): Document.
10413 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10414 default value to 0.
10415 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10417 (vect_analyze_loop_2): ...here. Don't check the vectorization
10418 factor against the number of loop iterations if the loop is
10421 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10422 Alan Hayward <alan.hayward@arm.com>
10423 David Sherwood <david.sherwood@arm.com>
10425 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10426 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10427 (dump_groups): Update accordingly.
10428 (iv_use::mem_type): New member variable.
10429 (address_p): New function.
10430 (record_use): Add a mem_type argument and initialize the new
10432 (record_group_use): Add a mem_type argument. Use address_p.
10433 Remove obsolete null checks of base_object. Update call to record_use.
10434 (find_interesting_uses_op): Update call to record_group_use.
10435 (find_interesting_uses_cond): Likewise.
10436 (find_interesting_uses_address): Likewise.
10437 (get_mem_type_for_internal_fn): New function.
10438 (find_address_like_use): Likewise.
10439 (find_interesting_uses_stmt): Try find_address_like_use before
10440 calling find_interesting_uses_op.
10441 (addr_offset_valid_p): Use the iv mem_type field as the type
10442 of the addressed memory.
10443 (add_autoinc_candidates): Likewise.
10444 (get_address_cost): Likewise.
10445 (split_small_address_groups_p): Use address_p.
10446 (split_address_groups): Likewise.
10447 (add_iv_candidate_for_use): Likewise.
10448 (autoinc_possible_for_pair): Likewise.
10449 (rewrite_groups): Likewise.
10450 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10451 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10452 (get_alias_ptr_type_for_ptr_address): New function.
10453 (rewrite_use_address): Rewrite address uses in calls that were
10454 identified by find_address_like_use.
10456 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10457 Alan Hayward <alan.hayward@arm.com>
10458 David Sherwood <david.sherwood@arm.com>
10460 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10462 * gimple-expr.h (is_gimple_addressable: Likewise.
10463 * gimple-expr.c (is_gimple_address): Likewise.
10464 * internal-fn.c (expand_call_mem_ref): New function.
10465 (expand_mask_load_optab_fn): Use it.
10466 (expand_mask_store_optab_fn): Likewise.
10468 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10469 Alan Hayward <alan.hayward@arm.com>
10470 David Sherwood <david.sherwood@arm.com>
10472 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10473 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10474 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10475 (cond_umax@var{mode}): Document.
10476 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10477 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10478 (cond_umin_optab, cond_umax_optab): New optabs.
10479 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10480 (COND_IOR, COND_XOR): New internal functions.
10481 * internal-fn.h (get_conditional_internal_fn): Declare.
10482 * internal-fn.c (cond_binary_direct): New macro.
10483 (expand_cond_binary_optab_fn): Likewise.
10484 (direct_cond_binary_optab_supported_p): Likewise.
10485 (get_conditional_internal_fn): New function.
10486 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10487 Cope with reduction statements that are vectorized as calls rather
10489 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10490 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10491 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10492 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10493 (UNSPEC_COND_EOR): New unspecs.
10494 (optab): Add mappings for them.
10495 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10496 (sve_int_op, sve_fp_op): New int attributes.
10498 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10499 Alan Hayward <alan.hayward@arm.com>
10500 David Sherwood <david.sherwood@arm.com>
10502 * optabs.def (while_ult_optab): New optab.
10503 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10504 * internal-fn.def (WHILE_ULT): New internal function.
10505 * internal-fn.h (direct_internal_fn_supported_p): New override
10506 that takes two types as argument.
10507 * internal-fn.c (while_direct): New macro.
10508 (expand_while_optab_fn): New function.
10509 (convert_optab_supported_p): Likewise.
10510 (direct_while_optab_supported_p): New macro.
10511 * wide-int.h (wi::udiv_ceil): New function.
10512 * tree-vectorizer.h (rgroup_masks): New structure.
10513 (vec_loop_masks): New typedef.
10514 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10515 and fully_masked_p.
10516 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10517 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10518 (vect_max_vf): New function.
10519 (slpeel_make_loop_iterate_ntimes): Delete.
10520 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10521 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10522 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10523 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10524 internal-fn.h, stor-layout.h and optabs-query.h.
10525 (vect_set_loop_mask): New function.
10526 (add_preheader_seq): Likewise.
10527 (add_header_seq): Likewise.
10528 (interleave_supported_p): Likewise.
10529 (vect_maybe_permute_loop_masks): Likewise.
10530 (vect_set_loop_masks_directly): Likewise.
10531 (vect_set_loop_condition_masked): Likewise.
10532 (vect_set_loop_condition_unmasked): New function, split out from
10533 slpeel_make_loop_iterate_ntimes.
10534 (slpeel_make_loop_iterate_ntimes): Rename to..
10535 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10536 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10537 (vect_do_peeling): Update call accordingly.
10538 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10540 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10541 mask_compare_type, can_fully_mask_p and fully_masked_p.
10542 (release_vec_loop_masks): New function.
10543 (_loop_vec_info): Use it to free the loop masks.
10544 (can_produce_all_loop_masks_p): New function.
10545 (vect_get_max_nscalars_per_iter): Likewise.
10546 (vect_verify_full_masking): Likewise.
10547 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10548 retries, and free the mask rgroups before retrying. Check loop-wide
10549 reasons for disallowing fully-masked loops. Make the final decision
10550 about whether use a fully-masked loop or not.
10551 (vect_estimate_min_profitable_iters): Do not assume that peeling
10552 for the number of iterations will be needed for fully-masked loops.
10553 (vectorizable_reduction): Disable fully-masked loops.
10554 (vectorizable_live_operation): Likewise.
10555 (vect_halve_mask_nunits): New function.
10556 (vect_double_mask_nunits): Likewise.
10557 (vect_record_loop_mask): Likewise.
10558 (vect_get_loop_mask): Likewise.
10559 (vect_transform_loop): Handle the case in which the final loop
10560 iteration might handle a partial vector. Call vect_set_loop_condition
10561 instead of slpeel_make_loop_iterate_ntimes.
10562 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10563 (check_load_store_masking): New function.
10564 (prepare_load_store_mask): Likewise.
10565 (vectorizable_store): Handle fully-masked loops.
10566 (vectorizable_load): Likewise.
10567 (supportable_widening_operation): Use vect_halve_mask_nunits for
10569 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10570 (vect_gen_while): New function.
10571 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10572 (aarch64_uqdec<mode>): New insn.
10574 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10575 Alan Hayward <alan.hayward@arm.com>
10576 David Sherwood <david.sherwood@arm.com>
10578 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10579 (reduc_xor_scal_optab): New optabs.
10580 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10581 (reduc_xor_scal_@var{m}): Document.
10582 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10583 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10584 internal functions.
10585 * fold-const-call.c (fold_const_call): Handle them.
10586 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10587 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10588 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10589 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10590 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10591 (UNSPEC_XORV): New unspecs.
10592 (optab): Add entries for them.
10593 (BITWISEV): New int iterator.
10594 (bit_reduc_op): New int attributes.
10596 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10597 Alan Hayward <alan.hayward@arm.com>
10598 David Sherwood <david.sherwood@arm.com>
10600 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10601 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10602 * optabs.def (vec_shl_insert_optab): New optab.
10603 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10604 (duplicate_and_interleave): Likewise.
10605 * tree-vect-loop.c: Include internal-fn.h.
10606 (neutral_op_for_slp_reduction): New function, split out from
10607 get_initial_defs_for_reduction.
10608 (get_initial_def_for_reduction): Handle option 2 for variable-length
10609 vectors by loading the neutral value into a vector and then shifting
10610 the initial value into element 0.
10611 (get_initial_defs_for_reduction): Replace the code argument with
10612 the neutral value calculated by neutral_op_for_slp_reduction.
10613 Use gimple_build_vector for constant-length vectors.
10614 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10615 but the first group_size elements have a neutral value.
10616 Use duplicate_and_interleave otherwise.
10617 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10618 Update call to get_initial_defs_for_reduction. Handle SLP
10619 reductions for variable-length vectors by creating one vector
10620 result for each scalar result, with the elements associated
10621 with other scalar results stubbed out with the neutral value.
10622 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10623 Require IFN_VEC_SHL_INSERT for double reductions on
10624 variable-length vectors, or SLP reductions that have
10625 a neutral value. Require can_duplicate_and_interleave_p
10626 support for variable-length unchained SLP reductions if there
10627 is no neutral value, such as for MIN/MAX reductions. Also require
10628 the number of vector elements to be a multiple of the number of
10629 SLP statements when doing variable-length unchained SLP reductions.
10630 Update call to vect_create_epilog_for_reduction.
10631 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10632 and remove initial values.
10633 (duplicate_and_interleave): Make public.
10634 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10635 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10637 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10638 Alan Hayward <alan.hayward@arm.com>
10639 David Sherwood <david.sherwood@arm.com>
10641 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10642 (can_duplicate_and_interleave_p): New function.
10643 (vect_get_and_check_slp_defs): Take the vector of statements
10644 rather than just the current one. Remove excess parentheses.
10645 Restriction rejectinon of vect_constant_def and vect_external_def
10646 for variable-length vectors to boolean types, or types for which
10647 can_duplicate_and_interleave_p is false.
10648 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10649 (duplicate_and_interleave): New function.
10650 (vect_get_constant_vectors): Use gimple_build_vector for
10651 constant-length vectors and suitable variable-length constant
10652 vectors. Use duplicate_and_interleave for other variable-length
10653 vectors. Don't defer the update when inserting new statements.
10655 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10656 Alan Hayward <alan.hayward@arm.com>
10657 David Sherwood <david.sherwood@arm.com>
10659 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10660 min_profitable_iters doesn't go negative.
10662 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10663 Alan Hayward <alan.hayward@arm.com>
10664 David Sherwood <david.sherwood@arm.com>
10666 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10667 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10668 * optabs.def (vec_mask_load_lanes_optab): New optab.
10669 (vec_mask_store_lanes_optab): Likewise.
10670 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10671 (MASK_STORE_LANES): Likewise.
10672 * internal-fn.c (mask_load_lanes_direct): New macro.
10673 (mask_store_lanes_direct): Likewise.
10674 (expand_mask_load_optab_fn): Handle masked operations.
10675 (expand_mask_load_lanes_optab_fn): New macro.
10676 (expand_mask_store_optab_fn): Handle masked operations.
10677 (expand_mask_store_lanes_optab_fn): New macro.
10678 (direct_mask_load_lanes_optab_supported_p): Likewise.
10679 (direct_mask_store_lanes_optab_supported_p): Likewise.
10680 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10682 (vect_load_lanes_supported): Likewise.
10683 * tree-vect-data-refs.c (strip_conversion): New function.
10684 (can_group_stmts_p): Likewise.
10685 (vect_analyze_data_ref_accesses): Use it instead of checking
10686 for a pair of assignments.
10687 (vect_store_lanes_supported): Take a masked_p parameter.
10688 (vect_load_lanes_supported): Likewise.
10689 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10690 vect_store_lanes_supported and vect_load_lanes_supported.
10691 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10692 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10693 parameter. Don't allow gaps for masked accesses.
10694 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10695 and vect_load_lanes_supported.
10696 (get_load_store_type): Take a masked_p parameter and update
10697 call to get_group_load_store_type.
10698 (vectorizable_store): Update call to get_load_store_type.
10699 Handle IFN_MASK_STORE_LANES.
10700 (vectorizable_load): Update call to get_load_store_type.
10701 Handle IFN_MASK_LOAD_LANES.
10703 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10704 Alan Hayward <alan.hayward@arm.com>
10705 David Sherwood <david.sherwood@arm.com>
10707 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10709 * config/aarch64/aarch64-protos.h
10710 (aarch64_sve_struct_memory_operand_p): Declare.
10711 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10712 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10713 (VPRED, vpred): Handle SVE structure modes.
10714 * config/aarch64/constraints.md (Utx): New constraint.
10715 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10716 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10717 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10718 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10719 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10720 structure modes. Split into pieces after RA.
10721 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10722 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10724 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10725 SVE structure modes.
10726 (aarch64_classify_address): Likewise.
10727 (sizetochar): Move earlier in file.
10728 (aarch64_print_operand): Handle SVE register lists.
10729 (aarch64_array_mode): New function.
10730 (aarch64_sve_struct_memory_operand_p): Likewise.
10731 (TARGET_ARRAY_MODE): Redefine.
10733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10734 Alan Hayward <alan.hayward@arm.com>
10735 David Sherwood <david.sherwood@arm.com>
10737 * target.def (array_mode): New target hook.
10738 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10739 * doc/tm.texi: Regenerate.
10740 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10741 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10742 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10743 targetm.array_mode.
10744 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10747 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10748 Alan Hayward <alan.hayward@arm.com>
10749 David Sherwood <david.sherwood@arm.com>
10751 * fold-const.c (fold_binary_loc): Check the argument types
10752 rather than the result type when testing for a vector operation.
10754 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10756 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10757 * doc/tm.texi: Regenerate.
10759 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10760 Alan Hayward <alan.hayward@arm.com>
10761 David Sherwood <david.sherwood@arm.com>
10763 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10764 (sve): Document new AArch64 extension.
10765 * doc/md.texi (w): Extend the description of the AArch64
10766 constraint to include SVE vectors.
10767 (Upl, Upa): Document new AArch64 predicate constraints.
10768 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10770 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10771 (msve-vector-bits=): New option.
10772 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10773 SVE when these are disabled.
10774 (sve): New extension.
10775 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10776 modes. Adjust their number of units based on aarch64_sve_vg.
10777 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10778 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10779 aarch64_addr_query_type.
10780 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10781 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10782 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10783 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10784 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10785 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10786 (aarch64_simd_imm_zero_p): Delete.
10787 (aarch64_check_zero_based_sve_index_immediate): Declare.
10788 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10789 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10790 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10791 (aarch64_sve_float_mul_immediate_p): Likewise.
10792 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10793 rather than an rtx.
10794 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10795 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10796 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10797 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10798 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10799 (aarch64_regmode_natural_size): Likewise.
10800 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10801 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10803 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10804 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10805 for VG and the SVE predicate registers.
10806 (V_ALIASES): Add a "z"-prefixed alias.
10807 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10808 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10809 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10810 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10811 (REG_CLASS_NAMES): Add entries for them.
10812 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10813 and the predicate registers.
10814 (aarch64_sve_vg): Declare.
10815 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10816 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10817 (REGMODE_NATURAL_SIZE): Define.
10818 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10820 * config/aarch64/aarch64.c: Include cfgrtl.h.
10821 (simd_immediate_info): Add a constructor for series vectors,
10822 and an associated step field.
10823 (aarch64_sve_vg): New variable.
10824 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10825 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10826 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10827 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10828 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10829 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10830 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10831 (aarch64_get_mask_mode): New functions.
10832 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10833 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10834 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10835 predicate modes and predicate registers. Explicitly restrict
10836 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10837 to store a vector mode if it is recognized by
10838 aarch64_classify_vector_mode.
10839 (aarch64_regmode_natural_size): New function.
10840 (aarch64_hard_regno_caller_save_mode): Return the original mode
10842 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10843 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10844 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10845 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10847 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10848 does not overlap dest if the function is frame-related. Handle
10850 (aarch64_split_add_offset): New function.
10851 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10852 them aarch64_add_offset.
10853 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10854 and update call to aarch64_sub_sp.
10855 (aarch64_add_cfa_expression): New function.
10856 (aarch64_expand_prologue): Pass extra temporary registers to the
10857 functions above. Handle the case in which we need to emit new
10858 DW_CFA_expressions for registers that were originally saved
10859 relative to the stack pointer, but now have to be expressed
10860 relative to the frame pointer.
10861 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10863 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10864 IP0 and IP1 values for SVE frames.
10865 (aarch64_expand_vec_series): New function.
10866 (aarch64_expand_sve_widened_duplicate): Likewise.
10867 (aarch64_expand_sve_const_vector): Likewise.
10868 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10869 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10870 into the register, rather than emitting a SET directly.
10871 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10872 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10873 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10874 (offset_9bit_signed_scaled_p): New functions.
10875 (aarch64_replicate_bitmask_imm): New function.
10876 (aarch64_bitmask_imm): Use it.
10877 (aarch64_cannot_force_const_mem): Reject expressions involving
10878 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10879 (aarch64_classify_index): Handle SVE indices, by requiring
10880 a plain register index with a scale that matches the element size.
10881 (aarch64_classify_address): Handle SVE addresses. Assert that
10882 the mode of the address is VOIDmode or an integer mode.
10883 Update call to aarch64_classify_symbol.
10884 (aarch64_classify_symbolic_expression): Update call to
10885 aarch64_classify_symbol.
10886 (aarch64_const_vec_all_in_range_p): New function.
10887 (aarch64_print_vector_float_operand): Likewise.
10888 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10889 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10890 and the FP immediates 1.0 and 0.5.
10891 (aarch64_print_address_internal): Handle SVE addresses.
10892 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10893 (aarch64_regno_regclass): Handle predicate registers.
10894 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10896 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10897 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10898 (aarch64_convert_sve_vector_bits): New function.
10899 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10900 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10901 rather than an rtx.
10902 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10903 Handle SVE vector and predicate modes. Accept VL-based constants
10904 that need only one temporary register, and VL offsets that require
10905 no temporary registers.
10906 (aarch64_conditional_register_usage): Mark the predicate registers
10907 as fixed if SVE isn't available.
10908 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10909 Return true for SVE vector and predicate modes.
10910 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10911 rather than an unsigned int. Handle SVE modes.
10912 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10914 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10916 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10917 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10918 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10919 (aarch64_sve_float_mul_immediate_p): New functions.
10920 (aarch64_sve_valid_immediate): New function.
10921 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10922 Explicitly reject structure modes. Check for INDEX constants.
10923 Handle PTRUE and PFALSE constants.
10924 (aarch64_check_zero_based_sve_index_immediate): New function.
10925 (aarch64_simd_imm_zero_p): Delete.
10926 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10927 vector modes. Accept constants in the range of CNT[BHWD].
10928 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10929 ask for an Advanced SIMD mode.
10930 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10931 (aarch64_simd_vector_alignment): Handle SVE predicates.
10932 (aarch64_vectorize_preferred_vector_alignment): New function.
10933 (aarch64_simd_vector_alignment_reachable): Use it instead of
10935 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10936 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10938 (MAX_VECT_LEN): Delete.
10939 (expand_vec_perm_d): Add a vec_flags field.
10940 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10941 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10942 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10944 (aarch64_evpc_rev): Rename to...
10945 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10946 (aarch64_evpc_rev_global): New function.
10947 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10948 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10950 (aarch64_evpc_sve_tbl): New function.
10951 (aarch64_expand_vec_perm_const_1): Update after rename of
10952 aarch64_evpc_rev. Handle SVE permutes too, trying
10953 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10954 than aarch64_evpc_tbl.
10955 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10956 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10957 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10958 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10959 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10960 (aarch64_expand_sve_vcond): New functions.
10961 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10962 of aarch64_vector_mode_p.
10963 (aarch64_dwarf_poly_indeterminate_value): New function.
10964 (aarch64_compute_pressure_classes): Likewise.
10965 (aarch64_can_change_mode_class): Likewise.
10966 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10967 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10968 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10969 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10970 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10971 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10972 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10973 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10975 (Dn, Dl, Dr): Accept const as well as const_vector.
10976 (Dz): Likewise. Compare against CONST0_RTX.
10977 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10978 of "vector" where appropriate.
10979 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10980 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10981 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10982 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10983 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10984 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10985 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10986 (v_int_equiv): Extend to SVE modes.
10987 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10989 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10990 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10991 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10992 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10993 (SVE_COND_FP_CMP): New int iterators.
10994 (perm_hilo): Handle the new unpack unspecs.
10995 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10997 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10998 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10999 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11000 (aarch64_equality_operator, aarch64_constant_vector_operand)
11001 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11002 (aarch64_sve_nonimmediate_operand): Likewise.
11003 (aarch64_sve_general_operand): Likewise.
11004 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11005 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11006 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11007 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11008 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11009 (aarch64_sve_float_arith_immediate): Likewise.
11010 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11011 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11012 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11013 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11014 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11015 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11016 (aarch64_sve_float_arith_operand): Likewise.
11017 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11018 (aarch64_sve_float_mul_operand): Likewise.
11019 (aarch64_sve_vec_perm_operand): Likewise.
11020 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11021 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11022 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11023 as well as const_vector.
11024 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11025 in file. Use CONST0_RTX and CONSTM1_RTX.
11026 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11027 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11028 Use aarch64_simd_imm_zero.
11029 * config/aarch64/aarch64-sve.md: New file.
11030 * config/aarch64/aarch64.md: Include it.
11031 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11032 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11033 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11034 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11035 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11036 (sve): New attribute.
11037 (enabled): Disable instructions with the sve attribute unless
11039 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11040 aarch64_expand_mov_immediate.
11041 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11042 CNT[BHSD] immediates.
11043 (movti): Split CONST_POLY_INT moves into two halves.
11044 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11045 Split additions that need a temporary here if the destination
11046 is the stack pointer.
11047 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11048 (*add<mode>3_poly_1): New instruction.
11049 (set_clobber_cc): New expander.
11051 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11053 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11054 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11055 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11056 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11057 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11058 Change innermode from fixed_mode_size to machine_mode.
11059 (simplify_subreg): Update call accordingly. Handle a constant-sized
11060 subreg of a variable-length CONST_VECTOR.
11062 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11063 Alan Hayward <alan.hayward@arm.com>
11064 David Sherwood <david.sherwood@arm.com>
11066 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11067 (add_offset_to_base): New function, split out from...
11068 (create_mem_ref): ...here. When handling a scale other than 1,
11069 check first whether the address is valid without the offset.
11070 Add it into the base if so, leaving the index and scale as-is.
11072 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11075 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11076 fold_for_warn before checking if arg2 is INTEGER_CST.
11078 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11080 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11081 (store_multiple_operation): Delete.
11082 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11083 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11084 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11085 guarded by TARGET_STRING.
11086 (rs6000_output_load_multiple): Delete.
11087 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11088 OPTION_MASK_STRING / TARGET_STRING handling.
11089 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11090 (const rs6000_opt_masks) <"string">: Change mask to 0.
11091 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11092 (MASK_STRING): Delete.
11093 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11095 (load_multiple): Delete.
11102 (store_multiple): Delete.
11109 (movmemsi_8reg): Delete.
11110 (corresponding unnamed define_insn): Delete.
11111 (movmemsi_6reg): Delete.
11112 (corresponding unnamed define_insn): Delete.
11113 (movmemsi_4reg): Delete.
11114 (corresponding unnamed define_insn): Delete.
11115 (movmemsi_2reg): Delete.
11116 (corresponding unnamed define_insn): Delete.
11117 (movmemsi_1reg): Delete.
11118 (corresponding unnamed define_insn): Delete.
11119 * config/rs6000/rs6000.opt (mno-string): New.
11120 (mstring): Replace by deprecation warning stub.
11121 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11123 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11125 * regrename.c (regrename_do_replace): If replacing the same
11126 reg multiple times, try to reuse last created gen_raw_REG.
11129 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11130 main to workaround a bug in GDB.
11132 2018-01-12 Tom de Vries <tom@codesourcery.com>
11135 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11137 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11139 PR rtl-optimization/80481
11140 * ira-color.c (get_cap_member): New function.
11141 (allocnos_conflict_by_live_ranges_p): Use it.
11142 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11143 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11145 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11148 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11149 (*saddl_se_1): Ditto.
11150 (*ssubsi_1): Ditto.
11151 (*ssubl_se_1): Ditto.
11153 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11155 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11156 rather than wi::to_widest for DR_INITs.
11157 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11158 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11159 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11161 (vect_analyze_group_access_1): Note that here.
11163 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11165 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11166 polynomial type sizes.
11168 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11170 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11171 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11172 (gimple_add_tmp_var): Likewise.
11174 2018-01-12 Martin Liska <mliska@suse.cz>
11176 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11177 (gimple_alloc_sizes): Likewise.
11178 (dump_gimple_statistics): Use PRIu64 in printf format.
11179 * gimple.h: Change uint64_t to int.
11181 2018-01-12 Martin Liska <mliska@suse.cz>
11183 * tree-core.h: Use uint64_t instead of int.
11184 * tree.c (tree_node_counts): Likewise.
11185 (tree_node_sizes): Likewise.
11186 (dump_tree_statistics): Use PRIu64 in printf format.
11188 2018-01-12 Martin Liska <mliska@suse.cz>
11190 * Makefile.in: As qsort_chk is implemented in vec.c, add
11191 vec.o to linkage of gencfn-macros.
11192 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11193 passing the info to record_node_allocation_statistics.
11194 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11196 * ggc-common.c (struct ggc_usage): Add operator== and use
11197 it in operator< and compare function.
11198 * mem-stats.h (struct mem_usage): Likewise.
11199 * vec.c (struct vec_usage): Remove operator< and compare
11200 function. Can be simply inherited.
11202 2018-01-12 Martin Jambor <mjambor@suse.cz>
11205 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11206 * tree-ssa-math-opts.c: Include domwalk.h.
11207 (convert_mult_to_fma_1): New function.
11208 (fma_transformation_info): New type.
11209 (fma_deferring_state): Likewise.
11210 (cancel_fma_deferring): New function.
11211 (result_of_phi): Likewise.
11212 (last_fma_candidate_feeds_initial_phi): Likewise.
11213 (convert_mult_to_fma): Added deferring logic, split actual
11214 transformation to convert_mult_to_fma_1.
11215 (math_opts_dom_walker): New type.
11216 (math_opts_dom_walker::after_dom_children): New method, body moved
11217 here from pass_optimize_widening_mul::execute, added deferring logic
11219 (pass_optimize_widening_mul::execute): Moved most of code to
11220 math_opts_dom_walker::after_dom_children.
11221 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11222 * config/i386/i386.c (ix86_option_override_internal): Added
11223 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11225 2018-01-12 Richard Biener <rguenther@suse.de>
11228 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11229 inline instance vars.
11231 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11234 * config/rx/rx.c (rx_is_restricted_memory_address):
11235 Handle SUBREG case.
11237 2018-01-12 Richard Biener <rguenther@suse.de>
11239 PR tree-optimization/80846
11240 * target.def (split_reduction): New target hook.
11241 * targhooks.c (default_split_reduction): New function.
11242 * targhooks.h (default_split_reduction): Declare.
11243 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11244 target requests first reduce vectors by combining low and high
11246 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11247 (get_vectype_for_scalar_type_and_size): Export.
11248 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11249 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11250 * doc/tm.texi: Regenerate.
11251 * config/i386/i386.c (ix86_split_reduction): Implement
11252 TARGET_VECTORIZE_SPLIT_REDUCTION.
11254 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11257 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11258 in PIC mode except for TARGET_VXWORKS_RTP.
11259 * config/sparc/sparc.c: Include cfgrtl.h.
11260 (TARGET_INIT_PIC_REG): Define.
11261 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11262 (sparc_pic_register_p): New predicate.
11263 (sparc_legitimate_address_p): Use it.
11264 (sparc_legitimize_pic_address): Likewise.
11265 (sparc_delegitimize_address): Likewise.
11266 (sparc_mode_dependent_address_p): Likewise.
11267 (gen_load_pcrel_sym): Remove 4th parameter.
11268 (load_got_register): Adjust call to above. Remove obsolete stuff.
11269 (sparc_expand_prologue): Do not call load_got_register here.
11270 (sparc_flat_expand_prologue): Likewise.
11271 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11272 (sparc_use_pseudo_pic_reg): New function.
11273 (sparc_init_pic_reg): Likewise.
11274 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11275 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11277 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11279 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11280 Add item for branch_cost.
11282 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11284 PR rtl-optimization/83565
11285 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11286 not extend the result to a larger mode for rotate operations.
11287 (num_sign_bit_copies1): Likewise.
11289 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11292 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11294 Use values-Xc.o for -pedantic.
11295 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11297 2018-01-12 Martin Liska <mliska@suse.cz>
11300 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11302 (possible_polymorphic_call_targets): Use it.
11303 (ipa_devirt): Likewise.
11305 2018-01-12 Martin Liska <mliska@suse.cz>
11307 * profile-count.h (enum profile_quality): Use 0 as invalid
11308 enum value of profile_quality.
11310 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11312 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11313 -mext-string options.
11315 2018-01-12 Richard Biener <rguenther@suse.de>
11317 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11318 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11319 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11321 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11323 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11325 * configure.ac (--with-long-double-format): Add support for the
11326 configuration option to change the default long double format on
11328 * config.gcc (powerpc*-linux*-*): Likewise.
11329 * configure: Regenerate.
11330 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11331 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11332 used without modification.
11334 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11336 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11337 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11338 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11339 MISC_BUILTIN_SPEC_BARRIER.
11340 (rs6000_init_builtins): Likewise.
11341 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11343 (speculation_barrier): New define_insn.
11344 * doc/extend.texi: Document __builtin_speculation_barrier.
11346 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11349 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11350 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11351 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11353 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11354 integral modes instead of "ss" and "sd".
11355 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11356 vectors with 32-bit and 64-bit elements.
11357 (vecdupssescalarmodesuffix): New mode attribute.
11358 (vec_dup<mode>): Use it.
11360 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11363 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11364 frame if argument is passed on stack.
11366 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11369 * ree.c (combine_reaching_defs): Optimize also
11370 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11371 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11373 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11375 PR middle-end/83189
11376 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11378 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11380 PR middle-end/83718
11381 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11382 after they are computed.
11384 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11386 PR tree-optimization/83695
11387 * gimple-loop-linterchange.cc
11388 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11389 reset cached scev information after interchange.
11390 (pass_linterchange::execute): Remove call to scev_reset_htab.
11392 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11394 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11395 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11396 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11397 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11398 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11399 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11400 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11401 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11402 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11403 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11404 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11405 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11406 (V_lane_reg): Likewise.
11407 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11409 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11410 (vfmal_lane_low<mode>_intrinsic,
11411 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11412 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11413 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11414 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11415 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11416 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11418 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11420 * config/arm/arm-cpus.in (fp16fml): New feature.
11421 (ALL_SIMD): Add fp16fml.
11422 (armv8.2-a): Add fp16fml as an option.
11423 (armv8.3-a): Likewise.
11424 (armv8.4-a): Add fp16fml as part of fp16.
11425 * config/arm/arm.h (TARGET_FP16FML): Define.
11426 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11428 * config/arm/arm-modes.def (V2HF): Define.
11429 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11430 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11431 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11432 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11433 vfmsl_low, vfmsl_high): New set of builtins.
11434 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11435 (vfml_op): New code attribute.
11436 (VFMLHALVES): New int iterator.
11437 (VFML, VFMLSEL): New mode attributes.
11438 (V_reg): Define mapping for V2HF.
11439 (V_hi, V_lo): New mode attributes.
11440 (VF_constraint): Likewise.
11441 (vfml_half, vfml_half_selector): New int attributes.
11442 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11444 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11445 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11447 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11448 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11449 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11450 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11452 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11453 Document new effective target and option set.
11455 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11457 * config/arm/arm-cpus.in (armv8_4): New feature.
11458 (ARMv8_4a): New fgroup.
11459 (armv8.4-a): New arch.
11460 * config/arm/arm-tables.opt: Regenerate.
11461 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11462 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11463 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11464 Add matching rules for -march=armv8.4-a and extensions.
11465 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11467 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11470 * config/rx/rx.md (BW): New mode attribute.
11471 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11473 2018-01-11 Richard Biener <rguenther@suse.de>
11475 PR tree-optimization/83435
11476 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11477 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11478 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11480 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11481 Alan Hayward <alan.hayward@arm.com>
11482 David Sherwood <david.sherwood@arm.com>
11484 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11486 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11487 (aarch64_print_address_internal): Use it to check for a zero offset.
11489 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11490 Alan Hayward <alan.hayward@arm.com>
11491 David Sherwood <david.sherwood@arm.com>
11493 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11494 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11495 Return a poly_int64 rather than a HOST_WIDE_INT.
11496 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11497 rather than a HOST_WIDE_INT.
11498 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11499 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11500 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11501 final_offset from HOST_WIDE_INT to poly_int64.
11502 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11503 to_constant when getting the number of units in an Advanced SIMD
11505 (aarch64_builtin_vectorized_function): Check for a constant number
11507 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11509 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11510 attribute instead of GET_MODE_NUNITS.
11511 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11512 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11513 GET_MODE_SIZE for fixed-size registers.
11514 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11515 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11516 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11517 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11518 (aarch64_print_operand, aarch64_print_address_internal)
11519 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11520 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11521 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11522 Handle polynomial GET_MODE_SIZE.
11523 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11524 wider than SImode without modification.
11525 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11526 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11527 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11528 passing and returning SVE modes.
11529 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11530 rather than GEN_INT.
11531 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11532 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11533 (aarch64_allocate_and_probe_stack_space): Likewise.
11534 (aarch64_layout_frame): Cope with polynomial offsets.
11535 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11536 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11537 polynomial offsets.
11538 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11539 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11540 poly_int64 rather than a HOST_WIDE_INT.
11541 (aarch64_get_separate_components, aarch64_process_components)
11542 (aarch64_expand_prologue, aarch64_expand_epilogue)
11543 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11544 (aarch64_anchor_offset): New function, split out from...
11545 (aarch64_legitimize_address): ...here.
11546 (aarch64_builtin_vectorization_cost): Handle polynomial
11547 TYPE_VECTOR_SUBPARTS.
11548 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11550 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11551 number of elements from the PARALLEL rather than the mode.
11552 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11553 rather than GET_MODE_BITSIZE.
11554 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11555 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11556 (aarch64_expand_vec_perm_const_1): Handle polynomial
11557 d->perm.length () and d->perm elements.
11558 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11559 Apply to_constant to d->perm elements.
11560 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11561 polynomial CONST_VECTOR_NUNITS.
11562 (aarch64_move_pointer): Take amount as a poly_int64 rather
11564 (aarch64_progress_pointer): Avoid temporary variable.
11565 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11566 the mode attribute instead of GET_MODE.
11568 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11569 Alan Hayward <alan.hayward@arm.com>
11570 David Sherwood <david.sherwood@arm.com>
11572 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11573 x exists before using it.
11574 (aarch64_add_constant_internal): Rename to...
11575 (aarch64_add_offset_1): ...this. Replace regnum with separate
11576 src and dest rtxes. Handle the case in which they're different,
11577 including when the offset is zero. Replace scratchreg with an rtx.
11578 Use 2 additions if there is no spare register into which we can
11579 move a 16-bit constant.
11580 (aarch64_add_constant): Delete.
11581 (aarch64_add_offset): Replace reg with separate src and dest
11582 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11583 Use aarch64_add_offset_1.
11584 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11585 an rtx rather than an int. Take the delta as a poly_int64
11586 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11587 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11588 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11589 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11590 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11591 and aarch64_add_sp.
11592 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11593 aarch64_add_constant.
11595 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11597 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11598 Use scalar_float_mode.
11600 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11602 * config/aarch64/aarch64-simd.md
11603 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11604 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11605 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11606 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11607 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11608 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11609 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11610 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11611 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11612 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11614 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11617 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11618 targ_options->x_arm_arch_string is non NULL.
11620 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11622 * config/aarch64/aarch64.h
11623 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11625 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11628 * expmed.c (emit_store_flag_force): Swap if const op0
11629 and change VOIDmode to mode of op0.
11631 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11633 PR rtl-optimization/83761
11634 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11635 than bytes to mode_for_size.
11637 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11639 PR middle-end/83189
11640 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11641 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11644 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11646 PR middle-end/83575
11647 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11648 when in layout mode.
11649 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11650 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11653 2018-01-10 Michael Collison <michael.collison@arm.com>
11655 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11656 * config/aarch64/aarch64-option-extension.def: Add
11657 AARCH64_OPT_EXTENSION of 'fp16fml'.
11658 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11659 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11660 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11661 * config/aarch64/constraints.md (Ui7): New constraint.
11662 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11663 (VFMLA_SEL_W): Ditto.
11666 (VFMLA16_LOW): New int iterator.
11667 (VFMLA16_HIGH): Ditto.
11668 (UNSPEC_FMLAL): New unspec.
11669 (UNSPEC_FMLSL): Ditto.
11670 (UNSPEC_FMLAL2): Ditto.
11671 (UNSPEC_FMLSL2): Ditto.
11672 (f16mac): New code attribute.
11673 * config/aarch64/aarch64-simd-builtins.def
11674 (aarch64_fmlal_lowv2sf): Ditto.
11675 (aarch64_fmlsl_lowv2sf): Ditto.
11676 (aarch64_fmlalq_lowv4sf): Ditto.
11677 (aarch64_fmlslq_lowv4sf): Ditto.
11678 (aarch64_fmlal_highv2sf): Ditto.
11679 (aarch64_fmlsl_highv2sf): Ditto.
11680 (aarch64_fmlalq_highv4sf): Ditto.
11681 (aarch64_fmlslq_highv4sf): Ditto.
11682 (aarch64_fmlal_lane_lowv2sf): Ditto.
11683 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11684 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11685 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11686 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11687 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11688 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11689 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11690 (aarch64_fmlal_lane_highv2sf): Ditto.
11691 (aarch64_fmlsl_lane_highv2sf): Ditto.
11692 (aarch64_fmlal_laneq_highv2sf): Ditto.
11693 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11694 (aarch64_fmlalq_lane_highv4sf): Ditto.
11695 (aarch64_fmlsl_lane_highv4sf): Ditto.
11696 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11697 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11698 * config/aarch64/aarch64-simd.md:
11699 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11700 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11701 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11702 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11703 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11704 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11705 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11706 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11707 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11708 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11709 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11710 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11711 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11712 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11713 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11714 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11715 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11716 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11717 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11718 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11719 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11720 (vfmlsl_low_u32): Ditto.
11721 (vfmlalq_low_u32): Ditto.
11722 (vfmlslq_low_u32): Ditto.
11723 (vfmlal_high_u32): Ditto.
11724 (vfmlsl_high_u32): Ditto.
11725 (vfmlalq_high_u32): Ditto.
11726 (vfmlslq_high_u32): Ditto.
11727 (vfmlal_lane_low_u32): Ditto.
11728 (vfmlsl_lane_low_u32): Ditto.
11729 (vfmlal_laneq_low_u32): Ditto.
11730 (vfmlsl_laneq_low_u32): Ditto.
11731 (vfmlalq_lane_low_u32): Ditto.
11732 (vfmlslq_lane_low_u32): Ditto.
11733 (vfmlalq_laneq_low_u32): Ditto.
11734 (vfmlslq_laneq_low_u32): Ditto.
11735 (vfmlal_lane_high_u32): Ditto.
11736 (vfmlsl_lane_high_u32): Ditto.
11737 (vfmlal_laneq_high_u32): Ditto.
11738 (vfmlsl_laneq_high_u32): Ditto.
11739 (vfmlalq_lane_high_u32): Ditto.
11740 (vfmlslq_lane_high_u32): Ditto.
11741 (vfmlalq_laneq_high_u32): Ditto.
11742 (vfmlslq_laneq_high_u32): Ditto.
11743 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11744 (AARCH64_FL_FOR_ARCH8_4): New.
11745 (AARCH64_ISA_F16FML): New ISA flag.
11746 (TARGET_F16FML): New feature flag for fp16fml.
11747 (doc/invoke.texi): Document new fp16fml option.
11749 2018-01-10 Michael Collison <michael.collison@arm.com>
11751 * config/aarch64/aarch64-builtins.c:
11752 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11753 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11754 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11755 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11756 (AARCH64_ISA_SHA3): New ISA flag.
11757 (TARGET_SHA3): New feature flag for sha3.
11758 * config/aarch64/iterators.md (sha512_op): New int attribute.
11759 (CRYPTO_SHA512): New int iterator.
11760 (UNSPEC_SHA512H): New unspec.
11761 (UNSPEC_SHA512H2): Ditto.
11762 (UNSPEC_SHA512SU0): Ditto.
11763 (UNSPEC_SHA512SU1): Ditto.
11764 * config/aarch64/aarch64-simd-builtins.def
11765 (aarch64_crypto_sha512hqv2di): New builtin.
11766 (aarch64_crypto_sha512h2qv2di): Ditto.
11767 (aarch64_crypto_sha512su0qv2di): Ditto.
11768 (aarch64_crypto_sha512su1qv2di): Ditto.
11769 (aarch64_eor3qv8hi): Ditto.
11770 (aarch64_rax1qv2di): Ditto.
11771 (aarch64_xarqv2di): Ditto.
11772 (aarch64_bcaxqv8hi): Ditto.
11773 * config/aarch64/aarch64-simd.md:
11774 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11775 (aarch64_crypto_sha512su0qv2di): Ditto.
11776 (aarch64_crypto_sha512su1qv2di): Ditto.
11777 (aarch64_eor3qv8hi): Ditto.
11778 (aarch64_rax1qv2di): Ditto.
11779 (aarch64_xarqv2di): Ditto.
11780 (aarch64_bcaxqv8hi): Ditto.
11781 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11782 (vsha512h2q_u64): Ditto.
11783 (vsha512su0q_u64): Ditto.
11784 (vsha512su1q_u64): Ditto.
11785 (veor3q_u16): Ditto.
11786 (vrax1q_u64): Ditto.
11787 (vxarq_u64): Ditto.
11788 (vbcaxq_u16): Ditto.
11789 * config/arm/types.md (crypto_sha512): New type attribute.
11790 (crypto_sha3): Ditto.
11791 (doc/invoke.texi): Document new sha3 option.
11793 2018-01-10 Michael Collison <michael.collison@arm.com>
11795 * config/aarch64/aarch64-builtins.c:
11796 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11797 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11798 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11799 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11800 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11801 (AARCH64_ISA_SM4): New ISA flag.
11802 (TARGET_SM4): New feature flag for sm4.
11803 * config/aarch64/aarch64-simd-builtins.def
11804 (aarch64_sm3ss1qv4si): Ditto.
11805 (aarch64_sm3tt1aq4si): Ditto.
11806 (aarch64_sm3tt1bq4si): Ditto.
11807 (aarch64_sm3tt2aq4si): Ditto.
11808 (aarch64_sm3tt2bq4si): Ditto.
11809 (aarch64_sm3partw1qv4si): Ditto.
11810 (aarch64_sm3partw2qv4si): Ditto.
11811 (aarch64_sm4eqv4si): Ditto.
11812 (aarch64_sm4ekeyqv4si): Ditto.
11813 * config/aarch64/aarch64-simd.md:
11814 (aarch64_sm3ss1qv4si): Ditto.
11815 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11816 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11817 (aarch64_sm4eqv4si): Ditto.
11818 (aarch64_sm4ekeyqv4si): Ditto.
11819 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11820 (sm3part_op): Ditto.
11821 (CRYPTO_SM3TT): Ditto.
11822 (CRYPTO_SM3PART): Ditto.
11823 (UNSPEC_SM3SS1): New unspec.
11824 (UNSPEC_SM3TT1A): Ditto.
11825 (UNSPEC_SM3TT1B): Ditto.
11826 (UNSPEC_SM3TT2A): Ditto.
11827 (UNSPEC_SM3TT2B): Ditto.
11828 (UNSPEC_SM3PARTW1): Ditto.
11829 (UNSPEC_SM3PARTW2): Ditto.
11830 (UNSPEC_SM4E): Ditto.
11831 (UNSPEC_SM4EKEY): Ditto.
11832 * config/aarch64/constraints.md (Ui2): New constraint.
11833 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11834 * config/arm/types.md (crypto_sm3): New type attribute.
11835 (crypto_sm4): Ditto.
11836 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11837 (vsm3tt1aq_u32): Ditto.
11838 (vsm3tt1bq_u32): Ditto.
11839 (vsm3tt2aq_u32): Ditto.
11840 (vsm3tt2bq_u32): Ditto.
11841 (vsm3partw1q_u32): Ditto.
11842 (vsm3partw2q_u32): Ditto.
11843 (vsm4eq_u32): Ditto.
11844 (vsm4ekeyq_u32): Ditto.
11845 (doc/invoke.texi): Document new sm4 option.
11847 2018-01-10 Michael Collison <michael.collison@arm.com>
11849 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11850 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11851 (AARCH64_FL_FOR_ARCH8_4): New.
11852 (AARCH64_FL_V8_4): New flag.
11853 (doc/invoke.texi): Document new armv8.4-a option.
11855 2018-01-10 Michael Collison <michael.collison@arm.com>
11857 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11858 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11859 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11860 * config/aarch64/aarch64-option-extension.def: Add
11861 AARCH64_OPT_EXTENSION of 'sha2'.
11862 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11863 (crypto): Disable sha2 and aes if crypto disabled.
11864 (crypto): Enable aes and sha2 if enabled.
11865 (simd): Disable sha2 and aes if simd disabled.
11866 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11868 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11869 (TARGET_SHA2): New feature flag for sha2.
11870 (TARGET_AES): New feature flag for aes.
11871 * config/aarch64/aarch64-simd.md:
11872 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11873 conditional on TARGET_AES.
11874 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11875 (aarch64_crypto_sha1hsi): Make pattern conditional
11877 (aarch64_crypto_sha1hv4si): Ditto.
11878 (aarch64_be_crypto_sha1hv4si): Ditto.
11879 (aarch64_crypto_sha1su1v4si): Ditto.
11880 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11881 (aarch64_crypto_sha1su0v4si): Ditto.
11882 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11883 (aarch64_crypto_sha256su0v4si): Ditto.
11884 (aarch64_crypto_sha256su1v4si): Ditto.
11885 (doc/invoke.texi): Document new aes and sha2 options.
11887 2018-01-10 Martin Sebor <msebor@redhat.com>
11889 PR tree-optimization/83781
11890 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11893 2018-01-11 Martin Sebor <msebor@gmail.com>
11894 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11896 PR tree-optimization/83501
11897 PR tree-optimization/81703
11899 * tree-ssa-strlen.c (get_string_cst): Rename...
11900 (get_string_len): ...to this. Handle global constants.
11901 (handle_char_store): Adjust.
11903 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11904 Jim Wilson <jimw@sifive.com>
11906 * config/riscv/riscv-protos.h (riscv_output_return): New.
11907 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11908 (riscv_attribute_table, riscv_output_return),
11909 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11910 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11911 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11913 (riscv_expand_prologue): Add early return for naked function.
11914 (riscv_expand_epilogue): Likewise.
11915 (riscv_function_ok_for_sibcall): Return false for naked function.
11916 (riscv_set_current_function): New.
11917 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11918 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11919 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11920 * doc/extend.texi (RISC-V Function Attributes): New.
11922 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11924 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11925 check for 128-bit long double before checking TCmode.
11926 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11927 128-bit long doubles before checking TFmode or TCmode.
11928 (FLOAT128_IBM_P): Likewise.
11930 2018-01-10 Martin Sebor <msebor@redhat.com>
11932 PR tree-optimization/83671
11933 * builtins.c (c_strlen): Unconditionally return zero for the empty
11935 Use -Warray-bounds for warnings.
11936 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11937 for non-constant array indices with COMPONENT_REF, arrays of
11938 arrays, and pointers to arrays.
11939 (gimple_fold_builtin_strlen): Determine and set length range for
11940 non-constant character arrays.
11942 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11944 PR middle-end/81897
11945 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11948 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11950 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11952 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11955 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11956 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11957 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11958 indexed_or_indirect_operand predicate.
11959 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11960 (*vsx_le_perm_load_v8hi): Likewise.
11961 (*vsx_le_perm_load_v16qi): Likewise.
11962 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11963 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11964 (*vsx_le_perm_store_v8hi): Likewise.
11965 (*vsx_le_perm_store_v16qi): Likewise.
11966 (eight unnamed splitters): Likewise.
11968 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11970 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11971 * config/rs6000/emmintrin.h: Likewise.
11972 * config/rs6000/mmintrin.h: Likewise.
11973 * config/rs6000/xmmintrin.h: Likewise.
11975 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11978 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11980 * tree.c (tree_nop_conversion): Return true for location wrapper
11982 (maybe_wrap_with_location): New function.
11983 (selftest::check_strip_nops): New function.
11984 (selftest::test_location_wrappers): New function.
11985 (selftest::tree_c_tests): Call it.
11986 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11987 (maybe_wrap_with_location): New decl.
11988 (EXPR_LOCATION_WRAPPER_P): New macro.
11989 (location_wrapper_p): New inline function.
11990 (tree_strip_any_location_wrapper): New inline function.
11992 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11995 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11996 stack_realign_offset for the largest alignment of stack slot
11998 (ix86_find_max_used_stack_alignment): New function.
11999 (ix86_finalize_stack_frame_flags): Use it. Set
12000 max_used_stack_alignment if we don't realign stack.
12001 * config/i386/i386.h (machine_function): Add
12002 max_used_stack_alignment.
12004 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12006 * config/arm/arm.opt (-mbranch-cost): New option.
12007 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12010 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12013 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12014 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12016 2018-01-10 Richard Biener <rguenther@suse.de>
12019 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12020 early out so it also covers the case where we have a non-NULL
12023 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12025 PR tree-optimization/83753
12026 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12027 for non-strided grouped accesses if the number of elements is 1.
12029 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12032 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12033 * i386.h (TARGET_USE_GATHER): Define.
12034 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12036 2018-01-10 Martin Liska <mliska@suse.cz>
12039 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12040 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12042 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12043 CLEANUP_NO_PARTITIONING is not set.
12045 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12047 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12048 for vectors, as a partial revert of r254296.
12049 * rtl.h (const_vec_p): Delete.
12050 (const_vec_duplicate_p): Don't test for vector CONSTs.
12051 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12052 * expmed.c (make_tree): Likewise.
12055 * common.md (E, F): Use CONSTANT_P instead of checking for
12057 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12058 checking for CONST_VECTOR.
12060 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12062 PR middle-end/83575
12063 * predict.c (force_edge_cold): Handle in more sane way edges
12064 with no prediction.
12066 2018-01-09 Carl Love <cel@us.ibm.com>
12068 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12070 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12071 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12072 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12073 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12074 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12075 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12076 * config/rs6000/rs6000-protos.h: Add extern defition for
12077 rs6000_generate_float2_double_code.
12078 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12080 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12081 (float2_v2df): Add define_expand.
12083 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12086 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12087 op_mode in the force_to_mode call.
12089 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12091 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12092 instead of checking each element individually.
12093 (aarch64_evpc_uzp): Likewise.
12094 (aarch64_evpc_zip): Likewise.
12095 (aarch64_evpc_ext): Likewise.
12096 (aarch64_evpc_rev): Likewise.
12097 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12098 instead of checking each element individually. Return true without
12100 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12101 whether all selected elements come from the same input, instead of
12102 checking each element individually. Remove calls to gen_rtx_REG,
12103 start_sequence and end_sequence and instead assert that no rtl is
12106 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12108 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12109 order of HIGH and CONST checks.
12111 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12113 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12114 if the destination isn't an SSA_NAME.
12116 2018-01-09 Richard Biener <rguenther@suse.de>
12118 PR tree-optimization/83668
12119 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12121 (canonicalize_loop_form): ... here, renamed from ...
12122 (canonicalize_loop_closed_ssa_form): ... this and amended to
12123 swap successor edges for loop exit blocks to make us use
12124 the RPO order we need for initial schedule generation.
12126 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12128 PR tree-optimization/64811
12129 * match.pd: When optimizing comparisons with Inf, avoid
12130 introducing or losing exceptions from comparisons with NaN.
12132 2018-01-09 Martin Liska <mliska@suse.cz>
12135 * asan.c (shadow_mem_size): Add gcc_assert.
12137 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12139 Don't save registers in main().
12142 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12143 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12144 * config/avr/avr.c (avr_set_current_function): Don't error if
12145 naked, OS_task or OS_main are specified at the same time.
12146 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12148 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12150 * common/config/avr/avr-common.c (avr_option_optimization_table):
12151 Switch on -mmain-is-OS_task for optimizing compilations.
12153 2018-01-09 Richard Biener <rguenther@suse.de>
12155 PR tree-optimization/83572
12156 * graphite.c: Include cfganal.h.
12157 (graphite_transform_loops): Connect infinite loops to exit
12158 and remove fake edges at the end.
12160 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12162 * ipa-inline.c (edge_badness): Revert accidental checkin.
12164 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12167 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12168 symbols; not inline clones.
12170 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12173 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12174 hard registers. Formatting fixes.
12176 PR preprocessor/83722
12177 * gcc.c (try_generate_repro): Pass
12178 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12179 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12182 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12183 Kito Cheng <kito.cheng@gmail.com>
12185 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12186 (riscv_leaf_function_p): Delete.
12187 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12189 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12191 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12193 (do_ifelse): New function.
12194 (do_isel): New function.
12195 (do_sub3): New function.
12196 (do_add3): New function.
12197 (do_load_mask_compare): New function.
12198 (do_overlap_load_compare): New function.
12199 (expand_compare_loop): New function.
12200 (expand_block_compare): Call expand_compare_loop() when appropriate.
12201 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12202 option description.
12203 (-mblock-compare-inline-loop-limit): New option.
12205 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12208 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12209 Reverse order of second and third operands in first alternative.
12210 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12211 of first and second elements in UNSPEC_VPERMR vector.
12212 (altivec_expand_vec_perm_le): Likewise.
12214 2018-01-08 Jeff Law <law@redhat.com>
12216 PR rtl-optimizatin/81308
12217 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12218 (process_switch): If group_case_labels makes a change, then set
12220 (pass_convert_switch::execute): If a switch is converted, then
12221 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12223 PR rtl-optimization/81308
12224 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12227 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12229 PR target/83663 - Revert r255946
12230 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12231 generation for cases where splatting a value is not useful.
12232 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12233 across a vec_duplicate and a paradoxical subreg forming a vector
12234 mode to a vec_concat.
12236 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12238 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12239 -march=armv8.3-a variants.
12240 * config/arm/t-multilib: Likewise.
12241 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12243 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12245 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12247 (cceq_ior_compare_complement): Give it a name so I can use it, and
12248 change boolean_or_operator predicate to boolean_operator so it can
12249 be used to generate a crand.
12250 (eqne): New code iterator.
12251 (bd/bd_neg): New code_attrs.
12252 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12253 a single define_insn.
12254 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12255 decrement (bdnzt/bdnzf/bdzt/bdzf).
12256 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12257 with the new names of the branch decrement patterns, and added the
12258 names of the branch decrement conditional patterns.
12260 2018-01-08 Richard Biener <rguenther@suse.de>
12262 PR tree-optimization/83563
12263 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12266 2018-01-08 Richard Biener <rguenther@suse.de>
12268 PR middle-end/83713
12269 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12271 2018-01-08 Richard Biener <rguenther@suse.de>
12273 PR tree-optimization/83685
12274 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12275 references to abnormals.
12277 2018-01-08 Richard Biener <rguenther@suse.de>
12280 * dwarf2out.c (output_indirect_strings): Handle empty
12281 skeleton_debug_str_hash.
12282 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12284 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12286 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12287 (emit_store_direct): Likewise.
12288 (arc_trampoline_adjust_address): Likewise.
12289 (arc_asm_trampoline_template): New function.
12290 (arc_initialize_trampoline): Use asm_trampoline_template.
12291 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12292 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12293 * config/arc/arc.md (flush_icache): Delete pattern.
12295 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12297 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12298 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12301 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12304 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12305 by not USED_FOR_TARGET.
12306 (make_pass_resolve_sw_modes): Likewise.
12308 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12310 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12313 2018-01-08 Richard Biener <rguenther@suse.de>
12315 PR middle-end/83580
12316 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12318 2018-01-08 Richard Biener <rguenther@suse.de>
12320 PR middle-end/83517
12321 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12323 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12325 PR middle-end/81897
12326 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12327 basic blocks with a small number of successors.
12328 (convert_control_dep_chain_into_preds): Improve handling of
12330 (dump_predicates): Split apart into...
12331 (dump_pred_chain): ...here...
12332 (dump_pred_info): ...and here.
12333 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12334 (can_chain_union_be_invalidated_p): Improve check for invalidation
12336 (uninit_uses_cannot_happen): Avoid unnecessary if
12337 convert_control_dep_chain_into_preds yielded nothing.
12339 2018-01-06 Martin Sebor <msebor@redhat.com>
12341 PR tree-optimization/83640
12342 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12343 subtracting negative offset from size.
12344 (builtin_access::overlap): Adjust offset bounds of the access to fall
12345 within the size of the object if possible.
12347 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12349 PR rtl-optimization/83699
12350 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12351 extract_bit_field_as_subreg to cases in which the extracted
12352 value is also a vector.
12354 * lra-constraints.c (process_alt_operands): Test for the equivalence
12355 substitutions when detecting a possible reload cycle.
12357 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12360 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12361 by default if flag_selective_schedling{,2}. Formatting fixes.
12363 PR rtl-optimization/83682
12364 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12365 if it has non-VECTOR_MODE element mode.
12366 (vec_duplicate_p): Likewise.
12368 PR middle-end/83694
12369 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12370 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12372 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12375 * config/i386/i386-builtin.def
12376 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12377 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12378 Require also OPTION_MASK_ISA_AVX512F in addition to
12379 OPTION_MASK_ISA_GFNI.
12380 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12381 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12382 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12383 to OPTION_MASK_ISA_GFNI.
12384 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12385 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12386 OPTION_MASK_ISA_AVX512BW.
12387 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12388 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12389 addition to OPTION_MASK_ISA_GFNI.
12390 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12391 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12392 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12393 to OPTION_MASK_ISA_GFNI.
12394 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12395 a requirement for all ISAs rather than any of them with a few
12397 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12399 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12400 bitmasks to be enabled with 3 exceptions, instead of requiring any
12401 enabled ISA with lots of exceptions.
12402 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12403 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12404 Change avx512bw in isa attribute to avx512f.
12405 * config/i386/sgxintrin.h: Add license boilerplate.
12406 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12407 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12408 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12409 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12411 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12412 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12413 temporarily sse2 rather than sse if not enabled already.
12416 * config/i386/sse.md (VI248_VLBW): Rename to ...
12417 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12418 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12419 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12420 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12421 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12422 mode iterator instead of VI248_VLBW.
12424 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12426 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12427 (record_modified): Skip clobbers; add debug output.
12428 (param_change_prob): Use sreal frequencies.
12430 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12432 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12433 punt for user-aligned variables.
12435 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12437 * tree-chrec.c (chrec_contains_symbols): Return true for
12440 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12443 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12444 of (x|y) == x for BICS pattern.
12446 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12448 PR tree-optimization/83605
12449 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12450 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12453 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12455 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12456 * config/epiphany/rtems.h: New file.
12458 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12459 Uros Bizjak <ubizjak@gmail.com>
12462 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12463 QIreg_operand instead of register_operand predicate.
12464 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12465 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12466 comments instead of -fmitigate[-_]rop.
12468 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12471 * cgraphunit.c (symbol_table::compile): Switch to text_section
12472 before calling assembly_start debug hook.
12473 * run-rtl-passes.c (run_rtl_passes): Likewise.
12476 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12478 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12479 range_int_cst_p rather than !symbolic_range_p before calling
12480 extract_range_from_multiplicative_op_1.
12482 2018-01-04 Jeff Law <law@redhat.com>
12484 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12485 redundant test in assertion.
12487 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12489 * doc/rtl.texi: Document machine_mode wrapper classes.
12491 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12493 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12494 using tree_to_uhwi.
12496 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12498 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12499 the VEC_PERM_EXPR fold to fail.
12501 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12504 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12505 to switched_sections.
12507 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12510 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12511 test for d.testing.
12513 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12516 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12517 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12519 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12522 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12523 is BLKmode and bitpos not zero or mode change is needed.
12525 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12528 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12531 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12534 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12535 instead of MULT rtx. Update all corresponding splitters.
12536 (*saddl_se): Ditto.
12537 (*ssub<modesuffix>): Ditto.
12538 (*ssubl_se): Ditto.
12539 (*cmp_sadd_di): Update split patterns.
12540 (*cmp_sadd_si): Ditto.
12541 (*cmp_sadd_sidi): Ditto.
12542 (*cmp_ssub_di): Ditto.
12543 (*cmp_ssub_si): Ditto.
12544 (*cmp_ssub_sidi): Ditto.
12545 * config/alpha/predicates.md (const23_operand): New predicate.
12546 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12547 Look for ASHIFT, not MULT inner operand.
12548 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12550 2018-01-04 Martin Liska <mliska@suse.cz>
12552 PR gcov-profile/83669
12553 * gcov.c (output_intermediate_file): Add version to intermediate
12555 * doc/gcov.texi: Document new field 'version' in intermediate
12556 file format. Fix location of '-k' option of gcov command.
12558 2018-01-04 Martin Liska <mliska@suse.cz>
12561 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12563 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12565 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12567 2018-01-03 Martin Sebor <msebor@redhat.com>
12569 PR tree-optimization/83655
12570 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12571 checking calls with invalid arguments.
12573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12575 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12576 (vectorizable_mask_load_store): Delete.
12577 (vectorizable_call): Return false for masked loads and stores.
12578 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12579 instead of gimple_assign_rhs1.
12580 (vectorizable_load): Handle IFN_MASK_LOAD.
12581 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12585 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12587 (vectorizable_mask_load_store): ...here.
12588 (vectorizable_load): ...and here.
12590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12592 * tree-vect-stmts.c (vect_build_all_ones_mask)
12593 (vect_build_zero_merge_argument): New functions, split out from...
12594 (vectorizable_load): ...here.
12596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12598 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12600 (vectorizable_mask_load_store): ...here.
12601 (vectorizable_store): ...and here.
12603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12605 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12607 (vectorizable_mask_load_store): ...here.
12609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12611 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12612 (vect_model_store_cost): Take a vec_load_store_type instead of a
12614 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12615 (vect_model_store_cost): Take a vec_load_store_type instead of a
12617 (vectorizable_mask_load_store): Update accordingly.
12618 (vectorizable_store): Likewise.
12619 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12623 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12624 IFN_MASK_LOAD calls here rather than...
12625 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12628 Alan Hayward <alan.hayward@arm.com>
12629 David Sherwood <david.sherwood@arm.com>
12631 * expmed.c (extract_bit_field_1): For vector extracts,
12632 fall back to extract_bit_field_as_subreg if vec_extract
12635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12636 Alan Hayward <alan.hayward@arm.com>
12637 David Sherwood <david.sherwood@arm.com>
12639 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12640 they are variable or constant sized.
12641 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12642 slots for constant-sized data.
12644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12645 Alan Hayward <alan.hayward@arm.com>
12646 David Sherwood <david.sherwood@arm.com>
12648 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12649 handling COND_EXPRs with boolean comparisons, try to find a better
12650 basis for the mask type than the boolean itself.
12652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12654 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12655 is calculated and how it can be overridden.
12656 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12657 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12659 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12663 Alan Hayward <alan.hayward@arm.com>
12664 David Sherwood <david.sherwood@arm.com>
12666 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12667 Remove the mode argument.
12668 (aarch64_simd_valid_immediate): Remove the mode and inverse
12670 * config/aarch64/iterators.md (bitsize): New iterator.
12671 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12672 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12673 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12674 aarch64_simd_valid_immediate.
12675 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12676 (aarch64_reg_or_bic_imm): Likewise.
12677 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12678 with an insn_type enum and msl with a modifier_type enum.
12679 Replace element_width with a scalar_mode. Change the shift
12680 to unsigned int. Add constructors for scalar_float_mode and
12681 scalar_int_mode elements.
12682 (aarch64_vect_float_const_representable_p): Delete.
12683 (aarch64_can_const_movi_rtx_p)
12684 (aarch64_simd_scalar_immediate_valid_for_move)
12685 (aarch64_simd_make_constant): Update call to
12686 aarch64_simd_valid_immediate.
12687 (aarch64_advsimd_valid_immediate_hs): New function.
12688 (aarch64_advsimd_valid_immediate): Likewise.
12689 (aarch64_simd_valid_immediate): Remove mode and inverse
12690 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12691 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12692 and aarch64_float_const_representable_p on the result.
12693 (aarch64_output_simd_mov_immediate): Remove mode argument.
12694 Update call to aarch64_simd_valid_immediate and use of
12695 simd_immediate_info.
12696 (aarch64_output_scalar_simd_mov_immediate): Update call
12699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12700 Alan Hayward <alan.hayward@arm.com>
12701 David Sherwood <david.sherwood@arm.com>
12703 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12704 (mode_nunits): Likewise CONST_MODE_NUNITS.
12705 * machmode.def (ADJUST_NUNITS): Document.
12706 * genmodes.c (mode_data::need_nunits_adj): New field.
12707 (blank_mode): Update accordingly.
12708 (adj_nunits): New variable.
12709 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12711 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12712 listed in adj_nunits.
12713 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12714 listed in adj_nunits. Don't emit case statements for such modes.
12715 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12716 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12717 nothing if adj_nunits is nonnull.
12718 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12719 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12720 (emit_mode_fbit): Update use of print_maybe_const_decl.
12721 (emit_move_size): Likewise. Treat the array as non-const
12723 (emit_mode_adjustments): Handle adj_nunits.
12725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12727 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12728 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12729 (VECTOR_MODES): Use it.
12730 (make_vector_modes): Take the prefix as an argument.
12732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12733 Alan Hayward <alan.hayward@arm.com>
12734 David Sherwood <david.sherwood@arm.com>
12736 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12737 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12738 for MODE_VECTOR_BOOL.
12739 * machmode.def (VECTOR_BOOL_MODE): Document.
12740 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12741 (make_vector_bool_mode): New function.
12742 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12744 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12745 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12747 * stor-layout.c (int_mode_for_mode): Likewise.
12748 * tree.c (build_vector_type_for_mode): Likewise.
12749 * varasm.c (output_constant_pool_2): Likewise.
12750 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12751 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12752 for MODE_VECTOR_BOOL.
12753 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12754 of mode class checks.
12755 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12756 instead of a list of mode class checks.
12757 (expand_vector_scalar_condition): Likewise.
12758 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12761 Alan Hayward <alan.hayward@arm.com>
12762 David Sherwood <david.sherwood@arm.com>
12764 * machmode.h (mode_size): Change from unsigned short to
12766 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12767 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12768 or if measurement_type is not polynomial.
12769 (fixed_size_mode::includes_p): Check for constant-sized modes.
12770 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12771 return a poly_uint16 rather than an unsigned short.
12772 (emit_mode_size): Change the type of mode_size from unsigned short
12773 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12774 (emit_mode_adjustments): Cope with polynomial vector sizes.
12775 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12777 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12779 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12780 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12781 * caller-save.c (setup_save_areas): Likewise.
12782 (replace_reg_with_saved_mem): Likewise.
12783 * calls.c (emit_library_call_value_1): Likewise.
12784 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12785 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12786 (gen_lowpart_for_combine): Likewise.
12787 * convert.c (convert_to_integer_1): Likewise.
12788 * cse.c (equiv_constant, cse_insn): Likewise.
12789 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12790 (cselib_subst_to_values): Likewise.
12791 * dce.c (word_dce_process_block): Likewise.
12792 * df-problems.c (df_word_lr_mark_ref): Likewise.
12793 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12794 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12795 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12796 (rtl_for_decl_location): Likewise.
12797 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12798 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12799 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12800 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12801 (expand_expr_real_1): Likewise.
12802 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12803 (pad_below): Likewise.
12804 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12805 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12806 * ira.c (get_subreg_tracking_sizes): Likewise.
12807 * ira-build.c (ira_create_allocno_objects): Likewise.
12808 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12809 (ira_sort_regnos_for_alter_reg): Likewise.
12810 * ira-costs.c (record_operand_costs): Likewise.
12811 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12812 (resolve_simple_move): Likewise.
12813 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12814 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12815 (lra_constraints): Likewise.
12816 (CONST_POOL_OK_P): Reject variable-sized modes.
12817 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12818 (add_pseudo_to_slot, lra_spill): Likewise.
12819 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12820 * optabs-query.c (get_best_extraction_insn): Likewise.
12821 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12822 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12823 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12824 * recog.c (offsettable_address_addr_space_p): Likewise.
12825 * regcprop.c (maybe_mode_change): Likewise.
12826 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12827 * regrename.c (build_def_use): Likewise.
12828 * regstat.c (dump_reg_info): Likewise.
12829 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12830 (find_reloads, find_reloads_subreg_address): Likewise.
12831 * reload1.c (eliminate_regs_1): Likewise.
12832 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12833 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12834 (simplify_binary_operation_1, simplify_subreg): Likewise.
12835 * targhooks.c (default_function_arg_padding): Likewise.
12836 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12837 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12838 (verify_gimple_assign_ternary): Likewise.
12839 * tree-inline.c (estimate_move_cost): Likewise.
12840 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12841 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12842 (get_address_cost_ainc): Likewise.
12843 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12844 (vect_supportable_dr_alignment): Likewise.
12845 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12846 (vectorizable_reduction): Likewise.
12847 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12848 (vectorizable_operation, vectorizable_load): Likewise.
12849 * tree.c (build_same_sized_truth_vector_type): Likewise.
12850 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12851 * var-tracking.c (emit_note_insn_var_location): Likewise.
12852 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12853 (ADDR_VEC_ALIGN): Likewise.
12855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12856 Alan Hayward <alan.hayward@arm.com>
12857 David Sherwood <david.sherwood@arm.com>
12859 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12861 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12862 or if measurement_type is polynomial.
12863 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12864 * combine.c (make_extraction): Likewise.
12865 * dse.c (find_shift_sequence): Likewise.
12866 * dwarf2out.c (mem_loc_descriptor): Likewise.
12867 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12868 (extract_bit_field, extract_low_bits): Likewise.
12869 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12870 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12871 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12872 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12873 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12874 * reload.c (find_reloads): Likewise.
12875 * reload1.c (alter_reg): Likewise.
12876 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12877 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12878 * tree-if-conv.c (predicate_mem_writes): Likewise.
12879 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12880 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12881 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12882 * valtrack.c (dead_debug_insert_temp): Likewise.
12883 * varasm.c (mergeable_constant_section): Likewise.
12884 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12887 Alan Hayward <alan.hayward@arm.com>
12888 David Sherwood <david.sherwood@arm.com>
12890 * expr.c (expand_assignment): Cope with polynomial mode sizes
12891 when assigning to a CONCAT.
12893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12894 Alan Hayward <alan.hayward@arm.com>
12895 David Sherwood <david.sherwood@arm.com>
12897 * machmode.h (mode_precision): Change from unsigned short to
12899 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12901 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12902 or if measurement_type is not polynomial.
12903 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12904 in which the mode is already known to be a scalar_int_mode.
12905 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12906 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12908 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12909 for GET_MODE_PRECISION.
12910 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12911 for GET_MODE_PRECISION.
12912 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12914 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12915 (expand_field_assignment, make_extraction): Likewise.
12916 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12917 (get_last_value): Likewise.
12918 * convert.c (convert_to_integer_1): Likewise.
12919 * cse.c (cse_insn): Likewise.
12920 * expr.c (expand_expr_real_1): Likewise.
12921 * lra-constraints.c (simplify_operand_subreg): Likewise.
12922 * optabs-query.c (can_atomic_load_p): Likewise.
12923 * optabs.c (expand_atomic_load): Likewise.
12924 (expand_atomic_store): Likewise.
12925 * ree.c (combine_reaching_defs): Likewise.
12926 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12927 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12928 * tree.h (type_has_mode_precision_p): Likewise.
12929 * ubsan.c (instrument_si_overflow): Likewise.
12931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12932 Alan Hayward <alan.hayward@arm.com>
12933 David Sherwood <david.sherwood@arm.com>
12935 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12936 polynomial numbers of units.
12937 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12938 (valid_vector_subparts_p): New function.
12939 (build_vector_type): Remove temporary shim and take the number
12940 of units as a poly_uint64 rather than an int.
12941 (build_opaque_vector_type): Take the number of units as a
12942 poly_uint64 rather than an int.
12943 * tree.c (build_vector_from_ctor): Handle polynomial
12944 TYPE_VECTOR_SUBPARTS.
12945 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12946 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12947 (build_vector_from_val): If the number of units is variable,
12948 use build_vec_duplicate_cst for constant operands and
12949 VEC_DUPLICATE_EXPR otherwise.
12950 (make_vector_type): Remove temporary is_constant ().
12951 (build_vector_type, build_opaque_vector_type): Take the number of
12952 units as a poly_uint64 rather than an int.
12953 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12955 * cfgexpand.c (expand_debug_expr): Likewise.
12956 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12957 (store_constructor, expand_expr_real_1): Likewise.
12958 (const_scalar_mask_from_tree): Likewise.
12959 * fold-const-call.c (fold_const_reduction): Likewise.
12960 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12961 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12962 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12963 (fold_relational_const): Likewise.
12964 (native_interpret_vector): Likewise. Change the size from an
12965 int to an unsigned int.
12966 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12967 TYPE_VECTOR_SUBPARTS.
12968 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12969 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12970 duplicating a non-constant operand into a variable-length vector.
12971 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12972 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12973 * ipa-icf.c (sem_variable::equals): Likewise.
12974 * match.pd: Likewise.
12975 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12976 * print-tree.c (print_node): Likewise.
12977 * stor-layout.c (layout_type): Likewise.
12978 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12979 * tree-cfg.c (verify_gimple_comparison): Likewise.
12980 (verify_gimple_assign_binary): Likewise.
12981 (verify_gimple_assign_ternary): Likewise.
12982 (verify_gimple_assign_single): Likewise.
12983 * tree-pretty-print.c (dump_generic_node): Likewise.
12984 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12985 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12986 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12987 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12988 (vect_shift_permute_load_chain): Likewise.
12989 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12990 (expand_vector_condition, optimize_vector_constructor): Likewise.
12991 (lower_vec_perm, get_compute_type): Likewise.
12992 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12993 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12994 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12995 (vect_recog_mask_conversion_pattern): Likewise.
12996 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12997 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12998 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12999 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13000 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13001 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13002 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13003 (supportable_widening_operation): Likewise.
13004 (supportable_narrowing_operation): Likewise.
13005 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13007 * varasm.c (output_constant): Likewise.
13009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13010 Alan Hayward <alan.hayward@arm.com>
13011 David Sherwood <david.sherwood@arm.com>
13013 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13014 so that both the length == 3 and length != 3 cases set up their
13015 own permute vectors. Add comments explaining why we know the
13016 number of elements is constant.
13017 (vect_permute_load_chain): Likewise.
13019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13020 Alan Hayward <alan.hayward@arm.com>
13021 David Sherwood <david.sherwood@arm.com>
13023 * machmode.h (mode_nunits): Change from unsigned char to
13025 (ONLY_FIXED_SIZE_MODES): New macro.
13026 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13027 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13028 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13030 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13031 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13032 or if measurement_type is not polynomial.
13033 * genmodes.c (ZERO_COEFFS): New macro.
13034 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13036 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13037 Use ZERO_COEFFS when emitting initializers.
13038 * data-streamer.h (bp_pack_poly_value): New function.
13039 (bp_unpack_poly_value): Likewise.
13040 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13041 for GET_MODE_NUNITS.
13042 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13043 for GET_MODE_NUNITS.
13044 * tree.c (make_vector_type): Remove temporary shim and make
13045 the real function take the number of units as a poly_uint64
13046 rather than an int.
13047 (build_vector_type_for_mode): Handle polynomial nunits.
13048 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13049 * emit-rtl.c (const_vec_series_p_1): Likewise.
13050 (gen_rtx_CONST_VECTOR): Likewise.
13051 * fold-const.c (test_vec_duplicate_folding): Likewise.
13052 * genrecog.c (validate_pattern): Likewise.
13053 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13054 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13055 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13056 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13057 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13058 * rtlanal.c (subreg_get_info): Likewise.
13059 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13060 (vect_grouped_load_supported): Likewise.
13061 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13062 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13063 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13064 (simplify_const_unary_operation, simplify_binary_operation_1)
13065 (simplify_const_binary_operation, simplify_ternary_operation)
13066 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13067 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13068 instead of CONST_VECTOR_NUNITS.
13069 * varasm.c (output_constant_pool_2): Likewise.
13070 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13071 explicit-encoded elements in the XVEC for variable-length vectors.
13073 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13075 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13078 Alan Hayward <alan.hayward@arm.com>
13079 David Sherwood <david.sherwood@arm.com>
13081 * coretypes.h (fixed_size_mode): Declare.
13082 (fixed_size_mode_pod): New typedef.
13083 * builtins.h (target_builtins::x_apply_args_mode)
13084 (target_builtins::x_apply_result_mode): Change type to
13085 fixed_size_mode_pod.
13086 * builtins.c (apply_args_size, apply_result_size, result_vector)
13087 (expand_builtin_apply_args_1, expand_builtin_apply)
13088 (expand_builtin_return): Update accordingly.
13090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13092 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13093 * cselib.c (cselib_hash_rtx): Likewise.
13094 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13095 CONST_VECTOR encoding.
13097 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13098 Jeff Law <law@redhat.com>
13101 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13102 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13103 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13104 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13107 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13108 explicitly probe *sp in a noreturn function if there were any callee
13109 register saves or frame pointer is needed.
13111 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13114 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13115 BLKmode for ternary, binary or unary expressions.
13118 * var-tracking.c (delete_vta_debug_insn): New inline function.
13119 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13120 insns from get_insns () to NULL instead of each bb separately.
13121 Use delete_vta_debug_insn. No longer static.
13122 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13123 delete_vta_debug_insns callers.
13124 * rtl.h (delete_vta_debug_insns): Declare.
13125 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13126 instead of variable_tracking_main.
13128 2018-01-03 Martin Sebor <msebor@redhat.com>
13130 PR tree-optimization/83603
13131 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13132 arguments past the endof the argument list in functions declared
13133 without a prototype.
13134 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13135 Avoid checking when arguments are null.
13137 2018-01-03 Martin Sebor <msebor@redhat.com>
13140 * doc/extend.texi (attribute const): Fix a typo.
13141 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13142 issuing -Wsuggest-attribute for void functions.
13144 2018-01-03 Martin Sebor <msebor@redhat.com>
13146 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13147 offset_int::from instead of wide_int::to_shwi.
13148 (maybe_diag_overlap): Remove assertion.
13149 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13150 * gimple-ssa-sprintf.c (format_directive): Same.
13151 (parse_directive): Same.
13152 (sprintf_dom_walker::compute_format_length): Same.
13153 (try_substitute_return_value): Same.
13155 2018-01-03 Jeff Law <law@redhat.com>
13157 PR middle-end/83654
13158 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13159 non-constant residual for zero at runtime and avoid probing in
13160 that case. Reorganize code for trailing problem to mirror handling
13163 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13165 PR tree-optimization/83501
13166 * tree-ssa-strlen.c (get_string_cst): New.
13167 (handle_char_store): Call get_string_cst.
13169 2018-01-03 Martin Liska <mliska@suse.cz>
13171 PR tree-optimization/83593
13172 * tree-ssa-strlen.c: Include tree-cfg.h.
13173 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13174 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13175 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13177 (strlen_dom_walker::before_dom_children): Call
13178 gimple_purge_dead_eh_edges. Dump tranformation with details
13180 (strlen_dom_walker::before_dom_children): Update call by adding
13181 new argument cleanup_eh.
13182 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13184 2018-01-03 Martin Liska <mliska@suse.cz>
13187 * cif-code.def (VARIADIC_THUNK): New enum value.
13188 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13191 2018-01-03 Jan Beulich <jbeulich@suse.com>
13193 * sse.md (mov<mode>_internal): Tighten condition for when to use
13194 vmovdqu<ssescalarsize> for TI and OI modes.
13196 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13198 Update copyright years.
13200 2018-01-03 Martin Liska <mliska@suse.cz>
13203 * ipa-visibility.c (function_and_variable_visibility): Skip
13204 functions with noipa attribure.
13206 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13208 * gcc.c (process_command): Update copyright notice dates.
13209 * gcov-dump.c (print_version): Ditto.
13210 * gcov.c (print_version): Ditto.
13211 * gcov-tool.c (print_version): Ditto.
13212 * gengtype.c (create_file): Ditto.
13213 * doc/cpp.texi: Bump @copying's copyright year.
13214 * doc/cppinternals.texi: Ditto.
13215 * doc/gcc.texi: Ditto.
13216 * doc/gccint.texi: Ditto.
13217 * doc/gcov.texi: Ditto.
13218 * doc/install.texi: Ditto.
13219 * doc/invoke.texi: Ditto.
13221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13223 * vector-builder.h (vector_builder::m_full_nelts): Change from
13224 unsigned int to poly_uint64.
13225 (vector_builder::full_nelts): Update prototype accordingly.
13226 (vector_builder::new_vector): Likewise.
13227 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13228 (vector_builder::operator ==): Likewise.
13229 (vector_builder::finalize): Likewise.
13230 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13231 Take the number of elements as a poly_uint64 rather than an
13233 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13234 from unsigned int to poly_uint64.
13235 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13236 (vec_perm_indices::new_vector): Likewise.
13237 (vec_perm_indices::length): Likewise.
13238 (vec_perm_indices::nelts_per_input): Likewise.
13239 (vec_perm_indices::input_nelts): Likewise.
13240 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13241 number of elements per input as a poly_uint64 rather than an
13242 unsigned int. Use the original encoding for variable-length
13243 vectors, rather than clamping each individual element.
13244 For the second and subsequent elements in each pattern,
13245 clamp the step and base before clamping their sum.
13246 (vec_perm_indices::series_p): Handle polynomial element counts.
13247 (vec_perm_indices::all_in_range_p): Likewise.
13248 (vec_perm_indices_to_tree): Likewise.
13249 (vec_perm_indices_to_rtx): Likewise.
13250 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13251 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13252 (tree_vector_builder::new_binary_operation): Handle polynomial
13253 element counts. Return false if we need to know the number
13254 of elements at compile time.
13255 * fold-const.c (fold_vec_perm): Punt if the number of elements
13256 isn't known at compile time.
13258 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13260 * vec-perm-indices.h (vec_perm_builder): Change element type
13261 from HOST_WIDE_INT to poly_int64.
13262 (vec_perm_indices::element_type): Update accordingly.
13263 (vec_perm_indices::clamp): Handle polynomial element_types.
13264 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13265 (vec_perm_indices::all_in_range_p): Likewise.
13266 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13268 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13269 polynomial vec_perm_indices element types.
13270 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13271 * fold-const.c (fold_vec_perm): Likewise.
13272 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13273 * tree-vect-generic.c (lower_vec_perm): Likewise.
13274 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13275 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13276 element type to HOST_WIDE_INT.
13278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13279 Alan Hayward <alan.hayward@arm.com>
13280 David Sherwood <david.sherwood@arm.com>
13282 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13283 rather than an int. Use plus_constant.
13284 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13285 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13288 Alan Hayward <alan.hayward@arm.com>
13289 David Sherwood <david.sherwood@arm.com>
13291 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13292 a HOST_WIDE_INT to a poly_int64.
13294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13295 Alan Hayward <alan.hayward@arm.com>
13296 David Sherwood <david.sherwood@arm.com>
13298 * calls.c (load_register_parameters): Cope with polynomial
13299 mode sizes. Require a constant size for BLKmode parameters
13300 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13301 forces a parameter to be padded at the lsb end in order to
13302 fill a complete number of words, require the parameter size
13303 to be ordered wrt UNITS_PER_WORD.
13305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13306 Alan Hayward <alan.hayward@arm.com>
13307 David Sherwood <david.sherwood@arm.com>
13309 * reload1.c (spill_stack_slot_width): Change element type
13310 from unsigned int to poly_uint64_pod.
13311 (alter_reg): Treat mode sizes as polynomial.
13313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13314 Alan Hayward <alan.hayward@arm.com>
13315 David Sherwood <david.sherwood@arm.com>
13317 * reload.c (complex_word_subreg_p): New function.
13318 (reload_inner_reg_of_subreg, push_reload): Use it.
13320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13321 Alan Hayward <alan.hayward@arm.com>
13322 David Sherwood <david.sherwood@arm.com>
13324 * lra-constraints.c (process_alt_operands): Reject matched
13325 operands whose sizes aren't ordered.
13326 (match_reload): Refer to this check here.
13328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13329 Alan Hayward <alan.hayward@arm.com>
13330 David Sherwood <david.sherwood@arm.com>
13332 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13333 that the mode size is in the set {1, 2, 4, 8, 16}.
13335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13336 Alan Hayward <alan.hayward@arm.com>
13337 David Sherwood <david.sherwood@arm.com>
13339 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13340 Use plus_constant instead of gen_rtx_PLUS.
13342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13343 Alan Hayward <alan.hayward@arm.com>
13344 David Sherwood <david.sherwood@arm.com>
13346 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13347 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13348 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13349 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13350 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13351 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13352 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13353 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13354 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13355 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13357 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13358 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13359 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13360 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13361 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13362 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13363 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13364 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13365 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13366 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13368 * expr.c (emit_move_resolve_push): Treat the input and result
13369 of PUSH_ROUNDING as a poly_int64.
13370 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13371 (emit_push_insn): Likewise.
13372 * lra-eliminations.c (mark_not_eliminable): Likewise.
13373 * recog.c (push_operand): Likewise.
13374 * reload1.c (elimination_effects): Likewise.
13375 * rtlanal.c (nonzero_bits1): Likewise.
13376 * calls.c (store_one_arg): Likewise. Require the padding to be
13377 known at compile time.
13379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13380 Alan Hayward <alan.hayward@arm.com>
13381 David Sherwood <david.sherwood@arm.com>
13383 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13384 Use plus_constant instead of gen_rtx_PLUS.
13386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13387 Alan Hayward <alan.hayward@arm.com>
13388 David Sherwood <david.sherwood@arm.com>
13390 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13391 rather than an int.
13393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13394 Alan Hayward <alan.hayward@arm.com>
13395 David Sherwood <david.sherwood@arm.com>
13397 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13398 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13399 via stack temporaries. Treat the mode size as polynomial too.
13401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13402 Alan Hayward <alan.hayward@arm.com>
13403 David Sherwood <david.sherwood@arm.com>
13405 * expr.c (expand_expr_real_2): When handling conversions involving
13406 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13407 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13408 as a poly_uint64 too.
13410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13411 Alan Hayward <alan.hayward@arm.com>
13412 David Sherwood <david.sherwood@arm.com>
13414 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13417 Alan Hayward <alan.hayward@arm.com>
13418 David Sherwood <david.sherwood@arm.com>
13420 * combine.c (can_change_dest_mode): Handle polynomial
13421 REGMODE_NATURAL_SIZE.
13422 * expmed.c (store_bit_field_1): Likewise.
13423 * expr.c (store_constructor): Likewise.
13424 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13425 and polynomial REGMODE_NATURAL_SIZE.
13426 (gen_lowpart_common): Likewise.
13427 * reginfo.c (record_subregs_of_mode): Likewise.
13428 * rtlanal.c (read_modify_subreg_p): Likewise.
13430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13431 Alan Hayward <alan.hayward@arm.com>
13432 David Sherwood <david.sherwood@arm.com>
13434 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13435 numbers of elements.
13437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13438 Alan Hayward <alan.hayward@arm.com>
13439 David Sherwood <david.sherwood@arm.com>
13441 * match.pd: Cope with polynomial numbers of vector elements.
13443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13444 Alan Hayward <alan.hayward@arm.com>
13445 David Sherwood <david.sherwood@arm.com>
13447 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13448 in a POINTER_PLUS_EXPR.
13450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13451 Alan Hayward <alan.hayward@arm.com>
13452 David Sherwood <david.sherwood@arm.com>
13454 * omp-simd-clone.c (simd_clone_subparts): New function.
13455 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13456 (ipa_simd_modify_function_body): Likewise.
13458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13459 Alan Hayward <alan.hayward@arm.com>
13460 David Sherwood <david.sherwood@arm.com>
13462 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13463 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13464 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13465 (expand_vector_condition, vector_element): Likewise.
13466 (subparts_gt): New function.
13467 (get_compute_type): Use subparts_gt.
13468 (count_type_subparts): Delete.
13469 (expand_vector_operations_1): Use subparts_gt instead of
13470 count_type_subparts.
13472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13473 Alan Hayward <alan.hayward@arm.com>
13474 David Sherwood <david.sherwood@arm.com>
13476 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13477 (vect_compile_time_alias): ...this new function. Do the calculation
13478 on poly_ints rather than trees.
13479 (vect_prune_runtime_alias_test_list): Update call accordingly.
13481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13482 Alan Hayward <alan.hayward@arm.com>
13483 David Sherwood <david.sherwood@arm.com>
13485 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13487 (vect_schedule_slp_instance): Likewise.
13489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13490 Alan Hayward <alan.hayward@arm.com>
13491 David Sherwood <david.sherwood@arm.com>
13493 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13494 constant and extern definitions for variable-length vectors.
13495 (vect_get_constant_vectors): Note that the number of units
13496 is known to be constant.
13498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13499 Alan Hayward <alan.hayward@arm.com>
13500 David Sherwood <david.sherwood@arm.com>
13502 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13503 of units as polynomial. Choose between WIDE and NARROW based
13506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13507 Alan Hayward <alan.hayward@arm.com>
13508 David Sherwood <david.sherwood@arm.com>
13510 * tree-vect-stmts.c (simd_clone_subparts): New function.
13511 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13514 Alan Hayward <alan.hayward@arm.com>
13515 David Sherwood <david.sherwood@arm.com>
13517 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13518 vectors as polynomial. Use build_index_vector for
13519 IFN_GOMP_SIMD_LANE.
13521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13522 Alan Hayward <alan.hayward@arm.com>
13523 David Sherwood <david.sherwood@arm.com>
13525 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13526 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13527 for variable-length vectors.
13528 (vectorizable_mask_load_store): Treat the number of units as
13529 polynomial, asserting that it is constant if the condition has
13530 already been enforced.
13531 (vectorizable_store, vectorizable_load): Likewise.
13533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13534 Alan Hayward <alan.hayward@arm.com>
13535 David Sherwood <david.sherwood@arm.com>
13537 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13538 of units as polynomial. Punt if we can't tell at compile time
13539 which vector contains the final result.
13541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13542 Alan Hayward <alan.hayward@arm.com>
13543 David Sherwood <david.sherwood@arm.com>
13545 * tree-vect-loop.c (vectorizable_induction): Treat the number
13546 of units as polynomial. Punt on SLP inductions. Use an integer
13547 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13548 cast of such a series for variable-length floating-point
13551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13552 Alan Hayward <alan.hayward@arm.com>
13553 David Sherwood <david.sherwood@arm.com>
13555 * tree.h (build_index_vector): Declare.
13556 * tree.c (build_index_vector): New function.
13557 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13558 of units as polynomial, forcibly converting it to a constant if
13559 vectorizable_reduction has already enforced the condition.
13560 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13561 to create a {1,2,3,...} vector.
13562 (vectorizable_reduction): Treat the number of units as polynomial.
13563 Choose vectype_in based on the largest scalar element size rather
13564 than the smallest number of units. Enforce the restrictions
13567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13568 Alan Hayward <alan.hayward@arm.com>
13569 David Sherwood <david.sherwood@arm.com>
13571 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13572 number of units as polynomial.
13574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13575 Alan Hayward <alan.hayward@arm.com>
13576 David Sherwood <david.sherwood@arm.com>
13578 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13579 * target.def (autovectorize_vector_sizes): Return the vector sizes
13580 by pointer, using vector_sizes rather than a bitmask.
13581 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13582 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13583 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13585 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13586 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13587 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13588 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13589 * omp-general.c (omp_max_vf): Likewise.
13590 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13591 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13592 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13593 * tree-vect-slp.c (vect_slp_bb): Likewise.
13594 * doc/tm.texi: Regenerate.
13595 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13597 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13598 the vector size as a poly_uint64 rather than an unsigned int.
13599 (current_vector_size): Change from an unsigned int to a poly_uint64.
13600 (get_vectype_for_scalar_type): Update accordingly.
13601 * tree.h (build_truth_vector_type): Take the size and number of
13602 units as a poly_uint64 rather than an unsigned int.
13603 (build_vector_type): Add a temporary overload that takes
13604 the number of units as a poly_uint64 rather than an unsigned int.
13605 * tree.c (make_vector_type): Likewise.
13606 (build_truth_vector_type): Take the number of units as a poly_uint64
13607 rather than an unsigned int.
13609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13610 Alan Hayward <alan.hayward@arm.com>
13611 David Sherwood <david.sherwood@arm.com>
13613 * target.def (get_mask_mode): Take the number of units and length
13614 as poly_uint64s rather than unsigned ints.
13615 * targhooks.h (default_get_mask_mode): Update accordingly.
13616 * targhooks.c (default_get_mask_mode): Likewise.
13617 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13618 * doc/tm.texi: Regenerate.
13620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13621 Alan Hayward <alan.hayward@arm.com>
13622 David Sherwood <david.sherwood@arm.com>
13624 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13625 * omp-general.c (omp_max_vf): Likewise.
13626 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13627 (expand_omp_simd): Handle polynomial safelen.
13628 * omp-low.c (omplow_simd_context): Add a default constructor.
13629 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13630 (lower_rec_simd_input_clauses): Update accordingly.
13631 (lower_rec_input_clauses): Likewise.
13633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13634 Alan Hayward <alan.hayward@arm.com>
13635 David Sherwood <david.sherwood@arm.com>
13637 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13638 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13639 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13640 (vect_analyze_slp_cost): Likewise.
13641 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13642 (vect_model_load_cost): Likewise.
13644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13645 Alan Hayward <alan.hayward@arm.com>
13646 David Sherwood <david.sherwood@arm.com>
13648 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13649 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13650 from an unsigned int * to a poly_uint64_pod *.
13651 (calculate_unrolling_factor): New function.
13652 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13655 Alan Hayward <alan.hayward@arm.com>
13656 David Sherwood <david.sherwood@arm.com>
13658 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13659 from an unsigned int to a poly_uint64.
13660 (_loop_vec_info::slp_unrolling_factor): Likewise.
13661 (_loop_vec_info::vectorization_factor): Change from an int
13663 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13664 (vect_get_num_vectors): New function.
13665 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13666 (vect_get_num_copies): Use vect_get_num_vectors.
13667 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13668 to an unsigned int *.
13669 (vect_analyze_data_refs): Change min_vf from an int * to a
13671 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13672 than an unsigned HOST_WIDE_INT.
13673 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13674 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13675 to an unsigned int *.
13676 (vect_analyze_data_ref_dependences): Likewise.
13677 (vect_compute_data_ref_alignment): Handle polynomial vf.
13678 (vect_enhance_data_refs_alignment): Likewise.
13679 (vect_prune_runtime_alias_test_list): Likewise.
13680 (vect_shift_permute_load_chain): Likewise.
13681 (vect_supportable_dr_alignment): Likewise.
13682 (dependence_distance_ge_vf): Take the vectorization factor as a
13683 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13684 (vect_analyze_data_refs): Change min_vf from an int * to a
13686 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13687 vfm1 as a poly_uint64 rather than an int. Make the same change
13688 for the returned bound_scalar.
13689 (vect_gen_vector_loop_niters): Handle polynomial vf.
13690 (vect_do_peeling): Likewise. Update call to
13691 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13692 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13694 * tree-vect-loop.c (vect_determine_vectorization_factor)
13695 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13696 (vect_get_known_peeling_cost): Likewise.
13697 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13698 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13699 (vect_transform_loop): Likewise. Use the lowest possible VF when
13700 updating the upper bounds of the loop.
13701 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13702 rather than an int.
13703 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13704 polynomial unroll factors.
13705 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13706 (vect_make_slp_decision): Likewise.
13707 (vect_supported_load_permutation_p): Likewise, and polynomial
13709 (vect_analyze_slp_cost): Handle polynomial vf.
13710 (vect_slp_analyze_node_operations): Likewise.
13711 (vect_slp_analyze_bb_1): Likewise.
13712 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13713 than an unsigned HOST_WIDE_INT.
13714 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13715 (vectorizable_load): Handle polynomial vf.
13716 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13718 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13721 Alan Hayward <alan.hayward@arm.com>
13722 David Sherwood <david.sherwood@arm.com>
13724 * match.pd: Handle bit operations involving three constants
13725 and try to fold one pair.
13727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13729 * tree-vect-loop-manip.c: Include gimple-fold.h.
13730 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13731 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13732 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13733 Add a path that uses a step of VF instead of 1, but disable it
13735 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13736 and niters_no_overflow parameters. Update calls to
13737 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13738 Create a new SSA name if the latter choses to use a ste other
13739 than zero, and return it via niters_vector_mult_vf_var.
13740 * tree-vect-loop.c (vect_transform_loop): Update calls to
13741 vect_do_peeling, vect_gen_vector_loop_niters and
13742 slpeel_make_loop_iterate_ntimes.
13743 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13744 (vect_gen_vector_loop_niters): Update declarations after above changes.
13746 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13748 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13749 128-bit round to integer instructions.
13750 (ceil<mode>2): Likewise.
13751 (btrunc<mode>2): Likewise.
13752 (round<mode>2): Likewise.
13754 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13756 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13757 unaligned VSX load/store on P8/P9.
13758 (expand_block_clear): Allow the use of unaligned VSX
13759 load/store on P8/P9.
13761 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13763 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13765 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13766 swap associated with both a load and a store.
13768 2018-01-02 Andrew Waterman <andrew@sifive.com>
13770 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13771 * config/riscv/riscv.md (clear_cache): Use it.
13773 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13775 * web.c: Remove out-of-date comment.
13777 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13779 * expr.c (fixup_args_size_notes): Check that any existing
13780 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13781 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13782 (emit_single_push_insn): ...here.
13784 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13786 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13787 (const_vector_encoded_nelts): New function.
13788 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13789 (const_vector_int_elt, const_vector_elt): Declare.
13790 * emit-rtl.c (const_vector_int_elt_1): New function.
13791 (const_vector_elt): Likewise.
13792 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13793 of CONST_VECTOR_ELT.
13795 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13797 * expr.c: Include rtx-vector-builder.h.
13798 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13799 directly on the tree encoding.
13800 (const_vector_from_tree): Likewise.
13801 * optabs.c: Include rtx-vector-builder.h.
13802 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13803 sequence of "u" values.
13804 * vec-perm-indices.c: Include rtx-vector-builder.h.
13805 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13806 directly on the vec_perm_indices encoding.
13808 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13810 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13811 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13812 * rtx-vector-builder.h: New file.
13813 * rtx-vector-builder.c: Likewise.
13814 * rtl.h (rtx_def::u2): Add a const_vector field.
13815 (CONST_VECTOR_NPATTERNS): New macro.
13816 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13817 (CONST_VECTOR_DUPLICATE_P): Likewise.
13818 (CONST_VECTOR_STEPPED_P): Likewise.
13819 (CONST_VECTOR_ENCODED_ELT): Likewise.
13820 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13821 (unwrap_const_vec_duplicate): Likewise.
13822 (const_vec_series_p): Check for a non-duplicated vector encoding.
13823 Say that the function only returns true for integer vectors.
13824 * emit-rtl.c: Include rtx-vector-builder.h.
13825 (gen_const_vec_duplicate_1): Delete.
13826 (gen_const_vector): Call gen_const_vec_duplicate instead of
13827 gen_const_vec_duplicate_1.
13828 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13829 (gen_const_vec_duplicate): Use rtx_vector_builder.
13830 (gen_const_vec_series): Likewise.
13831 (gen_rtx_CONST_VECTOR): Likewise.
13832 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13833 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13834 Build a new vector rather than modifying a CONST_VECTOR in-place.
13835 (handle_special_swappables): Update call accordingly.
13836 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13837 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13838 Build a new vector rather than modifying a CONST_VECTOR in-place.
13839 (handle_special_swappables): Update call accordingly.
13841 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13843 * simplify-rtx.c (simplify_const_binary_operation): Use
13844 CONST_VECTOR_ELT instead of XVECEXP.
13846 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13848 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13849 the selector elements to be different from the data elements
13850 if the selector is a VECTOR_CST.
13851 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13852 ssizetype for the selector.
13854 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13856 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13857 before testing each element individually.
13858 * tree-vect-generic.c (lower_vec_perm): Likewise.
13860 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13862 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13863 * selftest-run-tests.c (selftest::run_tests): Call it.
13864 * vector-builder.h (vector_builder::operator ==): New function.
13865 (vector_builder::operator !=): Likewise.
13866 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13867 (vec_perm_indices::all_from_input_p): New function.
13868 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13869 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13870 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13871 instead of reading the VECTOR_CST directly. Detect whether both
13872 vector inputs are the same before constructing the vec_perm_indices,
13873 and update the number of inputs argument accordingly. Use the
13874 utility functions added above. Only construct sel2 if we need to.
13876 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13878 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13879 the broadcast of the low byte.
13880 (expand_mult_highpart): Use an explicit encoding for the permutes.
13881 * optabs-query.c (can_mult_highpart_p): Likewise.
13882 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13883 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13884 (vectorizable_bswap): Likewise.
13885 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13886 explicit encoding for the power-of-2 permutes.
13887 (vect_permute_store_chain): Likewise.
13888 (vect_grouped_load_supported): Likewise.
13889 (vect_permute_load_chain): Likewise.
13891 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13893 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13894 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13895 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13896 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13897 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13898 (vect_gen_perm_mask_any): Likewise.
13900 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13902 * int-vector-builder.h: New file.
13903 * vec-perm-indices.h: Include int-vector-builder.h.
13904 (vec_perm_indices): Redefine as an int_vector_builder.
13905 (auto_vec_perm_indices): Delete.
13906 (vec_perm_builder): Redefine as a stand-alone class.
13907 (vec_perm_indices::vec_perm_indices): New function.
13908 (vec_perm_indices::clamp): Likewise.
13909 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13910 (vec_perm_indices::new_vector): New function.
13911 (vec_perm_indices::new_expanded_vector): Update for new
13912 vec_perm_indices class.
13913 (vec_perm_indices::rotate_inputs): New function.
13914 (vec_perm_indices::all_in_range_p): Operate directly on the
13915 encoded form, without computing elided elements.
13916 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13917 encoding. Update for new vec_perm_indices class.
13918 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13919 the given vec_perm_builder.
13920 (expand_vec_perm_var): Update vec_perm_builder constructor.
13921 (expand_mult_highpart): Use vec_perm_builder instead of
13922 auto_vec_perm_indices.
13923 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13924 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13925 or double series encoding as appropriate.
13926 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13927 vec_perm_indices instead of auto_vec_perm_indices.
13928 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13929 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13930 (vect_permute_store_chain): Likewise.
13931 (vect_grouped_load_supported): Likewise.
13932 (vect_permute_load_chain): Likewise.
13933 (vect_shift_permute_load_chain): Likewise.
13934 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13935 (vect_transform_slp_perm_load): Likewise.
13936 (vect_schedule_slp_instance): Likewise.
13937 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13938 (vectorizable_mask_load_store): Likewise.
13939 (vectorizable_bswap): Likewise.
13940 (vectorizable_store): Likewise.
13941 (vectorizable_load): Likewise.
13942 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13943 vec_perm_indices instead of auto_vec_perm_indices. Use
13944 tree_to_vec_perm_builder to read the vector from a tree.
13945 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13946 vec_perm_builder instead of a vec_perm_indices.
13947 (have_whole_vector_shift): Use vec_perm_builder and
13948 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13949 truncation to calc_vec_perm_mask_for_shift.
13950 (vect_create_epilog_for_reduction): Likewise.
13951 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13952 from auto_vec_perm_indices to vec_perm_indices.
13953 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13954 instead of changing individual elements.
13955 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13956 the vector in d.perm.
13957 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13958 from auto_vec_perm_indices to vec_perm_indices.
13959 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13960 instead of changing individual elements.
13961 (arm_vectorize_vec_perm_const): Use new_vector to install
13962 the vector in d.perm.
13963 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13964 Update vec_perm_builder constructor.
13965 (rs6000_expand_interleave): Likewise.
13966 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13967 (rs6000_expand_interleave): Likewise.
13969 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13971 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13972 to qimode could truncate the indices.
13973 * optabs.c (expand_vec_perm_var): Likewise.
13975 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13977 * Makefile.in (OBJS): Add vec-perm-indices.o.
13978 * vec-perm-indices.h: New file.
13979 * vec-perm-indices.c: Likewise.
13980 * target.h (vec_perm_indices): Replace with a forward class
13982 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13983 * optabs.h: Include vec-perm-indices.h.
13984 (expand_vec_perm): Delete.
13985 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13986 (expand_vec_perm_const): Declare.
13987 * target.def (vec_perm_const_ok): Replace with...
13988 (vec_perm_const): ...this new hook.
13989 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13990 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13991 * doc/tm.texi: Regenerate.
13992 * optabs.def (vec_perm_const): Delete.
13993 * doc/md.texi (vec_perm_const): Likewise.
13994 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13995 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13996 expand_vec_perm for constant permutation vectors. Assert that
13997 the mode of variable permutation vectors is the integer equivalent
13998 of the mode that is being permuted.
13999 * optabs-query.h (selector_fits_mode_p): Declare.
14000 * optabs-query.c: Include vec-perm-indices.h.
14001 (selector_fits_mode_p): New function.
14002 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14003 is defined, instead of checking whether the vec_perm_const_optab
14004 exists. Use targetm.vectorize.vec_perm_const instead of
14005 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14006 fit in the vector mode before using a variable permute.
14007 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14008 vec_perm_indices instead of an rtx.
14009 (expand_vec_perm): Replace with...
14010 (expand_vec_perm_const): ...this new function. Take the selector
14011 as a vec_perm_indices rather than an rtx. Also take the mode of
14012 the selector. Update call to shift_amt_for_vec_perm_mask.
14013 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14014 Use vec_perm_indices::new_expanded_vector to expand the original
14015 selector into bytes. Check whether the indices fit in the vector
14016 mode before using a variable permute.
14017 (expand_vec_perm_var): Make global.
14018 (expand_mult_highpart): Use expand_vec_perm_const.
14019 * fold-const.c: Includes vec-perm-indices.h.
14020 * tree-ssa-forwprop.c: Likewise.
14021 * tree-vect-data-refs.c: Likewise.
14022 * tree-vect-generic.c: Likewise.
14023 * tree-vect-loop.c: Likewise.
14024 * tree-vect-slp.c: Likewise.
14025 * tree-vect-stmts.c: Likewise.
14026 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14028 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14029 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14030 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14031 (aarch64_vectorize_vec_perm_const): ...this new function.
14032 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14033 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14034 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14035 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14036 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14037 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14038 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14040 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14041 check for NEON modes.
14042 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14043 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14044 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14045 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14047 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14048 the old VEC_PERM_CONST conditions.
14049 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14050 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14051 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14052 (ia64_vectorize_vec_perm_const_ok): Merge into...
14053 (ia64_vectorize_vec_perm_const): ...this new function.
14054 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14055 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14056 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14057 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14058 * config/mips/mips.c (mips_expand_vec_perm_const)
14059 (mips_vectorize_vec_perm_const_ok): Merge into...
14060 (mips_vectorize_vec_perm_const): ...this new function.
14061 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14062 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14063 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14064 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14065 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14066 (rs6000_expand_vec_perm_const): Delete.
14067 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14069 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14070 (altivec_expand_vec_perm_const_le): Take each operand individually.
14071 Operate on constant selectors rather than rtxes.
14072 (altivec_expand_vec_perm_const): Likewise. Update call to
14073 altivec_expand_vec_perm_const_le.
14074 (rs6000_expand_vec_perm_const): Delete.
14075 (rs6000_vectorize_vec_perm_const_ok): Delete.
14076 (rs6000_vectorize_vec_perm_const): New function.
14077 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14078 an element count and rtx array.
14079 (rs6000_expand_extract_even): Update call accordingly.
14080 (rs6000_expand_interleave): Likewise.
14081 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14082 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14083 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14084 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14085 (rs6000_expand_vec_perm_const): Delete.
14086 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14087 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14088 (altivec_expand_vec_perm_const_le): Take each operand individually.
14089 Operate on constant selectors rather than rtxes.
14090 (altivec_expand_vec_perm_const): Likewise. Update call to
14091 altivec_expand_vec_perm_const_le.
14092 (rs6000_expand_vec_perm_const): Delete.
14093 (rs6000_vectorize_vec_perm_const_ok): Delete.
14094 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14095 reference to the SPE evmerge intructions.
14096 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14097 an element count and rtx array.
14098 (rs6000_expand_extract_even): Update call accordingly.
14099 (rs6000_expand_interleave): Likewise.
14100 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14101 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14103 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14105 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14107 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14108 vector mode and that that mode matches the mode of the data
14110 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14111 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14112 directly using expand_vec_perm_1 when forcing selectors into
14114 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14116 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14118 * optabs-query.h (can_vec_perm_p): Delete.
14119 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14120 * optabs-query.c (can_vec_perm_p): Split into...
14121 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14122 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14123 particular selector is valid.
14124 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14125 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14126 (vect_grouped_load_supported): Likewise.
14127 (vect_shift_permute_load_chain): Likewise.
14128 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14129 (vect_transform_slp_perm_load): Likewise.
14130 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14131 (vectorizable_bswap): Likewise.
14132 (vect_gen_perm_mask_checked): Likewise.
14133 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14134 implementations of variable permutation vectors into account
14135 when deciding which selector to use.
14136 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14137 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14138 with a false third argument.
14139 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14140 to test whether the constant selector is valid and can_vec_perm_var_p
14141 to test whether a variable selector is valid.
14143 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14145 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14146 * optabs-query.c (can_vec_perm_p): Likewise.
14147 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14148 instead of vec_perm_indices.
14149 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14150 (vect_gen_perm_mask_checked): Likewise,
14151 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14152 (vect_gen_perm_mask_checked): Likewise,
14154 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14156 * optabs-query.h (qimode_for_vec_perm): Declare.
14157 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14158 (qimode_for_vec_perm): ...this new function.
14159 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14161 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14163 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14164 does not have a conditional at the top.
14166 2018-01-02 Richard Biener <rguenther@suse.de>
14168 * ipa-inline.c (big_speedup_p): Fix expression.
14170 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14173 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14176 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14180 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14181 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14182 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14183 cond_taken_branch_cost 3->4.
14185 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14187 PR tree-optimization/83581
14188 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14189 TODO_cleanup_cfg if any changes have been made.
14191 PR middle-end/83608
14192 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14193 convert_modes if target mode has the right side, but different mode
14196 PR middle-end/83609
14197 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14198 last argument when extracting from CONCAT. If either from_real or
14199 from_imag is NULL, use expansion through memory. If result is not
14200 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14201 the parts directly to inner mode, if even that fails, use expansion
14204 PR middle-end/83623
14205 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14206 check for bswap in mode rather than HImode and use that in expand_unop
14209 Copyright (C) 2018 Free Software Foundation, Inc.
14211 Copying and distribution of this file, with or without modification,
14212 are permitted in any medium without royalty provided the copyright
14213 notice and this notice are preserved.